DIABETES MELLITUS
from epidemiology to health policy
ClP-GEGEVENS KONINKLJJKE BlBLiOTHEEK, DEN HAAG
Ruwaard, Dirk
Diabetes mellitus: from epidemiology to health policy /
Dirk Ruwaard. - [S.l.: s.n.J.
Thesis Rotterdam. -With ref. - With summary in Dutch.
ISBN 90-9009749-X
Subject headings: diabetes / epidemiology / forecasts /
prospects / health policy
Cover:
'from data to scenarios', M.D. Comelissen-Kuyt, Katwijk
Printing: Drukkerij Elinkwijk B.V., Utrecht
© J 996 Ruwaard. No part of this publication may be reproduced, stored in a dataprocessing system or transmitted in any form by printing. photocopying, microfilming
or otherwise, without the prior permission of the author.
DIABETES MELLITUS
from epidemiology to health policy
DIABETES MELLITUS
van epidemiologie naar gezondheidsbeleid
Pl'oefschrift
ter verkrijging van de graad van doctor
aan de Erasmus Universiteit Rotterdam
op gezag van de rector magnificus
Prof.dr. P.W.C. Akkermans M.A.
en volgens besluit van het College voor Promoties.
De open bare verdediging zal plaatsvinden op
lVoensdag 2 oktober 1996 am 15.45 lIur
door
Dirk Ruwaard
geboren te Katwijk aan Zee
PROMOTIECOMMISSIE
Promotores:
Prof.dr.ir. D. Kromhout
Prof.dr. A.F. Casparie
Overige leden: Prof.dr. D.E. Grobbee
Prof.dr. P.J. van der Maas
Prof.dr. R.J. Heine
Co-promot~r:
Dr. H. Verkleij
Financial support by the Dutch Diabetic Research Foundation and the National Institute
of Public Health and the Environment for the publication of this thesis is gratefully
acknowledged.
Vaor Helina, Suzanne, Corne en Lotte
Voar mijn ouders
CONTENTS
I.
Introduction
9
2.
Diabetes mellitus: points of departure
19
3.
Increasing incidence of type I diabetes in the Netherlands:
the second nation-wide study among children under 20 years of age
41
4.
Is the incidence of diabetes increasing in all age groups in
the Netherlands? Results of the second study in the Dutch
Sentinel Practice Network
49
5.
Forecasting the number of diabetic patients in the Netherlands in 2005
63
6.
Forecasting the number of diabetic patients in the Netherlands in 2005:
an update
81
7.
Changing occurrence of diabetes mellitus and implications
for health policy: from a global to a national perspective
99
8.
General discussion
127
Summary
Samenvatting
Dankwoord
About the author
157
163
169
171
7
CHAPTER 1
Introduction
Introduction
EPIDEMIOLOGY AND HEALTH POLICY
This thesis deals with the interrelationship between epidemiology and health policy.
According to Rothman (I), the clearest of many definitions of epidemiology that have
been proposed has been attributed to Anderson, who defined epidemiology as "the
science of disease occurrence" (2). The primary task of epidemiological research is to
quantify the occurrence of illness in order to evaluate hypotheses about the causation
of illness and its sequelae, and to relate disease occurrence to characteristics of people
and their environment.
Health policy in the broadest sense relates to the actions of government, doctors and
other players who aim to maintain and improve the state of health of individuals and
the population. More specifically, a distinction can be made between policy designed
for health care and for prevention. Health care involves the organisation of diagnostics,
treatment, nursing and care ('cure and care'). Broadly speaking, it covers activities
directed at people who already have health problems. Prevention focuses on measures
and activities with the purpose of averting health problems. It may involve collective
measures to prevent specific diseases (vaccination programmes, screening
programmes), measures to promote health (health information and education), and
measures to improve safety (health protection, for example regulations on the safety of
food)(3).
Although the physicist Kelvin emphasized the importance of measurement in science,
his words also apply to the importance of epidemiology for health policy: "I often say
that when YOll can measure what you are speaking about, and express it in !lumbers,
YOll know something about it; but when you cannot e.'press it ill numbers, your
knowledge is of a meagre and IInsatisfactory kind; it may be the beginning of
knowledge, bllt YOII iIave scarcely, ill YOllr thollgilts, advanced to tile stage of Science,
whatel'er the matter may be" (4). Despite the fact that the choices that need to be made
in health policy are social or political in nature, the availability of data from
epidemiological research to express "it" (Le. health problems) in numbers should
actually playa cmcial role in unde,rpinning decision-m"aking.
The intenelationship between epidemiology and health policy is presented in Figure I.
It summarizes the policy cycle which consists of four steps (5). As Figure 1 shows,
epidemiology has two functions in the policy cycle:
I. to provide data which can be used in the preparation of new policy, and
2. to provide data which can be used to evalllate Cllrrent policy.
11
Chapter I
Figure I: The cycle of health policy alld the place of epidemiology ill it.
Source: adapted from Nota Gezondheidsbeleid, 1991 (5).
HEALTH POLICY AND HEALTH STATUS
The way health policy can influence health status can be illustrated by the conceptual
model devised when drawing up the document entitled 'Public Health Status and
Forecasts: the health status of the Dutch population over the period 1950-2010' (3).
This model, which will be extended in the next issue of the document due in 1997 (6),
has the following functions:
I. to provide a stmcture for the development of ideas;
2. to define boundaries;
3. to offer a structure for orderly handling of the subject matter;
4. to serve as a basis for producing a formal model in mathematical terms.
The conceptual model contains elements of previously published models (7-10) and is
shown in its basic form in Figure 2. The model shows that health status is defined by
detenninants and that it results in health care utilisation. Health status and the ensuing
health care utilisation (and costs) playa part in giving direction to health policy. By
means of prevention and/or health care, this policy aims to influence health status
12
Introduction
through the determinants of health. This dynamic process is affected by autonomous
demographic, sociocultural, economic and technological developments.
The structure of the conceptual model, which is discussed in more detail in the 'Public
Health Status and Forecasts' document, is essentially static. It describes a pattern of
effects in a qualitative sense. In fact we are interested in the dynamics of the system:
we want to know how the health status and the resultant burden on care facilities have
changed over time as a consequence of all those different influences. This is of great
importance for making jlltllre projections. Preparing policy and setting priorities not
only require information about the current situation but also about past and expected
future developments.
autonomous
developments
Figure 2: A cOJ1ceplflal model for public healtlt.
Source: Ruwaard et a!., 1995 (6).
THE CHOICE FOR DIABI,TES MELLITUS
The inspiration for choosing diabetes mellitus as the theme of this thesis came from the
scenario project entitled 'The future burden of chronic diseases for Dutch society:
scenarios for diabetes mellitus, cluonic non-specific lung diseases and rheumatoid
13
Chapter I
arthritis I 990-200S' (II-IS). This project started in 1988 and was performed on behalf
of the Steering Committee on Future Health Scenarios (a body set up in 1983 by the
State Secretary of the Ministry of Welfare, Health and Cultural Affairs) in order to
give an overview of anticipated future developments in the field of public health and
health care. Scenario projects had already been carried out for cardiovascular diseases
(16), cancer (17), accidents and traumas (18) and for ageing (19) and medical
technology (20).
Chronic diseases have been selected as the subject of a separate scenario study because
these diseases represent the major public health problem of our age, in the same way
as infectious diseases afflicted
Ollr
society in the last century. Chronic diseases may be
defined in various ways; generally these cover diseases from which people suffer
continuously or intermittently over a period of years. Due to their severity they involve
long-term disabilities and handicaps which impose a lasting burden on the health care
system (IS,21-23).
Chronic diseases are many and varied, which makes it impossible to lump them all
together. One of the first questions that therefore needs to be answered in conducting a
study on chronic diseases is to define which diseases are to be regarded as chronic, and
which of these are to be included in the study. Since scenario studies had already been
conducted for cardiovascular diseases and cancer, which account for the main part of
total mortality, it was clear that attention should also be directed towards a number of
chronic diseases that are less important in terms of mortality but which by contrast
contribute considerably to morbidity.
At the request of the Steering Committee, the TNO Institute for Prevention and Health
(formerly TNO Institute for Preventive Health Care) carried out a pilot study in order
to make a considered choice. On the basis of three criteria - occurrence, average
duration and severity/intensity of care - researchers at this institute recommended that
scenario studies be set lip in the field of diabetes mellitus and chronic non-specific
lung disease. As a third disorder, the choice was narrowed down to rheumatoid arthritis
or multiple sclerosis (24,2S). The Steering Committee ultimately chose rheumatoid
arthritis because of its more frequent occurrence.
The 'Chronic Diseases' scenario study posed two sets of questions:
I. What are the likely trends in the number of patients, the number and severity of
complications, the quality of life and the degree of care/self-care in response to
changes in determinants (risk factors), developments in medical technology,
14
Introduction
changes in the structure of care and treatment, and demographic and sociocultural
developments; and what implications will this have in tenns of the demands made
on health care over the next ten to fifteen years?
2.
What means do the government or other organizations and groups in society have
to influence these developments and what possible effect might these have?
Given the rapid rate of change in this field and the large degree of uncertainty that
must be taken into account, it became clear that it would not be possible to make
considered judgments for a period of more than ten to fifteen years. For this reason the
scenarios have not been taken beyond the year 2005.
This thesis is restricted to diabetes mellitus and deals only with developments in the
occurrence of the disease. After finishing the scenario study for diabetes, research
projects were started that provided insight into trends in the occurrence of diabetes
mellitus. Results from the scenario study and in particular from the consecutive
research activities are presented in this thesis.
AIM OF THE STUDY
The overall objective of the studies described in this thesis is to elucidate the
occurrence of diabetes mellitus in the Netherlands and its development over time. As
diabetes mellitus is a major cause of prolonged ill health and premature mortality
which requires a substantial amount of health care, the implications of the changing
OCCUITence for health policy are raised as well. More specifically, the study aims to
answer four main questions:
I. what is the occurrence of diabetes mellitus?
2. has the occurrence of diabetes mellitus changed in recent years?
3. what are possible future developments in the occulTence of diabetes mellitus?
4. what are the likely implications of these developments for health policy?
STRUCTURE OF THE THESIS
The four questions that need to be answered serve as a guideline for the stmcture of
this thesis. The poi/lts of departllre are dealt with in Chapter 2, which gives a general
description of the disease and the concepts employed. It also substantiates the choices
that need to be made to ensure that the most suitable sources are lIsed for incidence,
15
Chapter I
prevalence, remission and mortality data. These choices are based on the available
information at the time the scenario project was conducted (the first question in the
thesis).
The second question is dealt with in Chapters 3 and 4, which focus on changes in the
occurrence over a ten-year period. Chapter 3 describes the changes in the incidence of
insulin-dependent diabetes mellitus among children under 20 years of age. The findings
are based on the first (1978-1980) and second (1988-1990) nation-wide retrospective
studies covering the total Dutch population. Chapter 4 addresses the assessment of
possible changes in the incidence of diabetes mellitus in all age groups in the
Netherlands in the period 1980-1983 and 1990-1992. It describes the results of a large
registration network of sentinel stations (the Dutch Sentinel Practice Network)
consisting of about I % of the Dutch population.
Possible fllfllre developments are presented in Chapters 5 and 6 (the third question). In
Chapter 5 projections of the number of patients are based on the information provided
by the scenario project (background study). In addition, we updated the projections in
Chapter 6 by using the 'trend' data from Chapters 3 and 4.
Finally, Chapter 7 addresses the fourth question by paying attention to the possible
implications of these findings for health policy. The general discussion (Chapter 8) is
followed by a summary.
REFERENCES
I.
2.
3.
4.
5,
6.
7.
16
Rothman KJ. Modern Epidemiology. Bostonrroronlo: Lillie, Brown and Company, 1986.
Cole P. The evolving case-control study. J Chron Dis 1979; 32: 15·27.
Ruwaard D. Kramers PGN, Berg JClhs A van den, AcJuerberg PW. Public Status and Forecasts: the
health slatus of the Dulch population over the period 1950-2010. National Institute of Public Health and
Environmental Protection. TIle Hague: Sdu Uilgeverij, 1994.
Beiser A. The World of Physics. New York: McGraw-Hili, 1960.
Nola Gczondheidsbeleid 1992. Gezondheid met beleid. Tweede Kroner, vergaderjaar 1991·1992 Session,
22459, fif. 1. 's-Gravenhage: Sdu Uitgeverij, 1991.
Ruwaard D, Berg Jelhs A van den, Jansen J. Kramers PON, Giessen A van der. Genugten J\1LL. Gijsen
R. Harteloh PPM, Maas lAM, Poos MJJC, Vcrklcij H. Achterberg PW. Definitie voor de opzel van de
studie Volksgezondheid Tockomst verkenning 1997. Rapport nr. 431501013. Bilthoven: Rijksinstituut
voor Volksgezondhcid en Milieuhygi{!ne, 1995.
Lalonde M. A new perspective on the health of the Canadians. Onawa: Ministry of National Health and
Welfare. 1974.
Introduction
Nota 2000. Over lie ontwikkcling van gezondhcidsbeleid: reiten, bcschouwingcn en beleidsvoome-mcns.
Tweede kroner, vcrgaderjaar 1985-1986, 19500. nrs. 1-2. Rijswijk: Miniserie van WYC, 1986.
9. Stuurgroep Tockomstscenruio's Gezondheidszorg. Scenario's in de Yolksgezondheid: Inleiding in de
Illcthodick van de STG. Utrecht: J(Ul van Arkc1. 1989.
10, Dijk FJH van, Donnolen M van, Kompier MAJ, Meijman TF. Herwaardcring model belastingbelastbaarheid. T Soc Gezondheidsz 1990: 68: 3-10.
11. Verkleij H, Casparie AF, Ruwa...'lfd D, Kromhoul D, Vclde UK van del'. Toekomslscenruio-sludie van
8,
start gegaan. Diabetes mellitus, CARA, reumatordc arthritis. Medisch Contact 1989: 44: 438·440.
12. Steering Committee on Future Health Scenarios. Chronic Diseases in the year 2005. Volume I:
Scenarios on Diabetes Mellitus 1990·2005. Dordrechl/Boston/London: Kluwer Academic Publishers,
199J.
13. Steering Committee on Future Health Scenarios. Chronic Discases in the year 2005. Volume 2:
Scenarios on Chronic Non-Specific Lung Diseases 1990·2005. DordrechlJ BostonILondon: Kiuwer
Academic Publishers, 1993.
14. Sleering Committee on Future Health Scenarios. Chronic Diseases in the year 2005. Volwne 3:
Sccnarios on Rheumatoid Arthritis 1990·2005. Dordrecht/Boston/London: Kluwcr Academic Publishers.
1994.
15. Sluurgroep Toekomslscencuio's Gezondheidszorg. Chronische ziekten in het ja.'U' 2005. Deel 4:
Scenario's voor bcleid. Houten/Antwcrpen: Bolm Staflcu van Loghum, 1992.
16. Steering Committee on Future Health Scenarios. The heart of the future - the future of the he..mh.
Scenarios on cardiovascular disease in the period 1985·2010. volume I and 2. Dordrecht/BostOn/Lancaster: Martinus Nijhoff Publishers, 1987.
17. Steering Committee on Future Health Scenarios. Canccr in the Nethcrlands, volume I and 2. Dordrccht:
Kluwer, 1988.
18. Steering Committee on Future Health Scenarios. Accidents in the year 2000, volume I and 2.
Dordredlt/BostonjLondon: Kluwer Academic Publishers, 1989.
19. Steering Committee on Future Health Scenarios. Growing old in the future. Scenarios on health and
ageing in the period 1984-2000. Dordrecht: Kluwer, 1987.
20. Steering Committee on Future Health Scenarios. Anticipating and assessing health care technology,
volume 1-8. Dordrecht/BostonjLancaster: Martinus Nijhoff Publishers, 1987.
21. Voorn ThB. Chroni:;;chc ziekten in de hui$artpraktijk. Proerschrift. Nijmegen: Katholieke Universiteit
Nijmegen, 1983.
22. Bos GMf van den. Zorgen van en voor chronisch ziekcn. Proefschrift. Universiteit van Amsterdam.
Utrecht-Antwerpen: Bohn, Scheltema en Holkcma, 1989.
23. Casparie AF. Gezondheidsonderloek bij cnronische ziekten. Parruneters van kwaliteit van zorg. T Soc
Gezondheidsz 1991; 69: 242-245.
24. Dftvidse W, Watcr HPA van de, Vaandrager GJ. Toekomstscenario's chronische ziektcn - voorondcrzoek. Leiden: NIPG-1NO, 1987.
25. Davidse W, Water HPA van de, Vaandmger OJ. Sclcctie van chronische patienten voor scenario·
onderzoek. Mcthodologische keuzeproblemen. Medisch Contact 1988: 43: 493·494.
17
CHAPTER 2
Diabetes mellitus: points of departure
'General description of the disease' (first section) is based on the manuscript:
Dirk RlIwaard, Edith J.M. Feskells. Diabetes mellitlls. III: RlIwaard D. Kramers PGN
(red.) VolksgezolldheidToekomst Verkellllillg. De gezolldheidstoestalldvall de Nederlalldse
bel'olkillg ill de periode 1950-2010. RijksillStifllllt 1'00r Volksgezolldheid ell Miliellhygielle.
Dell Haag: Sdll Uitgeverij, 1993 1'1'.303-308.
'Concepts and approach' (second section) and 'Current knowledge of incidence,
prevalence, remission and mortality' (third section) are based on the report:
Steerillg Committee 011 FlIflIre Health Scellarios. Chrollic diseases ill the year 2005.
Voillme /, Scellarios 011 diabetes melliflls /990-2005. DordrechtlBostolllLolldoll: Killwer
Academic PlIblishers, 1991.
Diabetes mellitus: points of departure
ABSTRACT
This chapter describes the principles underlying this thesis. As the subject is diabetes
mellitus, it starts by providing a general description of the disease. The (changes in)
diagnostic criteria and classification, its symptoms and course, and the determinants are
briefly reviewed.
Secondly, the conceptual model employed and the stepwise approach selected to answer
the questions which are dealt with in this thesis are the subjects of the section 'Concepts
and approach'. There are two major prerequisites for constructing sound future projections
for health policy purposes: the transcription of the conceptual model into a dynamic
(mathematical) model and the availability of data.
Finally, the section 'Current knowledge of incidence, prevalence, remission and mortality'
focuses on the availability of epidemiological data on the occurrence of diabetes mellitus.
It substantiates the choices made to ensure that the most suitable sources were used for
incidence, prevalence, remission and mortality (available at the time we conducted the
scenario study). There are two reasons why this information is included in 'the points of
departure'. It is not only the first step in the stepwise approach (performing a background
study to gain insight into the occurrence of diabetes). In addition, the scenario study also
revealed a lack of recent representative data. This observation was the starting-point for
further research activities to provide insight into trends in the occurrence of diabetes.
21
Chapter 2
GENERAL DESCRIPTION OF THE DISEASE
Diagnostic Cl'iteria and classification
Diabetes mellitus is a chronic metabolic disorder. It is associated with excessive levels of
blood glucose, on which the diagnosis is based. In 1985 the World Health Organization
(WHO) defined cutoff values above which diabetes mellitus is regarded as being present
(I). These values for diabetes mellitus (and for impaired glucose tolerance and normal
glucose tolerance) according to the oral glucose tolerance test (OGTT) are presented in
Table I. The glucose values differ, depending on the glucose load (fasting or 2 hrs after
a 75 g glucose load), where the blood sample is taken (i.e. from a capillary or vein) and
whether the glucose level is measured in whole blood or in plasma.
Table 1,' Diagllostic values for the oral glucose to/erallce lest.
Glucose concentration, mmol/litre
Whole blood
Diabetes mellitus
Fa'\ting value
2 hrs valuea
Plasma
Venous
Capillary
Venous
Capillary
'? 6.7
'?1O.0
'? 6.7
'?Il.l
'? 7.8
'?Il.l
'? 7.8
'?12.2
< 6.7
6.7-10.0
< 6.7
7.8-1l.l
< 7.8
7.8-1l.l
< 7.8
8.9-12.2
< 6.7
< 6.7
< 6.7
< 7.8
< 7.8
< 7.8
< 7.8
< 8.9
Impaired glucose tolerance
Fa"iting value
2 hrs valuea
Norma! glucose tolerance
Fasting value
2 hrs valuea
a: two hours after a 75 g glucose load under standardized conditions.
Source; World Health Organization, 1985 (1).
The OGTT is often used in epidemiological research (or during pregnancy). However, in
clinical practice the diagnosis diabetes is usually based on the presence of the classic
diabetic symptoms (polyuria, hunger, thirst, weight loss, tiredness, dizziness, drowsiness
or, in extremis, coma) combined with a single abnonnal blood glucose level or on two
22
Diabetes mellitus: points of departure
abnormal levels without complaints measured on different occasions. An abnormal level
is defined as a capillary blood glucose fasting value equal to or exceeding 6.7 mmol/litre
(although a fasting state is difficult to prove) and/or a randomly measured capillary blood
glucose value equal to or exceeding 11.1 mmol/litre. If the blood glucose levels are
inconclusive, an OOTT is recommended. Instead, a high carbohydrate breakfast test is
often used in everyday practice.
The above-mentioned 1985 cutoff values used to diagnose diabetes mellitus were adapted
from the 1980 criteria (2). In 1980 the cutoff values were rounded to the nearest mmol/
litre, whereas in 1985 they were rounded to the nearest tenth of a mmol/litre. The first
time the WHO published diagnostic criteria was in 1965 (3). Depending on age, the
diagnosis could be established two hours after a 50 g or 100 g glucose load in the case
of a blood glucose value equal to or exceeding 7.2 or 7.8 mmol/litre measured in venous
whole blood or capillary whole blood, respectively. If these criteria are compared with the
current ones (established in 1985), the fonner are less strict: the diagnosis 'diabetes
mellitus' based on the WHO 1965 criteria is assumed to correspond roughly to the
diagnosis 'diabetes mellitus including impaired glucose tolerance' according to the WHO
1985 criteria (4,5). When interpreting trend data on the occurrence of diabetes, it is
essential to be aware of these changes in diagnostic criteria.
There are various forms of diabetes mellitus. The classification of diabetes mellitus and
associated categories of glucose intolerance adopted by the WHO in 1985 is given in
Table 2. The t)VO most common forms are insulin-dependent diabetes mellitus (IDDM) and
non-insulin-dependent diabetes mellitus (NIDDM). IDDM represents about 10-20% and
NIDDM about 80-90% of all diabetic patients (6,7).
Instead of IDDM and NIDDM the older terms type I and type II diabetes are still widely
used. Although IDDM and NTDDM are clinically descriptive subclasses and type I and
type Il represent different pathogenetic mechanisms, IDDM and type I diabetes on the one
hand and NIDDM and type IT diabetes on the other were regarded as completely
synonymous (Le. carrying no etiological or pathogenetic implications)(l). In clinical
practice and for research purposes the pragmatic assumption is sometimes made that
diabetes diagnosed before the age of 20-30 will be mainly IDDM and above 20-30 years
mainly NIDDM (amongst others, 8).
Recently, the classification presented in Table 2 has been disputed. As a result of
increasing knowledge about the different causes of diabetes, it seems likely that further
refinement or revision of the classification will soon be possible. Therefore, it is no longer
23
Chapter 2
recommended to use the terms IDDM and NIDDM as synonymous with type I and type
II diabetes, respectively (9). For instance, some patients clinically diagnosed as NIDDM
appeared to have type I diabetes according to the pathogenetic process (Latent Autoimmune Diabetes in Adults; LADA)(IO). [n addition, it has long been observed that
NIDDM can also be subdivided into several forms. One subtype of NIDDM in adolescence and in young adults shows a strong autosomal dominant inheritance pattern (11). The
term maturity-onset diabetes of the young (MODY) is used to describe this subtype, which
in itself is genetically heterogeneous (12-15).
Table 2: Classification 0/ diabetes mellitus alld associated categories of glucose illtolerance.
A.
Clinical classes
Diabetes mellitus (DM)
- Insulin-dependent diabetes mellitus (lDDM)
- Non-insulin-dependent diabetes mellitus (NIDDM)
(a) Non-obese
(b) Obese
- Malnutrition-related diabetes mellitus (MRDM)
- Other types of diabetes associated with certain conditions and syndromes:
(I) pancreatic disease; (2) disease of hormonal etiology; (3) drug-induced or
chemical-induced conditions; (4) abnormalities of insulin or its receptors; (5)
certain genetic syndromes; (6) miscellaneous.
Impaired glucose tolerance (IGT)
(a) Non-obese
(b) Obese
(c) Associated with certain conditions and syndromes
Gestational diabetes mellitus (GDM)
B.
Statistical risk classes (subjects with normal glucose tolerance but a substantially increased
risk of developing diabetes)
Previous abnormality of glucose tolerance
Potential abnonnality of glucose tolerance
Source: World Health Organization, 1985 (1).
24
Diabetes mellitus: points of departure
Symptoms and course
Patients suffering from the metabolic disorder diabetes mellitus often exhibit symptoms
such as frequent urination, heavy eating and drinking. weight loss, tiredness and dizziness.
In particular in IDDM, coma - caused by acidification and excessive blood glucose levels
combined with dehydration - can sometimes be the first sign of the presence of the
disease. Over time, multiple chronic complications may occur, resulting from damage to
large and small blood vessels and nervous tissue. This can lead to complications such as
myocardial infarction, stroke, circulation disorders in the legs, blindness, kidney diseases
and loss of sensitivity and/or pain in the limbs (see Table 3). The epidemiological aspects
of the acute and chronic complications of both IDDM and NIDDM are reviewed in the
scenario report (16).
In patients with NIDDM, chronic complications may also be found at the time of
diagnosis, or the complications may actually be the reason for suspecting the presence of
diabetes (8,17-20). This is partly due to the fact that a large number of NIDDM patients
(at least 50%) are not known to be carrying the disease and are therefore as yet
undiagnosed (5,21-24). The occurrence of the symptoms and complications can severely
reduce life expectancy and the quality of life (not only in physical tenl1S but also as
regards mental and social consequences) and is responsible for a substantial degree of
health care utilisation in Dutch society. In the Netherlands the health care costs related to
diabetes are estimated to be about 1% of total health care costs. However, this
underestimates the real costs, as it only represents the costs incurred when diabetes is the
primary diagnosis (16).
Table 3,' Acute alld chronic complications arisillg from diabetes mellitus.
acute complications:
ketoacidosis/ketoacidotic coma
hyperosmolar non-ketoacidotic coma
hypoglycaemia/hypoglycaemic coma
chronic complications:
coronary heart disease
cerebrovascular accident
peripheral vascular disease
diabetic foot
neuropathy
nephropathy
retinopathy
25
Chapter 2
Determinants
lDDM arises in genetically susceptible individuals who are exposed to putative environmental or exogenous triggers that may activate immunological mechanisms, leading to a
progressive loss of pancreatic islet beta cells (25-27). The excessive glucose concentration
stems from a lack of the hormone insulin (insulin deficiency), which is produced by the
beta cells. The immunological and inflammatory mechanisms concerned have not yet been
clearly defined. It is an insidious process which may occur over many years. During the
'pre-diabetic' stage of the evolution of the disease, individuals can often be recognized by
the presence of immunological markers and a decline in pancreatic beta-cell function (28).
Studies callied out on healthy children in the Dutch population have shown that in the
presence of certain immunological markers. 50% of the children concerned develop lDDM
within eight years (29).
The fact that hereditary factors playa role is apparent from studies of identical twins.
which demonstrate a higher concordance rate for lDDM in monozygotic twins (25-40%)
than in dizygotic twins (5-10%) (30-32). In addition. the overall risk of lDDM among
whites in the United States of America is 0.2-0.4%, while the risk in siblings of probands
with IDDM is about 5% and in the offspring of diabetic parents 2-3% (if the mother has
the disease) and 5-6% (if the father has the disease)(32). The major genetic predisposition
is conferred by genes located on the short anTI of chromosome 6, either within or in close
approximation to the major histocompatibility complex (33.34).
Viruses, such as mumps, rubella and coxsackie, are suspected exogenous determinants
(35). It is thought that incidence peaks noted by some observers in autumn and early
winter could be explained by the presence of vimses (36). In addition, there are
indications that diet may also playa role. For example, breast-feeding could have a
protective effect. while a high level of consumption of protein-rich foods and carbohydrates or nitrosamine-containing foods could increase the risk of lDDM (37-39). However.
a recently published meta-analysis of 17 case-control studies showed that the increased
risk of lDDM associated with early infant diet exposures is small and may be explained
by methodological shortcomings (40). Quantification of the relative contribution of the
different determinants has so far not been possible.
Primarily. IDDM and NIDDM are clinically descriptive subclasses of diabetes. Whereas
lDDM appears to be the result of an auto-immune disease process (i.e. a type I
pathogenetic process). the etio'ogy and pathogenesis of NlDDM is heterogenous.
26
Diabetes mellitus: points of departure
Many patients with NIDDM and individuals with an impaired glucose tolerance (IGT)
exhibit insulin resistance and hyperinsulinaemia in association with dyslipoproteinaemia.
central obesity and hypertension. This cluster of cardiovascular detenninants has been
described by a number of names such as 'syndrome X' or 'Reaven syndrome' (41), the
chronic metabolic syndrome, and the insulin resistance syndrome (42). The extent to
which this cluster of determinants represents a single disease process is still unclear (43).
Persons with a history of IGT and gestational diabetes are at increased risk of developing
NIDDM. The rate of progression from IGT to NIDDM is about 2-3% per year in studies
carried out in the UK and the USA, and the incidence of NIDDM in women with
gestational diabetes is about 3-5% per year (44,45).1
Whereas IDDM is most commonly encountered at a younger age « 20-30 years), NIDDM
is a form of diabetes particularly associated with advancing age. Apart from age, heredity
is currently considered as an important risk factor associated with NIDDM, apparently
playing an even greater role than it does in IDDM. In the case of identical twins, a
concordance rate of 95-100% for NIDDM can be found (30). The risk of developing
NIDDM in individuals where one or both parents have NIDDM is almost three times as
great as in those whose parents are free of the disease (47).
Irrespective of the preseQ.ce or absence of the disease in the family, ovenveight increases
the risk of the occurrence of NIDDM by a factor of two to three (48). There are
indications that people must be overweight for some time before it becomes a risk factor
for diabetes (49,50). Also the distriblltion offat around the body is an important factor.
Abdominally localised body fat (a 'paunch') imposes an additional risk (51).
It has been demonstrated that physical inactivity promotes the development of NIDDM
(44,47). The composition of the diet is also a risk factor. In line with the 'Guidelines for
a healthy diet' published by the Netherlands Nutrition Council in 1986 (52), the
consumption of high-fibre foods and unsaturated fatty acids at the expense of foods rich
in saturated fatty acids is encouraged to prevent NIDDM (48). To what extent alcohol lise
and smoking are independent determinants for NIDDM is still unclear. The results of
several studies on alcohol use (53-58) as well as smoking (56-61) are contradictory. Table
4 summarizes the relative risks of established determinants for NIDDM.
I
In addition, there are indications that 10-20% of NIDDM is the result of an auto-immune disease process as found in
lDOM (Latent Auto-immune Diabetes in Adults) and that 2-4% is the result of specific gene mutations (46).
27
Chapter 2
Table 4: Relative risks of established determillallls for NlDDM.
Determinant
Indicator
RR
Reference
I or 2 parents
wilh diabetes
2.9
47
BM] >25kg/m'
waist/hip ratio> I
2.5
2.5
48
51
energy expenditure
1.7
47
0.7
48
Endogenous:
genetic factors
body weight
Exogenous:
physical activity
<2,000 kcaVweek
nutrition
Guidelines for a
healthy diet
Bl\.fI: Body Mass Index (weight (kg)/height 2 (m»
Note: for the risk of NlDDM in those wHh IGT and gestational diabetes, see text.
CONCEPTS AND APPROACH
The ultimate question to be addressed in this thesis is: what are the likely implications of
future developments in the occurrence of diabetes mellitus for health policy (Chapter 7)?
Future developments may be explored in various ways. In the Netherlands the scenario
method has been applied extensively in recent years. By drawing a distinction between
likely developments on the one hand and the potential for influencing these developments
on the other, an attempt has been made to provide a structure for the wide range of future
possibilities.
In a scenario study, a number of possible images of the future are drawn up with the aid
of scenarios. In general, a scenario can be defined as follows: "A scenario describes how
the present situation for a part of society may be changed into a particular image of the
flllllre by means of a series of possible developmellls. The aim in doing so is to obtain
greater illsight illto the underlying mechanisms and the ways ill which these call be
inflllenced" (62).
A distinction is often drawn between exploratory and strategic scenarios. A strategic
scenario takes a particular (desired) image of the future and explores the various strategic
paths in which this might be achieved. In an exploratory scenario the future is studied by
examining a number of different possible images of the future. In this thesis the future is
28
Diabetes mellitus: points of departure
confined to exploratory scenarios with regard to the occurrence of diabetes until 2005
(Chapters 5 and 6).
The two major prerequisites for constructing sound future projections for health policy
purposes are: (I) the transcription of the conceptual model into a dynamic (mathematical)
model and (2) the availability of epidemiological trend data.
From a conceptual to a mathematical model
The conceptual model shown in Figure I describes the interrelationships between the
epidemiological concepts incidence, prevalence, remission and mortality or life expectancy. Table 5 summarizes the definitions used.
incidence
~
remission
~
prevalence of
diabetes
------.. mortaHtyllife expectancy
Figure I.' Prevalence of diabetes mellitus as a result of incidence, remission alld mortalityllife
e.\pee/alley.
Figure I indicates that the number of patients with diabetes mellitus (prevalence) is the
net result of an inflow and outflow of patients. The inflow is equal to the incidence, while
the outflow is the sum of remission and mortality. These concepts can be interrelated by
using such a conceptual model, which describes the pattern of effects in a qualitative
sense. But what we are ultimately interested in is the dynamics of the system: we want
to quantify the number of diabetic patients projected for the period up to 2005, taking into
account several influences, some of which may be disease specific (e.g. changes in
exposure to risk factors) while others are autonomous (e.g. demographic developments).
The more we know about potential developments, the more relevant it is to reshape the
conceptual model into a dynamic model in which the relationships between the concepts
as defined in Table 5 are presented in formal mathematical terms (Chapters 5 and 6).
29
Chapter 2
Table 5: Definitions of concepts llsed in relalioll to diabetes mellillls.
prevalence:
the number of diagnosed diabetic patients at a moment in time in a defined
population (e.g. expressed per 10,000 persons)
incidence:
the number of newly-diagnosed diabetic patients per unit of time in a defined
population (e.g. expressed per 10,000 persons per year)
remission:
the chance of recovery from diabetes mellitus per unit of time (e.g. per year)
mortality:
the risk of patients with diabetes mellitus dying per unit of time (e.g. per year)
life expectancy: the number of years before dying from age at onset of diabetes mellitus
Availability of data
To judge what sources will be used to obtain the necessary data to make projections, every
scenario study begins with a background study resulting in a description of the present
situation (see next section in this chapter). In addition, this background study is also
designed to identify gaps in information needed to make accurate projections.
One of the
information
population.
(Chapters 3
problems arising from the background study for diabetes was the lack of
on trends in the occurrence of diabetes representative for the Dutch
Therefore, additional studies were conducted to deal with this problem
and 4).
In scenario studies extra information is often collected by means of a Delphi study to
identify the factors that will determine tWllre developments. A systematic inventory of
future expectations is compiled by means of repeated rounds of questionnaires sent to
experts with interim anonymous feedback reports. While a Delphi study provides the
building blocks for future scenarios, these blocks then need to be combined in an
acceptable manner. In order to check for consistency - which is an indispensable element a workshop was convened to bring together some of the participants in the Delphi study.
The results of the Delphi study and the workshop provided information for Chapters 5 and
6 of this thesis. The stepwise approach used to answer the questions in this thesis is
summarized in Table 6.
30
Diabetes mellitus: points of departure
Table 6: Stepwise approach.
Step
Question to be answered
Chapter
1. Background study
What is the occurrence of diabetes mellitus?
23
2. Two 'trend' studies
Has the occurrence changed in recent years?
3 and 4
3. Drawing up scenariosb
What are possible future developments?
5 and 6
4. interpretation
What are the likely implications for health policy?
7
a: sec next section in this chapter.
b: with the aid of a mathematical model and additional inrormation from the Delphi Study.
CURRENT KNOWLEDGE OF INCIDENCE, PREVALENCE, REMISSION AND
MORTALITY
The background study provided a description of current knowledge about the occun'ence,
i.e. the incidence, prevalence, remission and mortality of diabetes mellitus in the
Netherlands. The findings were also compared with other countries. The study was
completed in May 1989 and is briefly reviewed in this section with respect to the
OCCUITence of diabetes mellitus in the Netherlands. The results have been described in
detail in previous publications (16,63).
In 1988 a search revealed that the incidence andlor prevalence of diabetes mellitus in the
Netherlands in the period 1971-1987 had been studied in 18 surveys. In terms of the
method of data collection the surveys can be divided into four categories:
I.
morbidity registration by general practitioners (9 surveys);
2.
questionnaire forming part of population survey (6 surveys);
3.
questionnaire submitted to internists and paediatricians (I survey);
4.
screening survey for glucose tolerance in general practice (2 sUlveys).
As we were particularly interested in the burden on health care in the Netherlands, the
following basic principles were applied to choose the most appropriate sources: the data
should represent clinically-known patients and not those as yet undiagnosed, and they
should be representative for the Dutch population as a whole in terms of age, gender,
degree of urbanization and geographical variation. The incidence among 0-19 year-aids
was most reliably represented by the questionnaire survey conducted among all Dutch
31
Chapter 2
paediatricians and internists in the period 1978-1980 (64,65). The Dutch Sentinel Practice
Network, which represents about one percent of the Dutch population as a whole, was
used to estimate the incidence from age 20 onwards (recorded in the period 1980-1983)
and the prevalence for all age groups (recorded in 1980)(66). Due to the lack of genderspecific figures in the Sentinel Practice Network, additional information about the sex
ratios as identified in the continuous morbidity registration of the Nijmegen University
General Practitioners Institute was used (63).
Table 7 presents the incidence, expressed as an annual average, and the prevalence
according to age and gender. The table reveals that both the incidence and prevalence
increase with age and that the prevalence is higher among women than among men. The
table also shows the absolute number of patients with diabetes mellitus, calculated on the
basis of the size and composition of the population in 1980 (67). The number of clinicallyknown diabetic patients appeared to be 191,000 in 1980, which corresponded to a
prevalence of 1.35% of the population. The annual number of newly-diagnosed patients
amounted to 17,300, which corresponded to an incidence of 0.12% in 1980.
Table 7: llIcidellce alld prevalellce of diabetes melliflls per 10,000 illitabitallls alld the absolllle
number of patients with diabetes mellillls ill the Netherlands according to age alld sex ill 1980.
Age
Incidence
men
women
Prevalence
men women
Absolute number
men
women
" 65
0.7
1.2
1.4
1.2
1.4
3.3
7.0
20.8
26.1
49.9
0.6
1.0
1.5
0.9
1.6
2.7
6.5
18.7
26.1
52.7
2
2
7
7
10
25
46
13l
280
599
2
2
14
21
31
25
64
127
342
728
88
108
431
440
605
3,048
4,107
9,859
17,615
40,104
84
104
833
1,276
1,797
2,838
5,317
9,537
23,378
69,385
Total'
11.3
13.2
109
161
76,375
114,549
0- 4
5- 9
10-14
15-19
20-24
25-34
35-44
45-54
55-64
a:
the total figures for the incidence and prevalence have been standardized by age on the basis of the size and
composition of the population in 1980 (67).
Source: Steering Committee on Future Health Scenarios (16).
32
Diabetes mellitus: points of departure
The interpretation of these incidence and prevalence figures is subject to a number of
problems:
l. The incidence and prevalence figures relate to all forms of diabetes mellitus. It may
be assumed that by far the largest contribution is made by lDDM (10-20%) and
especially NlDDM (80-90%)(6,7).
2. The data do not permit a classification into lDDM and NIDDM over the entire age
range. Nor do the other surveys allow any distinction to be made in this respect. Only
the survey among paediatricians and internists provides information on the incidence
of lDDM < 20 years (64,65).
3. A comparison with the other surveys indicates a certain degree of dispersion in the
estimates, which needs to be taken into account when making projections.
4. Depending on the quality of the registration, underreporting needs to be taken into
account (Le. the degree of meticulousness with which general practitioners in fact
record patients diagnosed as having diabetes mellitus).
5. The morbidity patterns registered in general practice specifically reflect the health
problems presented by those who make an appeal to the health care system. As the
general practitioner operates as a 'gatekeeper' in the Netherlands, most problems (e.g.
diabetes) will be detected. However, people living in institutions for a long time (e.g.
nursing homes) and relying on institutional doctors will be missed. This may result in
an underestimation of the incidence and prevalence when diabetes is relatively more
prominent among such people.
6. A large American survey (1976-1980) among 20-74 year-olds indicated that only 50%
of all patients with diabetes mellitus were diagnosed as such (21). When our search
was conducted, there were indications to suggest that there is a high prevalence of
undiagnosed glucose metabolism disorders among individuals aged over 65 years in
the Netherlands as well (22). Recently, this was confirmed by several other studies
(5,23,24).
The extent to which remission (in addition to mortality) makes a significant contribution
to the outflow of the diabetic population is unknown. Nor is it clear to what extent
reported cases of remission are due to misclassification (e.g. instances in which it later
turns out that a patient has been diagnosed as having diabetes mellitus on the basis of
inadequate data). In so far as remission takes place this will be confined to NlDDM. It is
possible for overweight patients with NIDDM to achieve an improvement - i.e. a drop in the blood glucose level by losing weight. Despite the remission, these 'patients' will
remain under a certain degree of medical supervision. For this reason, it has been decided
in this study to assume that no remission takes place.
33
Chapter 2
Assuming no remISSIOn, the outflow from the diabetic population wiII be wholly
determined by lIIortality. In the Netherlands, statistics on the causes of death are compiled
by Statistics Netherlands (formerly the Central Bureau of Statistics). On the basis of the
International Classification of Diseases, both the underlying (primary) cause of death and
other diseases that could have contributed towards death (i.e. secondary causes) are
recorded. The sum of diabetes mellitus as primary and secondary cause of death can give
an impression of mortality. There is however need for caution. On the one hand, death
certificates are mainly concerned with the primary cause of death. On the other hand,
multi-morbidity - especially at a later age - can mean that chronic diseases (including
diabetes mellitus) are not sufficiently registered as secondary causes of death.
When this background study was can'ied out, no other studies had been conducted in the
Netherlands to examine the underreporting of diabetes mellitus as primary and secondary
cause of death. Studies in other countries revealed that in the case of deceased patients
with diabetes mellitus, the disease was not mentioned at all on the death certificate in
between 25-77% of all cases (68). Recently, it has become evident that in the Netherlands
only about 50% of ali registered cases of diabetes on the death certificates were coded in
the statistics with diabetes specified as cause of death (69).
Another means of determining the outflow from the diabetic population resulting from
mortality is based on the reduction of life expectancy of patients with diabetes mellitus
in relation to the total population. A number of sUlveys in other countries have indicated
that the younger the age at which diabetes mellitus is diagnosed the greater the reduction
of life expectancy will be. This is shown in Figure 2 (70).
In this respect it should be noted that:
I. The spread in reduction of life expectancy falis the later the age at which diabetes
mellitus is diagnosed.
2. No distinction can be made between IDDM and NlDDM.
3. Whether or not men and women with diabetes mellitus suffer the same reduction of
life expectancy is disputed (74).
For the purposes of making future projections, it has been decided that the reduction of
life expectancy as identified in studies from other countries provides a more reliable
measure of the outflow from the diabetic population than the Dutch cause of death
statistics. Besides, data on life expectancy provide additional information compared with
mortality data, because the duration of diabetes is also included.
34
Diabetes mellitus: points of departure
30
25
VI
~ 0
~
'"
»
'0 »
c: 0c
0
u
0
U
" 'a."
'" '"x
"0
20
15
10
5
n:
0
0
10
Age at onse1
Figure 2.' Reduction of life e.\pectallCY among jJatieJ1ls with diabetes mellitus according to age at
oJ/set.
Note: the shaded area in Figure 2 is the area behveen the highest and lowest estimates.
Source: the data come from Marks & Krall (71), Goodkin (72). and Panuam (73), adapted by Van der Veen (70).
According to this background study as a fIrst step to answering the questions that are
addressed in this thesis, it was decided in conclusion:
to choose the questionnaire survey conducted among all Dutch paediatricians and
internists to estimate the incidence among 0-19 year-aids (recorded in the period 19781980)(64,65);
to choose the continuous morbidity registration of the Dutch Sentinel Practice Network
to estimate the incidence from age 20 onwards (recorded in 1980-1983) and the
prevalence for all age groups (recorded in 1980)(66);
to assume that no remission takes place;
to choose the reduction of life expectancy as identified in studies from other countries
instead of Dutch mortality statistics (70-73),
35
Chapter 2
REFERENCES
1.
WHO Study Group on Diabetes Mellitus. WHO Tcdnical Report Series 727. Geneva: World Health
2.
Organization, 1985.
WHO Expert Committee on Diabetes Mellitus. WHO Technical Report Series 646, Geneva: World Health
Organization. 1980.
3.
WHO Expen Committee on Diabetes Mellitus. WHO Technical Report Series 310. Geneva: World He..1lth
Organization, 1965.
4.
5.
West Kl\1. Epidemiology of diabetes and its vascular lesions. New York: E1scvier/North-HollMd. Inc, 1978.
Feskcns ElM. Bowles CH, Kromhout D. Intra- and interindividual variability of glucose tolerance in an
elderly population. J Clio Epidemiol 1991; 44: 947-953.
6. Herman WH. Sinnock P, Brenner E, Brimberry JL, L1ngford D. Nakashima A. Sepe SJ, Teutsch SM,
Mane RS. An epidemiologic model for diabetes mellitus: incidence, prevalence and mortality. Diabetes
Care 1984; 7: 367-371.
7. Lartkso M. Py6rillti K. Age of onset and type of diabetes. Diabetes Care 1985; 8: 114-117.
8. Klein R. Klein BEK, Moss SE, Davis MD. DeMetes DL. The Wisconsin epidemiologic study of
retinopathy. III. Prevalence and risk of diabetic retinopathy when age at diagnosis is 30 or more years. Arch
Ophthalmol 1984; 102: 527-532.
9. WHO Report of a WHO Study Group. Prevcntion of diabetes mellitus. WHO Technical Report Series 844.
Geneva: World Health Organization, 1994.
10. Tuomi T. Groop LC, Zimmet P, Rowley MJ. Knowles WC, Mackay IR. Antibodies to glulanlic acid
II.
12.
13.
14.
15.
decarboxylase reveal latent autoimmune diabetes mellitus in adults with a non-insulin dependent onset of
disease. Diabetes 1993; 42: 359-362.
Tattersall RB, Fajans SS. A diffcrence between the inheritance of classicaljuveniIe·onset and maturity-onsct
type dinbcles of young people. Diabetes 1975, 24: 44-53.
Bell GI. Xiang KS, Newman MY. Wu SH, Wright LG, Fajans SS, SpieumUl RS, Cox NJ. Gene for noninsulin-dependcnt diabetes mellitus (maturity-onset diabetes of the young MODY subtype) is linked to DNA
polymorphism on human chromosome 20 q. Proc Natl Acad Sci USA 1991; 88: 1484-1488.
FrogueJ P. Yaxillaire M, Sun F, Velho G, Zouali H. Brutel MO, Lesage S, Vionett N, Clement K,
Fougerousse F. Close linkage of glucokinase locus on chromosome 7p to early·onset non-insulin-dependent
diabetes mellitus. Nature 1992; 356: 162-164.
Hatters!ey AT, Turner RC. Pennuu MA. Patel p, Tanizawa Y, Chiu KC, O'Rahilly R, Watkins PJ.
Wainscoat JS. Linkage of type 2 diabetes to the glucokinase gene. Lancet 1992; 339: 1307-1310.
Vaxillaire M, Boccio V, Phillippi A, Vigouroux C. Terwilliger J, Pnssa P. Beckmann JS, Velho G. Lathrop
GM, Froguel P. A gene for maturity onset diabetes of the young (MODY) maps to chromosome 12q. Nat
Genet 1995: 9: 418-423.
16. Steering Committee on Future Health Scenarios. Chronic diseases in the year 2005. Volume I. Scenarios
on diabetes mellitus 1990·2005. DordrechtlBostonfLondon: Kluwer Academic Publishers. 1991.
17. Pirart J. Diabetes mellitus and its degenerative complications: a prospective study of 4,400 patients observed
betwecn 1947 and 1973. Part 1. Diabetes Care 1978; I: 168-188.
18. Pieart J. Diabetes mellitus and its degenerative complications: a prospective study of 4,400 patients observed
between 1947 and 1973. Part 2. Diabetes Care 1978; I: 252-263.
19. Dwyer MS. Melton m.LJ, Ballard OJ, Palumbo PJ. Trautmann JC, Chu CPo Incidence of diabetic
retinopathy and blindness: a population-based study in Rochester, Minnesota. Diabetes Care 1985; 8: 316-
322.
36
Diabetes mellitus: points of departure
20. Harris MI, Klein R, Welborn TA, Knuiman
21.
22.
23.
24.
~1W.
Onset of NIDDM occurs at least 4·7 yr before clinical
diagnosis. Diabetes Care 1992: 15: 815·819.
Harris lvII, Hadden we, Knowler we, DenneU PH: Prevalence of diabetes mellitus and impaired glucose
tolerance and plasma glucose levels in U.S. population aged 20-74 yr. Diabetes 1987; 36: 523·534.
Cromme PVM. Glucose tolerance in a typical Dutch community. Thesis. Amsterdam, Ihe Netherlands: Free
University Amsterdam. 1991.
Stolk RP, Grabbee DE. De prevaIcntie van diabetes mellitus ell hyperinsulinemie bij auderen. Hel ERGO
onderLOck. T Soc Gezondheidsz 1994; 72 (middcnkatem): 49.
Mooy JM, Grootcnhuis PA, Vries H de, Valkcnburg HAt Bouter LM, Koslense Pl. Heine Rl. Prevalence
and detenninants of glucose intolerance in a Dutch Caucasian population. TIle Hoom Study. Diabetes Care
1995; 18: 1270-1273.
25. Eisenbarth OS. Type I diabetes mellitus: a chronic autoimmune disease. N Engl J Med 1986; 314: 1360-
1368.
26. Krolewski AS, Warram JH, Rand LI, Kahn CR. Epidemiologic approach to the ctiology of type I diabetes
mellitus and its complications. N Eng! J Med 1987; 317: 1390-1398.
27. Atkinson.MA, Maclaren NK. The pathogenesis of insulin-dependent diabetes mellitus. N Engl J Med 1994;
331: 1428-1436.
28. Palmer JP. Predicting lDDM-1991. Diabetes Reviews 1993; I: 104-115.
29. Bruining GJ, Grobbee DE, Scheffer CJ, Bruyn MID, Molenaar JL, Hofman A. Ten-year follow-up study
of islet-cell antibodies and childhood diabetes mellitus. Lancet 1989; 333: 1100-1103.
30. Lebovitz HE. Ethiology and pathogenesis of diabetes mellitus. Paediatrics Clinics of North America 1984:
31: 521-530.
31. Olmos P, A'Hem R, Heaton DA, Millward BA, Risley D, Pyke DA. Leslie ROO. TIle significance of
concordance rote for type I (insulin-dependent) diabetes in identical twins. Diabetologia 1988; 31; 747-750.
32. Skyler J. Immune intervention in Type I diabetes mellitus. Diabcles/Metabolism Reviews 1993; 1: 15-42.
33. NerupJ, Mandrup-Poulsen T, MolvigJ. The HLA-IDDM association: implications for etiology and pathogenesis ofIDDM. Diabetes/Metabolism Reviews 1987; 3: 779-802.
34. Nepom OT, Ehrlich. H. MHC Class II molecules and autoimmunity. Annual Rev Immunol 1991; 9: 493-
525.
35. Gamble DR. The epidemiology of insulin-dependent diabetes. with particular reference to the relationship
of virus infection to its etiology. Epidemiol Rev 1980; 2: 49-70.
36. Gleason RE, Kahn CB, Funk lB, Seasonal incidence of insulin-dependent diabetes (IODM) in
Massachusetts. 1964-1973. Int J Epidemiol1982; 11: 39-45.
37. Borch-Johnsen K. loner G, Mandrup-Poulsen T. Christy M, Zachau-Christiansen B, Kastrup K, Nerup J.
Relation between breast-feeding and incidence ofinsulin-dependent diabetes mellitus. Lancet 1984; 2: 1083-
1086.
38. Dahlquist G, Blom LO, Persson L. Sandstrom AIM. Wall SOL Dietary factors and the risk of developing
100M in childhood. Brit Moo J 1990; 300: 1302-1306.
39. Virtanen SM. Rtistinen L, Aro A. Lindstrom J, Sippola H. LOWlamaa R, Toivanen L, Tuomilehto },
Akerblom HK. Infant feeding in Finnish children < 7 yr of age with newly diagnosed IDDM_ Diabetes Care
1991: 14: 415-417.
40. Norris JM, Scoll FW. A meta-analysis of infant diet and insulin-dependent diabetes mellitus: do biases play
a role? Epidemiology 1996; 7: 87-92.
41. Reaven OM. Role of insulin resistance in human disease. Diabetes 1988; 37: 1595-1607.
42. Zimmet P. Challenges in diabetes epidemiology - from West to tile rest. Diabetes Care 1992; 15: 232-252.
37
Chapter 2
43. Zimmet P. The epidemiology of diabetes mellitus and associated disorders, In: Alberti KGMM. Krall LP
(eds.), The Diabetes AnnuaJj6. Amsterdam: Elsevier, 1991: 24: 1·19
44. Tuomilehto J, Knowler we, Zimme! P. Primary prevention of non·insulin·dcpcndent diabetes mellitus.
Diabetes/Mctabolism Reviews 1992; 8: 339-353.
45. Tuomilchlo J, Tuomilehto-Wolf E, Zimmet P. Alberti KGMM. Keen H. Primary prevention of diabetes
mellitus. In: Alberti KGMM, DeFronzo RA, Keen H. Zimmet P (cds.), International textbook of diabetes
mellitus. London: John Wiley & Sons Ltd, 1992: 1655-1673.
46. Zimmet P. The pathogenesis and prevention of diabetes in adults: genes. autoimmunity, and demography.
Diabetes Care 1995; 18: 1050-1064.
47. Helmrich SP, Ragland DR, Leung RW, Paffcnbarger RS. Physical activity and reduced occurrence of noninsulin-dependent diabetes mellitus. N Engl J Med 1991; 325: 147-152.
48. Feskcns EJM. NulritionaJ factors and the etiology of non-insulin-dependent diabetes mellitus: an
cpidcmiologicaJ overview. World Rev Nutr Diet 1992: 69: 1-39.
49. Modan M, Karasik A, Halkin H, Fuchs Z, Lusky A, Shitrit A, Modan B. Effect orpasl and concurrent body
mass index on prevalence of glucose intolerance and type 2 (non-insulin-dependent) diabetes and on insulin
response: The Israel study of glucose intolerance. obesity and hypertension. Diabetoiogia 1986; 29: 82-89.
50. Everhart)E, PeUitt DJ, Slaine KR, Knowler WC. Duration of obesity is a risk factor for non-insulindependent diabetes mellitus. Am J Epidemiol 1986; 124: 525.
51. Ohlson L-O, Larsson B, Svtirdsudd K, Welin L, Eriksson H, Wilhelmsen L. Bjomtorp p. Tibblin G. The
innuence of body fat distribution on the incidence of diabetes mellitus. 13.5 years of follow-up of the
participants in the study of men born in 1913. Diabetes 1985; 34: 1055-1058.
52. Voedingsrtls.1.d. Advies Richtlijncn Goede Voeding. 's-Gravenhage: Voedingsraad, 1986.
53. Papoz L, Eschwcge E, Warnel J-M, Richard J-L. Claude J-R. Incidence and risk factors of diabetes in the
Paris Prospective Study (GREA). In: Eschwege E (ed.) Advances in diabetes epidemiology. Amsterdam:
Elsevier Biochemical Press. 1982, 113-122.
54. Strunpfer MJ, Colditz GA. Willen WC, Munson JE, Arky RA, HelUlCkens CH, Spcizcr FEA. A prospective
study of moderate alcohol drinking and risk of diabetes in women. Am J Epidemiol 1988; 128: 549-558.
55. Holbrook TL. Barrett-Conner E. Wingard DL. A prospective population-based study of alcohol use and non·
insulin-dependent diabetes mellitus. Am J Epidemiol 1990; 132: 902-909.
56. Feskens EJ1..1, Kromhout D. Cardiovascular risk factors and the 25·year incidence of diabetes mellitus. TIle
Zutphen Study. Am J Epidemiol 1989; 130: 1101-1108.
57. Rimm EB, Chan J. Slanlpfcr MJ, Colditz GA, Willett WC. Prospective study of cigarette smoking, alcohol
use. and the risk of diabetes in men. BMJ 1995; 310; 555·559.
58. Perry IJ, Wannamcthee SO, Walker MK. Thomson AG, Whincup PH, Shaper AG. Prospective study of risk
factors for developing of non-insulin dependent diabetes in middle aged British men. BMJ 1995; 310: 560·
S64,
59. MedaJie JH, Papie eM, Ooldbourt U, Herman JB. Major factors in the development of diabetes mellitus
in 10,000 men. Arch Intern Mcd 1975; 135: 811·87.
60. Wilson PWF, Anderson KM, Kannel WB. Epidemiology of diabetes in the elderly. Am J Med 1986:
80(slIppi SA): 3·9,
61. Rimm EB, Manson JE, Stampfer IvU, Colditz GA. Willet WC. Rosner B, Hennekens CH, Speizer FE.
Cigarette smoking and the risk of diabetes in womcn, Am J Public HeaJth 1993; 83: 211-214.
62. Stuurgroep Toekomstsceoario's Gezondheidszorg. Scenario's in de volksgezondheid. Inleiding in de
methodiek van de STG. Utrecht: Jan van Arkel, 1989.
3R
Diabetes mellitus: points of departure
63. Hoogenveen RT. RUWfu'lfd D, Vclde LJK van der, Verklcij H. Incidenlie. prevalcntie en ziekteduur. Een
dynamischc beschrijving. Rapportnr. 958606002. Bil'hoven: RIVM, 1990.
64. Vaandrager OJ, Vcenhof FJ, Klauw MM van deT. Dc incidenlie van insuline-afhankelijke diabetes mellitus
bij O~19 jnrigen in Nederland (1978-1980), Lciden: NlPG-TNO, 1984.
65. Vaandragcr GJ, Bruining OJ, Veenhof FJ, Drayer HM. Incidence of childhood diabetes in the Netherlands:
a decrease from north to south over North Western Europe? Diabctoiogia 1984; 27: 203-206.
66. Continuolls Morbidity Registration, Sentinel Stations. the Netherlands. Annual Reports. Utrecht, the
Netherland<;; NIVEL, 1980-1983.
67. CCIltlTh:'li Bureau voor de Slatisliek. Statistisch zukboek 1980. 's-Gravcnhage: Staatsuitgcvcrij, 1980.
68. Fuller JH. Cause..<; of death in diabetes mellitus. Horn Metab Res 1985; 15 (Supp1): 3-9.
69. Mackenbach JP. Sncls IAK. Friden-KiU LM. Diabetes mellitus als doodsoorzaak. Ned Tijdschr Geneeskd
1991: 135: 1492-1496.
70. Veen EA van der. Epidemiology of diabetes mellitus and risk factors for end-organ disease. Postgmd Med
J 1988: 64 (Suppl 3): 5-9.
71. Marks HH & Krall LP. Onset, course, prognosis and mortality in diabetes mellitus. In: Marble A. White
P, Bradley RF & Krall LP (eds). Joslin's Diabetes mellitus. 11th cdn. Lea and Feabiger. Philadelphia 1971:
209-254.
72. Goodkin G. Mortality factors in diabetes. J Occup Med 1975; 17: 716-721.
73. Panzram G, Zabel-Langhennig R. Prognosis of diabetes mellitus in a gcogmphically defined population.
Diabetologia 1981; 20: 587-591.
74. Panzram G. Mortality and survival in type 2 (non-insulin-dependent) diabetes mellitus. Diabetologia 1987;
30: 123-131.
39
CHAPTER 3
Increasing incidence of type I diabetes in
the Netherlands:
the second nation-wide study among
children under 20 years of age
Publishes as:
Dirk Ruwaard1, Remy A. HirasiJlg 2, H. Maartell Reese,J, Stefvan BllIlrell 2, Karel Bakker4 , Rob
J. Heill~, Rolf A. Geerdill/f', C. Jan Bruining7. Gerrit J. Vaalldragel 2, S. Pauli11e VerlooveVa/lhol'ick'. Diabetes Care 1994; 17: 599-601.
1.
2.
3.
4.
5.
6.
7.
National Institute of Public Health and the Environment, Bilthoven, the NeUlcrlands
TNO Institute fOf Prevention and Health. Leiden, the NeUlcrlands
Juliana Children's Hospital, The Hague, the Netherlands
Spaame Hospital Heemstede, Hecmstede, the NeUlcrlands
Department ofIntemal Medicine, Free University Hospital Amsterdam. Amsterdam, the Netherlands
Dutch Diabetes Association, Amersfoort. the NcUlcrlrulds
Department of Paediatrics. Academic Hospital Rotterdam/Sophia Children's Hospital. Rotterdam,
the NcUlerlands
Increasing incidence of childhood diabetes
ABSTRACT
Objective - A nation-wide retrospective study was conducted to assess the incidence of
type I diabetes in the Netherlands among children <20 years of age in 1988-1990. The
first study with a similar design covered 1978-1980.
Research design and methods - The capture-recapture census method was choseil for
analysis of the data. A questionnaire was sent to all Dutch paediatricians and internists,
and for the ascertainment, a similar questionnaire was sent out separately to members of
the Dutch Diabetes Association, which is the national patient association.
Results - The average achieved ascertainment rate was 81%. The ascertainment-adjusted
annual incidence was 13.2/100,000 for 0-19 year-old children, indicating an increase of
23% compared with the 1978-1980 survey; for 0-14 year-olds, the increase amounted to
17%.
Conelusions - This study suggests a sustained increase of type I diabetes in the
Netherlands because the cumulative incidence studied previously in the 1960-1970 birth
cohorts of male army conscripts 18 years of age was also found to rise. In contrast to
Northern European countries, an increase in incidence for the age category 0-4 years could
not be found.
43
Chapter 3
INTRODUCTION
During the past decades, an increase in incidence of type I diabetes has been found in
several countries (1-6). A study amongst the 1960-1970 birth cohorts of 18 year-old male
army conscripts revealed that the incidence of type I diabetes is also rising in the
Netherlands (7). Our study offers an opportunity to assess recent changes in incidence of
type I diabetes in the Netherlands in both sexes because this second nation-wide study
covering 1988-1990 had a design similar to the previous 1978-1980 study (8).
RESEARCH DESIGN AND METHODS
A questionnaire was sent in January 1991 to all paediatricians and internists to obtain data
on children <20 years of age newly diagnosed with type I diabetes during the years 1988,
1989, or 1990. The questionnaire requested infonnation on the child's initials, gender, date
of birth, date of first insulin injection, and residence at that time.
As in the first study (8), the national Dutch Diabetes Association (DDA) was selected as
a secondary source for validation. In April 199 I, the DDA mailed a questionnaire to all
members registered since 1988 and born since 1968. Registration and reporting by the
specialists and DDA have not changed since the first study.
The ascertainment rate was defined as the proportion of responding patients from the DDA
who also were reported by the specialists. The method used to estimate the incidence and
its confidence intervals (CIs) is based on the capture-recapture census described by Hook
et al. (9). The same formulas have been used to recalculate the incidence figures for 19781980. Changes in incidence estimates per 100,OOO/year were considered significant when
the CIs did not overlap. For the 0-19 and 0-14 year-olds, the incidence rates were
standardized to the age (5-year intervals) and sex distribution of the population during
1978-1980.
44
Increasing incidence of childhood diabetes
RESULTS
On 1 November 1991, 100% and 87% of the paediatricians and internists, respectively,
did respond. The paediatricians reported 840 youngsters and the internists reported 329
youngsters 0-19 years of age in whom insulin treatment was initiated in 1988-1990. Of
the DDA questionnaires received, 799 met the criteria. From the 1,169 patients reported
by the specialists, 643 (55%) were also responding members of the DDA. Of the 156
DDA members not reported by the specialists, 36 were treated by non-responding
specialists, 117 were treated by responding specialists who failed to report, and 3 were
under treatment by doctors who were missing on the mailing list (two general practitioners
and one paediatrician abroad).
No significant differences were found in ascertainment rates according to sex, month of
fIrst insulin injection. year of first insulin injection, and province of residence at that time
(two-way X2 test; P<0.05). Ascertainment by paediatricians (91 %) was significantly higher
than by internists (54%), and consequently, the overall decrease in ascertainment rate with
age was significant (Table I).
Table J: Cases of type I diabetes reponed by specialists alld members of the DDA according
to type of specialist, age starting on illsulin, alld rate of ascertainment.
Cases of type I diabetes (n)
Rate of
Specialists
Type of specialist
Paediatrician
Intemist
DDA
members
Both
ascertainment
sources
(%)
Significance of
difference in
ascertainment rate
840
329
570
227
520
123
91
54
P<O.OOI
160
292
436
281
110
204
290
195
99
181
256
107
90
89
88
55
P<O.OOI
Age starting on insulin
(years)
0- 4
5- 9
10-14
15-19
Note: The sum of DDA members according to type of specialist (797) is not equal to the sum of members according
fo age startiog on insulin (799) because two members were treated by general practitioners.
45
Chapter 3
The ascertainment-adjusted incidence rate was 13.2/100,000 (95% CI 12.7-13.7) per year
for 0-19 years-aids and 12.4/100,000 (95% CI 12.1-12.7) per year for 0-14 year-aids. For
boys as well as girls, the ascertainment-adjusted incidence increased with the first three
5-year age categories, after which a decline was observed in the age category 15-19 years
(Table 2).
Table 2: Ascertainmelll-adjusted 3-year i/lcidellce of type J diabetes and the anllilal incidellce
per 100,000 according to age alld sex.
Mean
Ascertainment adjusted number
Age (years)
( 1988-1990)
population
over the
3 years
Ascertainmentadjusted incidence
per lOO,OOO/year
95% CI
Boys
0- 4
5- 9
10-14
15-19
96
159
259
287
470,081
453,269
465,841
578,051
6.8
11.7
18.5
16.6
6.3- 7.2
11.0-12.4
17.9-19.1
14.3-18.9
Girls
0- 4
5- 9
10-14
15-19
83
173
235
221
450,499
433,872
444,098
554,671
6.1
13.3
17.7
13.3
5.9- 6.3
12.9-13.7
16.9-18.4
11.5-15.1
When comparing our data with the ascertainment-adjusted annual incidence rates for 19781980 (10.9/100,000 (95% CI 10.5-11.4) for 0-19 year-aids; 11.1/100,000 (95% CI 10.711.5) for 0-14 year-aids), a significant increase of21 and 12% was apparent for the ageranges 0-19 and 0-14 years, respectively. The standardized ascertainment-adjusted
incidence in 1988-1990 was 13.5/100,000 (95% CI 13.0-14,0) per year for 0-19 year-aids
and 12.9/100,000 (95% C[ 12.6-13.2) per year for 0-14 year-aIds, indicating an even larger
increase of 23 and nearly 17%, respectively. Table 3 shows the ascertainment-adjusted
incidence rates for both periods according to age. With the exception of the age category
0-4 years, the incidence increased significantly.
46
Increasing incidence of childhood diabetes
Table 3: Ascertaillmellf-adjllsled anllual incidence a/type I diabetes per 100,000 with its CIs
accordillg to age ill /978-/980 alld /988-/990.
1978-1980
Age (years)
0- 4
5- 9
10-14
15-19
Incidence per
lOO,OOO/year
6.8
11.0
14.3
lOA
95% CI
6.6- 7.1
10.3-11.6
13.4-15.3
9.3-11.6
1988-1990
Incidence per
lOO,OOO/year
6.4
12.4
18.1
15.0
95% C[
6.2- 6.7
12.0-12.7
17.6-18.6
13.5-16.5
CONCLUSIONS
This study is the fIrst one in which the change in incidence over time has been estimated
by the capture-recapture census method (9). Comparing the standardized results of this
study (1988-1990) with the first nation-wide study (1978-1980), a signifIcant increase in
incidence of type I diabetes was found of 23 and 17% for the age-ranges 0-19 and 0-14
years, respectively. This increase could not be attributed to factors leading to a spurious
rise, such as improvement of diagnosis, changes in case definition, or declining diseasespecific mortality (10). Especially in the 0-14 year age category with high ascertainment
rates, the increase could not be attributed to underreporting.
Drykoningen et al. (7) studied the cumulative incidence of type I diabetes in male anny
conscripts 18 years of age in the Netherlands over a IO-year period. A significant
nonlinear increase in the birth cohorts of 1960-1970 was found (on the average 4.4% with
each annual birth cohort). Although the cumulative incidences for birth cohorts are not
directly comparable to the incidences found in the current study, both studies suggest a
sustained signifIcant increase in incidence of type I diabetes in the Netherlands.
The Diabetes Epidemiology Research International Group reported an annual increase in
incidence ranging from 10.1 % in New Zealand to 2.8% in Nonvay (3). Although the rise
established in our study is lower than in those countries, it is a substantial increase. In our
study the absence of increase in the 0-4 year-olds is striking. This contrasts with the
findings in Sweden and Finland (4,5). In Leicestershire the most prominent increase could
even be found in the youngest age categories (3). In both our studies a north-south
gradient was not present in our small but densely populated country.
47
Chapter 3
The causes of the increasing incidence, as observed in several countries, are unknown. It
is unlikely that it can be attributed to changes in genetic susceptibility (11,12). Although
etiologically important factors in the environment have not been identified with certainty,
observed differences in incidence over time and between countries may be helpful in the
search for environmental determinants of type I diabetes.
ACKNOWLEDGMENTS
This study, supported by the Stichting Research Fonds Diabetes Mellitus Grant 89.5, has
been performed at the TNO Institute for Prevention and Health, Leiden, the Netherlands.
REFERENCES
1.
2,
Bingley Pl. Gale EAM. Rising incidence of IDDM in Europe. Diabetes Care 1989; 12: 289-95.
loner G. S¢wik O. Increa<;ing incidence of diabetes mellitus in Norwegian children 0-14 years of age, 19731982. Diabctologia 1989: 32: 79-83.
3.
Diabetes Epidemiology Research International Group. Secular trends in incidence of childhood IDDM in
10 countries. Diabetes 1990; 39; 858-864.
4.
Nystrom L, Dahlquist G, Rewers M, Wall S. The Swedisch Childhood Diabetes Study. An analysis of the
temporal variation in diabetes incidence 1978-1987. Int J Epidemiol 1990; 19: 141-146.
5.
Tuomilehto J, Rewers M, Reunanen A, Lounamall P. Lounamaa R. Tuomilehto-Wolf E. Akerblom HK.
Increasing trend in Type 1 (insulin-dependent) diabetes mellitus in childhood in Finland. Diabetoiogia 1991;
6.
Green A, Andersen PK, Svendsen AJ, Mortensen K: Increasing incidence of early onset Type 1 (insulindependent) diabetes mellitus: a study of Danish male birth cohorts. Diabetologia 1992; 35: 178-182.
Drykoningen CEM, Mulder ALM, Vaandmger GJ, LaPorte RE, Bruining GJ: The incidence of male
childhood Type I (insulin-dependent) diabetes mellitus is rising rapidly in the Netherlands. Diabetologia
34: 282-287.
7.
1992; 35: 139-142.
8.
Vaandrager, GJ, Bruining. GJ. VeenhofFJ. Drayer NM: Incidence of childhood diabetes in the Netherlands:
a decrease from north to south over North-Western Europe? Diabetologia 1984; 27:203-206.
9. Hook EB, Regal RR. The value of capture-recapture methods even for apparent exhaustive surveys. Am J
EpidemioI1992; 135: 1060-1067.
10. Central Office for Statistics: Death according to causes by death certificates. age and sex, 1940-1990.
Annual Reports. Voorburg, the Netherland~: Central Office for Statistics, 1941-1991.
11. Green A. Svejgaard A. Platz P, Ryder LP. Jakobsen BK. Morton NE, MacLean CJ: The genetic
susceptibility to insulin-dependent diabetes mellitus (IDDM): combined segregation and linkage analysis.
Genet Epidemiol 1985; 2: 1-15.
12. Dahlquist G, Blom L, Tuvemo T, Nystrom L, Sandstrom A. Wall S: The Swedish Childhood Diabetes Study
-results from a nine year case register and a one year case-referent study indicating that Type 1 (insulindependent) diabetes mellitus is associated wilh both Type 2 (non-insulin-dependent) diabetes mellitus and
autoimmune disorders. Diabetologia 1989; 32: 2-6.
48
CHAPTER 4
Is the incidence of diabetes increasing in
all age groups in the Netherlands?
Results of the second study in the Dutch
Sentinel Practice Network
Published as:
Dirk Ruwaardl , ROllald Gijsell l , Aad I.M. Bartelds', Remy A. Hirasing', Harry
Verklei/, Daall Kromhout l .,. Is the incidence of diabetes increasillg ill all age groups
ill the Netherlands? Results of the secolld study in the DlIIch Selltillel Practice
Network. Diabetes Care 1996; 19: 214-218.
I.
2.
3.
4.
National Institute of Public Health and the Environment. Bilthovcn. Ole Netherlands
Netherlands Institute fOf Primary Health Care. Utrecht, the Netherlands
TNO Institute for Prevention and Health, Leiden, the NcUlcriands
Department of Epidemiology and Public Health, Agricultural University Wageningen, the
Netherlands
Is diabetes increasing?
ABSTRACT
Objective - To assess possible changes in the incidence of diabetes in all age groups
in the Netherlands during a lO-year period (1980-1983/1990-1992).
Research desigll alld methods - Since 1970, a network of sentinel stations (the Dutch
Sentinel Practice Network) consisting of about I % of the Dutch population has been in
operation to gain insight into the morbidity patterns of the Dutch population as
recorded by general practitioners. One of the items recorded from 1990 to 1992 was
the incidence of diabetes. The fIrst study with a similar design that registered the
incidence of diabetes was conducted from 1980 to 1983.
Results - The overall incidence of diabetes increased significantly by 12.1 % in the
period between the two studies. This overall increase can largely be attributed to a
statistically signifIcant increase in the age group 45-64 years (30.5%). Although not
statistically significant, the 36% increase of diabetes in the age group 0-19 years is in
accordance with the increase of type I diabetes based on the first and second nationwide retrospective studies covering the total Dutch popUlation.
COllclusions - There is a marked increase in the incidence of diabetes in the age group
45-64 years. This selective increase is probably not due to a real rise caused by
changes in exposure to risk factors but to an earlier recognition of symptoms and signs
of diabetes followed by blood glucose measurements and/or to more intensive case
finding in general practice.
51
Chapter 4
INTRODUCTION
In recent decades, an increase in the incidence of type I (insulin-dependent) diabetes
has been found in several countries (1-6). A study among the 1960-1970 birth cohorts
of 18 year-old male army conscripts (7) and a comparison of the first (1978-1980) and
second (1988-1990) nation-wide retrospective studies (1988-1990) among individuals
<20 years of age (8,9) revealed that the incidence of type I diabetes is also rising in
the Netherlands.
Whether the incidence of diabetes is also increasing for those> 19 years of age in the
Netherlands is not known. One of the results of a Delphi investigation that we
conducted in 1989-1990 among 33 experts on diabetes in the Netherlands indicated an
average expected increase in incidence of 8% for the period 1980-2005 (10,11). The
study presented here offers an opportunity to assess empirically based changes in the
incidence of diabetes, especially for those> 19 years of age, during a lO-year period
( 1980-1983/1990-1992).
RESEARCH DESIGN AND METHODS
In the Netherlands, general practices are a useful source for gaining insight into the
morbidity patterns of the population. In the Dutch health care system, everyone has
their own general practitioner, who operates as a 'gatekeeper', This implies that health
problems will first be presented to the general practitioner and that no patient will visit
a specialist without being referred by his or her general practitioner. In addition, the
specialist informs the general practitioner about clinical or policlinical findings (such as
diagnosis and laboratory results). However, it should be emphasized that the morbidity
patterns registered in general practice specifically reflect the health problems presented
by those who make an appeal to the health care system.
Since 1970, a network of sentinel stations (the Dutch Sentinel Practice Network) has
been in operation to gain insight into the morbidity patterns of the Dutch population as
recorded by general practitioners (12). This network has been designed to be as
representative for the Dutch population as possible (for age, sex and degree of
urbanization) and covers about I % of the population. It was realized when recmiting
the 'spotter' physicians that there could be no question of a random sample of Dutch
general practitioners; an expressly positive attitude on the part of the participating
52
Is diabetes increasing?
physicians was called for, plus an intention to participate for a number of years.
Primarily physicians that had participated in the forerunner of this network (the first
Dutch National Morbidity Survey), which consisted of 50 general practitioners (12),
were involved. In addition, interested general practitioners who applied themselves or
were recommended by others were selected, taking into account the criteria that the
network should be representative for the Dutch population and cover I % of the total.
The same criteria were used when a general practitioner left the network and had to be
replaced by another. To determine how representative the study sample is compared
with the total Dutch popUlation, a census is performed every 2 years. Since 1970, the
network has consisted of 60-65 general practitioners working in about 45 sentinel
stations.
A committee decides annually which items will be recorded on a registration form that
has to be filled in by the general practitioner and sent to the Central Project Bureau
once a week. At this bureau, the forms are checked, and in the case of uncertainties,
the general practitioner is contacted. One of the items recorded in 1990-I 992 by 63
general practitioners (43 sentinel stations) was the incidence of diabetes, which is
defined as the number of patients newly diagnosed during that period per 1,000 personyears according to the diagnostic criteria formulated in 1985 by the World Health
Organization (WHO) (13). The overall denominator (expressed in person-years)
represents the sum of the separate denominators per sentinel station. To estimate the
denominator per sentinel station during the period 1990-1992, the number of people
present in that station (according to the census of 1991) was multiplied by the
registration period (mostly the full period of 3 years, in one station 2 years, and in
another station half a year). For every recorded patient, a supplementary questionnaire
was filled in to collect information about the diagnostic approach, the treatment given,
and the complications present at age of onset.
The first time the incidence of diabetes was recorded in the Dutch Sentinel Practice
Network was in the period 1980-1983 (62 general practitioners working in 46 sentinel
stations). To estimate the denominator, the censuses of 1979, 1981 and 1983 were
used. At that time, the WHO criteria of 1980 (14) were used instead of the 1985
criteria (13). Depending on the circumstances in which the blood glucose value was
measured (whole blood/plasma, venous/capillary, fasting/2 h after a 75-g glucose load),
the diagnostic cutoff levels according to the 1980 and 1985 criteria differ 0.0-0.3
mmollI from each other. For instance, according to the 1985 criteria, the diagnostic
fasting cutoff value measured in capillary whole blood amounted to 'e.6.7 mmollI, while
53
Chapter 4
this value was "-7.0 mmol/lusing the 1980 criteria. On the other hand, the diagnostic
cutoff value in capillary whole blood 2 h after a 75-g oral glucose load was "-Il.l and
"-11.0 mmol/!, applying the 1985 and 1980 criteria, respectively. As we recorded the
glucose values in our second study (1990-1992) and retrospectively traced the glucose
values of the incident cases (who were still alive) in our first study (1980-1983), it
became possible to detect spurious changes in the incidence caused by differences in
diagnostic criteria. It appeared that only one newly diagnosed patient (out of 654) in
the second study (based on 1985 criteria) would not have been diagnosed using the
1980 criteria (a thirsty 61 year-old woman with a fasting blood glucose of 6.9 mmol/l),
while all newly diagnosed patients in the first study (based on 1980 criteria) would
have been diagnosed according to the 1985 criteria.
To COlTeet for changes in incidence caused by demographic developments, all data was
standardized (by 5-year age groups and sex) to the Dutch population of 1990. Because
the first study did not distinguish between men and women, we pooled these figures in
the second study. This was also done with the subsequent age groups >65 years.
Changes in incidence were then calculated for the age groups 0-19, 20-44, 45-64, and
>64 years. Statistical significance was tested with the 1 test to compare two proportions
(P<0.05). In addition, the 95% Cis of the differences in incidence were estimated using
the normal approximation for the binomial distribution.
RESULTS
Table I shows the incidence of diabetes. Note that the incidence increases up to 80
years of age, after which a decline can be seen. This applies to men as weB as to
women. A significant difference between men and women according to age group (z
test; P<0.05) could not be found. However, the absolute number of newly diagnosed
diabetic patients is the largest in the age group 45-65 years.
Table 2 presents the changes in incidence by comparing our recent study in the sentinel
network (1990-1992) with the former one (1980-1983). The figures indicate that the
overall incidence of diabetes increased significantly by >12% over a period of 10
years. This overall increase can largely be attributed to a significant increase in the age
group 45-64 years. For the other age groups, the increase is not significant, although
the relative increase is most prominent in the youngest age group.
54
Table 1: Estimated incidence (per 1,000 person-years) and total number of newly diagnosed patients with diabetes in the Netherlands in
1990 according to the Dutch Sentinel Practice Network.
Men
2
Total
Women
3
4
2
3
4
2
3
4
Age group (years)
0-19
20-44
45-64
65-79
~80
Total
10
45
132
110
23
320
52_236
89.192
46.255
19.438
4.186
211.307
0.2
0.5
2.9
5.7
5.5
1.5
(0.1-0.4)
(0.4-0.7)
(2.4-3.4)
(4.6-6.8)
(3.5-8.2)
(1.3-1.6)
372
1.605
4.443
3.597
732
10.749
8
25
143
121
37
334
50.411
88.029
47.336
24.922
8.735
219.433
0.2
0.3
3.0
4.9
4.2
1.5
(0.1-0.3)
(0.2-0.4)
(2.5-3.5)
(4.0-5.7)
(3.0-5.8)
(1.3-1.7)
296
868
4.710
4.152
1.270
11.295
18
70
275
231
60
654
102.647
177.221
93.591
44.360
12.921
430.740
0.2
0.4
3.0
5.2
4.6
1.5
(0.1-0.3)
(0.3-0.5)
(2.6-3.3)
(4.5-5.9)
(3.5-6.0)
(1.4-1.6)
668
2.473
9.153
7.749
2.002
22.044
Data are 1} observed number of newly diagnosed patients in the Sentinel Practice Network in the period 1990-1992; 2) total number of person-years in the Sentinel Practice
Network (1990-1992); 3) estimated incidence per 1.000 person-years. standardized to the Dutch population in 1990 (95% eI): 4) estimated total number of newly diagnosed
patients in the Netherlands in 1990.
~
"-
i·
0-
~
~
:i"
";;:
'"
~.
~
v.
v.
·00
Chapter 4
Table 2: Estimated illcidellce of diabetes per 1,000 persoll-years ill 1980-1983 alld ill 19901992, stalldardized to the Dutch poplliatioll ill 1990.
Age group (years)
0-19
20-44
45-64
265
Total
1980-1983
1990-1992
0.13
0.41
2.26
5.15
1.33
0.17
0.43
2.95
5.16
1.49
Absolute increase
(95% CI)
Increase
(%)
0.05 (-0.03-0.13)
35.9
3.4
30.5
0.3
12.1
om (-0.09-0.12)
0.69 ( 0.25-1.13)
0.01 (-0.80-0.82)
0.16 (0.01-0.31)
CONCLUSIONS
To obtain incidence estimates that are less prone to chance. a rather large population
size is needed. The Dutch Sentinel Practice Network, consisting of more than 140,000
people, has the largest denominator of all Dutch continuous morbidity registrations in
primary care. The others contain 50,000 people or fewer (15-18). Nevertheless, we
recorded the incidence in a 3-year period to increase the denominator even more.
To correct for an undercount of cases, the capture-recapture census method is
recommended in the literature, using an independent secondary source for
ascertainment (19). Even though a secondary source for validation is lacking, the
incidence figures found in the Dutch Sentinel Practice Network are likely to be
reliable. The network has been in operation for a long period of time, and the general
practitioners who participate in it are not only highly motivated but also experienced in
recording health problems. Most general practitioners have been participating for many
years; about 66% of the general practitioners are still involved after a to-year period.
Besides, in the Dutch health care system, general practitioners play a central role
because they operate as gatekeepers. In spite of the fact that diabetic patients are
diagnosed or treated by a specialist, the general practitioner is infonned by the
specialist and is therefore able to record health problems detected by the specialist.
According to the results of the supplementary questionnaire from our recent study
(1990-1992), 17.9% of all newly diagnosed patients were recorded in this way.
Nevertheless, just a few cases might have been missed at the end of the recording
period because of a delay in transferring information. However, the same applies to the
previolls study.
56
Is diabetes increasing?
On the other hand, when the incidence figures from the Dutch Sentinel Practice
Network are compared with those from the relatively small samples of other
continuous morbidity registrations in primary care, the incidence in the sentinel
network is \.5- to 2-fold lower (15-18). Discrepancies in the objectives, design,
definition of the numerator (such as diagnostic criteria), extent and definition of the
denominator, and length of the recording period are assumed to be responsible for the
differences (20). For example, the Dutch Sentinel Practice Network is specially
designed to obtain incidence and prevalence figures in primary practice, while other
registrations focus more on recording medical consumption, or include uncertain
diagnoses.
This study is the first in the Netherlands to assess possible changes in the incidence of
diabetes for all age groups based on a rather large denominator. When the results of
this study were compared with the former study with a similar design, it appeared that
the relative increase (nearly 36%) was greatest in the age group 0-19 years. However,
the number of cases in this younger age group is too few to obtain statistically
significant changes, despite the fact that the sentinel network covers about I % of all
Dutch inhabitants and that in both studies several years were used to estimate the
average annual incidence. Nevertheless, the change in incidence of diabetes in the age
group 0-19 years is indicative of an important increase. This finding is in line with the
23% increase of type I diabetes based on the first (1978-1980) and second (1988-1990)
nation-wide retrospective studies involving all paediatricians and internists and
covering the total Dutch popUlation (8,9). The causes of this increasing incidence,
observed in several countries (1-9), are unknown.
It is striking that above 20 years of age, a statistically significant increase in the
incidence is only found in the age group 45-64 years. As stated earlier, this rise is
probably not the result of changes in diagnostic criteria. Besides, it seems unlikely that
changes in exposure to risk factors, especially for type 11 (non-insulin-dependent)
diabetes, are responsible for this 30.5% age-specific increase. Three large screening
projects on cardiovascular risk factors in which height and weight were measured
indicated that in the period 1974-1991 there was no change in the mean BMl (kg/m2 )
or marked increase in the age-adjusted prevalence of obesity (BMl :2:30 kg/m2) in the
Dutch population (21,22). Data from the last screening project among 36,000 men and
women aged 20-59 years showed a stable BMl in the period 1987-1991, with a slight
significant increase in obesity of 0.3% per year for men. The mean prevalence of
obesity amounted to 7.4% for men and 9.0% for women, respectively (22). A marked
57
Chapter 4
change in the prevalence of physical inactivity in the period 1987-1991 was not
observed either (22).
Recently, general practitioners have become very aware of diabetes as a public health
problem. In 1988, the Dutch College of General Practitioners published its 'Standard
Diabetes Type lJ' (23). This standard contains guidelines on the diagnosis, treatment,
and support of non-insulin-using type II patients. One of the recommendations is to
examine every person with an impaired glucose tolerance annually. In 1988-1990 the
Steering Committee on Future Health Scenarios emphasized the phenomenon of
underreporting diabetes and the importance of the disease as a major and growing
cause of prolonged ill health and premature mortality (10). The Steering Committee
brought to the attention of medical practitioners the resuits of the second National
Health and Nutrition Examination Survey (NHANES II, 1976-1980) carried out in the
U.S. It was found that diagnosed patients in the age range 20-74 years represent only
50% of all patients with diabetes (24). Recently, it appeared that in the Netherlands,
many individuals also suffer from undiagnosed disturbances in glucose metabolism (25-
27). In a cross-sectional study among 2,472 persons aged 50-74 years in the Dutch
town of Hoarn, the prevalence of previously diagnosed diabetes was 4.2%, while
diabetes was newly diagnosed in 4.8% by means of an oral glucose tolerance test (27).
The Hoom Study findings are in line with those from the NHANES: roughly 50% of
patients with diabetes were undiagnosed. In accordance with pronouncements in the
literature (28,29), the Steering Committee is cautious about the establishment and
administration of large screening programmes for type 11 diabetes. It was recommended
instead to explore the possibilities of case finding in general practice among people
>50 years of age with obesity and/or a positive family history of type II diabetes
and/or the existence of complications that might be attributable to diabetes. These
developments may have influenced the general practitioners' diagnostic behaviour and
might be responsible for the increase found in the age group 45-64 years, because this
group is of special interest with respect to case-finding activities in general practice.
To verify this hypothesis, a questionnaire was sent to the general practitioners who
participated during both recording periods. It was confirmed that for this age group
there is a tendency to measure blood glucose in those who make an appeal to the
health care system for other health problems (case finding). On the other hand, the
general practitioners observed that people are better able to recognize symptoms
associated with diabetes, whereas a greater alertness on the part of the general
practitioner leading to earlier "cognition may also be of importance. In addition, blood
58
Is diabetes increasing?
glucose measurements to confirm the diagnosis conducted by the general practitioner
when symptoms are found are common nowadays (replacing the less sensitive
measurements of glucose in the urine used in the past).
To underpin these observations quantitatively, it is valuable to compare the use of
dia·gnostic tests and the presence of symptoms in newly diagnosed patients over time.
An increase in case-finding activities will be accompanied by a decline in the existence
of initial symptoms. However, when an increase in case-finding activities as well as an
increase in early recognition of symptoms and signs is evident, the results will be
difficult to interpret. Unfortunately, because valid additional information was not
available from the former study, it is difficult to establish changes in the diagnostic
approach. The recent study indicated that in addition to glucose measurements to
establish the diagnosis, 65.4% of the cases also initially presented symptoms and signs
associated with diabetes.
Our findings illustrate that to interpret trend data, one must be aware of different kinds
of developments that not only are confined to etiological factors per se, but also take
into account changes in health care practice. This is necessary not only to interpret
time trends in the incidence (and prevalence) of type II diabetes found within one
study, but also to make comparisons between studies and countries. The ideal solution
for disentangling real trends from trends due to changes in the proportion of diagnosed
and undiagnosed patients is to link periodic or continuous morbidity registrations
(physician-diagnosed cases) both in time and on an individual level with intermittently
perfonned population-based (screening) surveys. It is worthwhile to explore the most
cost-effective ways to achieve this ideal.
ACKNOWLEDGMENTS
The study was supported by grant 89-111 from the Ministry of Health, Welfare and
Sport. We thank the participating general practitioners and the secretary of the Central
Project Bureau of the Dutch Sentinel Practice Network for efforts in recording and
collecting the data.
59
Chapler 4
REFERENCES
l.
2.
Bingley Pl, Gale EAM. Rising incidence of lDDM in Europe. Diabetes Care 1989: 12: 289-295.
Joncr G, S¢vik O. Increasing incidence of diabetes mellitus in Norwegian children 0-14 years of age.
3.
Diabetes Epidemiology Research International Group. Secular trends in incidence of childhood IDDM in
4.
Nystrom L, Dahlquist G, Rewcrs M, Wall S. TIle Swedisch Childhood Diabetes Study. An analysis of
the lempomi variation in diabetes incidence 1978-1987. Int J Epidcmiol 1990; 19: 141-146.
Tuomilehto), Rewcrs M, Reunanen A, Lounamaa P, Lounamaa R, Tuomilehlo-Wolf E, Akerblom HK.
Increasing trend in type 1 (insulin-dependent) diabetes mellitus in childhood in Finland. Diabelologia
1973-1982. Diabctologia 1989; 32: 79-83.
10 countries. Diabetes 1990; 39: 858-64.
5.
1991; 34: 282-87.
6.
7.
Green A. Andersen PK. Svendsen AJ, Mortensen K. Increasing incidence of early onset type 1 (insulindependent) diabetes mellitus: a study of Danish male birth cohorts. Diabetoiogia 1992: 35: 178-82.
Drykoningen CEM. Mulder ALM, Vaandrager GJ, L"lPorte RE. Bruining GJ. TIle incidence of male
childhood type 1 (insulin-dependent) diabetes mellitus is rising rapidly in tile Netherlands. Diabetologia
1992; 35: 139-142.
8.
9.
10.
II.
12.
13.
14.
15.
16.
17.
18.
19.
60
Vaandrager. GJ, Bruining, GJ. Veenhof FJ, Drayer Nlvf. Incidence of childhood diabetes in tile
Netherlands: a decrease from north to south over North-Western Europe? Diabetologia 1984; 27: 203-
206.
Ruwaard D. Hirasing RA. Reeser HM. Bakker K. Heine RJ, Geerdink RA. Bruining GJ. Vaandrager GJ,
Verloove-Vanhorick SP. Increasing incidence of type I diabetes in the Netherlands: the second
nationwide study among children under 20 years of age. Diabetes Care 1994: 17: 599-601.
Steering Committee on Future Health Scenarios. Chronic Diseases in the year 2005. Volume I:
Scenarios on Diabetes Mellitus 1990-2005. Dordrecht-Boston-London: Kluwer Academic Publishers.
1991.
Ruwaard D. Hoogenveen RT, Verkleij H. Kromhout D. Casparie AF, Veen EA van der. Forecasting the
number of diabetic patients in the Netherlands in 2005. Am J Public Health 1993; 83: 989-995.
Bartelds AIM, Fracheboud J, Zee J van der. TIle Dutch Sentinel Practice Network; relevance for public
health policy. Utrecht. the Netherlands: Institute for Primary Health Care, 1989.
WHO Study Group on Diabetes Mellitus. WHO Technical Report Series 727. Geneva: World Health
Organization, 1985.
WHO Expert Committee on Diabetes Mellitus. WHO Technical RejXlrt Series 646. Geneva: World
HealtIl Organization, 1980.
Lisdonk EH van de, Bosch WJHM, Huygen FlA, Lagro-Jansen ALM. Ziekten in de huisartspraktijk
(Diseases in general practice). Utrecht: Wetenschappelijke Uilgeverij Bunge, 1990.
Lrunberts H, Brouwer Hl. Mohrs J. Reason for encounter- & cpisode- & process-oriented standard
output from the Transition Project. Part 1 and 2. Amsterdam: Department of General Practice, 1991.
Oskam SK, Brouwer ill, Mohrs J. TRANS, an interactional access program for standard output of the
Transition Project. Amsterdam: Dep.:'1rtment of General Practice, 1994.
Metsemakcrs lFM, H6ppener P, Knottnerus JA, Umonard ChBO. Health problems and diagnoses in
family practice. Maastricht: Registration Network Family Practices University of Limburg, 1992.
LaPorte RE. McCarthy D, Bruno G, Tajima N, Baba S. Counting diabetes mellitus in the next
millennium: application of captUf~-recapturc tedmology. Diabetes Care 1993; 16: 528-534.
Is diabetes increasing?
20. Meyboom-de Jong B. Morbidity registration in general practice in the Netherlands. Huisarts &
\Vetenschap 1993; 36(suppJ): 49-53.
21. Blokstra A, Kromhout D. Trends in obesity in young adults in the Netherlands from 1974 to 1986. lnt J
Obesity 1991; 15; 513-521.
22. Verschuren WMM, Smit HA, Leer EM van, Berns lvWH, Blokstra A, Steenbrink·vM Woerden JA,
Seidell JC. Prevalentie van risicofactorcn voor harl· en vaatziekten en veranderingen daarin in de
periode 1987·1991. Eindrapportagc Peilstatiollsproject Hart· en Vaatziekten 1987·1991. Reportnr:
528901011. Bilthoven: National Institute of Public Health and the Environment, 1994.
23. Nederlands Huisartsen Genootschap (Dutch College of General Practitioners). NHG-Standaard diabetes
mellitus type TI. Utrecht: Nederlands Huisartsen Genootschap, 1988.
24. Harris IvfI, Hadden WC, Knowler WC, Bennett PH. Prevalence of diabetes mellitus and impaired
glucose tolerance and plasma glucose levels in U.S. population aged 20·74 yr. Diabetes 1987; 36:
523-534.
25. Cromme PVM. Glucose tolerance in a typical Dutch community. Thesis. Amsterdam, the Netherlands:
Free University Amsterdam, 1991.
26. Feskens EJM, Bowles CH, Kromhout D. Intra- and interindividual variability of glucose tolerance in an
elderly population. J Clin Epidemiol 1991; 44: 947·953.
27. Mooy JM, Grootenhuis PA, Vries H de, Heine RJ, Bouler LM. The Room Study: disorders of glucose
tolenmce in a general Caucasian population (Abstmct). Nelh J Med 1992; 41: A29·A30.
28. Singer DE. Samet JR. Coley CM, Nathan OM. Screening for diabetes mellitus. Ann Intern Mcd 1988:
109: 639-649.
29. U.S. Department of Health and Human Services. Screening for diabetes: No. 16. Guide 10 clinical
preventive services. Atlanta, GA: Centers for Disease Control. 1989.
61
CHAPTER 5
Forecasting the number of diabetic
patients in the Netherlands in 2005
Published as:
Dirk RUll'aard', Rudolf T. Hoogenveen', Harry Verkleij', Daan KromllOut lJ , Anton F.
Casparie 3, Ed A. van der Veen'. Forecasting tfle number of diabetic patients in tfle
Netherlands in 2005. Am J Public Health 1993; 83: 989-995.
I. National Institute of Public Health and the Environment, Bilthoven, the Netherlands
2. Department of Epidemiology and Public Health, Agricultural University Wageningen, tile
Netherlands
3. Department of Health Policy and Management, Erasmus University Rolterdam, Rotterdam. the
Netherlands
4. Department of Endocrinology, Free University Hospital Amsterdam, Amsterdam, the Netherlands
Forecasting diabetes mellitus
ABSTRACT
Objectives - There is evidence from past decades that the number of diabetic patients has
increased independently of changes in demography. A static model that takes into account
only demographic changes is therefore unable to forecast the expected number of diabetic
patients correctly.
Methods - We developed a dynamic model in which actual incidence, prevalence and life
expectancy data are used and alternative assumptions about future trends in these
parameters can be incorporated.
Results - This dynamic model forecasts higher numbers of diabetic patients than the less
sophisticated static model. According to the dynamic model, a 46% increase in the number
of diabetic patients in the Netherlands can be expected, from 244,000 in 1990 to 355,000
in 2005 (about 2.5% annually). The static model forecasts a 22% increase.
Conc/usions - Diabetes mellitus will become a more serious public health problem than
can be expected from demographic changes only. In planning future health care,
mOllitoring of trends in incidence, prevalence, remission, and mortality or life expectancy
is a necessary prerequisite.
65
Chapter 5
INTRODUCTION
In planning future health care for an aging Western popUlation, one of the main problems
is the number of patients with chronic diseases expected in the next few decades. Diabetes
mellitus is a major and growing cause of prolonged ill health and premature mortality that
affects tens of millions of people in countries at all levels of development (I). Therefore,
diabetes mellitus was selected as a case study for further investigation in the Netherlands
(2,3).
The results of two prognostic models (static and dynamic) for computing the number of
patients with diabetes mellitus expected in the Netherlands in 2005 are presented. The
static model takes into account only actual prevalence data and demographic changes. The
dynamic model also makes use of information about actual incidence as well as life
expectancy data. The dynamic model was developed because there is evidence that in past
decades type I (insulin-dependent) as well as type II (non-insulin-dependent) diabetes
mellitus increased independently of demographic changes (4-8), which makes the static
model inadequate.
With a historic simulation procedure, it was possible to compute prevalence figures and
compare them with actual prevalences. This procedure can be considered a validation of
the dynamic model. Varying the actual incidence, prevalence, and life expectancy data on
diabetic patients made it possible to test the sensitivity of the dynamic model in
forecasting the number of diabetic patients in 2005 (sensitivity analysis).
To our knowledge this is the first study that uses more than demographic changes to
compute the number of diabetic patients expected.
METHODS
Description of the two models
Two distinct models are used to compute the number of patients expected in the year
2005. The first is called a 'static' or 'equilibrium' model. In this model the assumption
is made that the age and sex -specific prevalence of diabetes mellitus will remain constant
over time. Apart from the prevalence, the only parameter of importance is demography
(changing quantity and composition of the Dutch popUlation until 2005). A simple
66
Forecasting diabetes mellitus
multiplication of the age and sex-specific prevalence by the population estimates. at a
certain moment yields the expected number of diabetic patients. This model makes it
possible to determine the influence of demographic changes on the number of patients
expected.
The second model is called a 'dynamic' or 'disequilibrium' model. In this model the age
and sex-specific prevalence is not presumed in advance to be stable over time. Extra data
are needed on the age and sex-specific influx of new patients into and the age and sexspecific efflux of known patients out of the pool of diabetic patients. The incidence
represents the influx. The efflux is the sum of diabetic patients who die and the diabetic
patients who recover from the disease. In our model, data on the reduction of life
expectancy from the moment diabetes mellitus is diagnosed, instead of mortality figures
from death certificates, are used to define the efflux. Data based on death certificates are
considered to be unreliable. On 25% to 77% of the death certificates of patients with
diabetes mellitus, this disease is not mentioned at all (9). Recovery from diabetes mellitus
is not unlikely, but often temporary and medical care in terms of blood glucose and weight
control is still recommended. Therefore the assumption is made that recovery does not
occur.
In the dynamic model two variants are used. In the first variant the incidence remains
constant over time; in the second variant a regular age and sex-specific increase in
incidence is taken into account. In both variants the reduction of life expectancy is kept
constant over time, because in the literature there is no evidence that life expectancy has
changed substantially for the majority of the patients (type II patients). The impact of
improved survival, of type I patients only, on the projections would be limited. First, type
I patients represent only 10% to 20% of all patients. Second, because of demographic
changes in the period from 1980 to 2005 (i.e., the aging of the Dutch population), the
absolute numbers of type II patients will strongly increase, while the proportion of type
I patients will decrease. For a more detailed description of the models, see Appendix A
and Hoogenveen et aJ. (10).
To start the dynamic model we had to estimate the age and sex-specific distribution of the
diabetic patients over the years of remaining life expectancy in the first year of the
simulation period. A more detailed description of this precalculation procedure is given
in Appendix B.
67
Chapter 5
Data used
As baseline for computing the number of diabetic patients expected in 2005, the year 1980
was selected. The most reliable and representative data for the Netherlands stem from
periods around this year. The age and sex-specific prevalence in 1980 is presented in
Figure I. It represents known diabetic patients registered in 1980 in a Dutch sentinel
network of general practitioners, distributed all over the country and covering about
160,000 inhabitants of all ages (1.2% of the Dutch population)(lI). The incidence in 1980
is presented in Figure 2. The incidence for age categories older than 19 years was
recorded in the period from 1980 through 1983; the data were obtained from the same
study as the prevalence figures (II). Although the sentinel network was covering about
160,000 inhabitants and 4 years were used to estimate the average annual incidence, the
incidence of diabetes mellitus in the younger age categories is too small to obtain reliable
figures. For the age categories from birth through 19 years, therefore, we used the nationwide retrospective study of children younger than 20 years, which registered all new type
I diabetic patients in the period 1978 through 1980 (12). The method chosen was a
questionnaire distributed to all Dutch paediatricians and internists. To correct for
undercount of cases, the same questionnaire was given separately to members of the Dutch
Diabetes Association, employing the capture-recapture census method for calculating the
ascertainment-corrected incidence figures. (13), Because diabetes in persons from birth
through 19 years of age is almost entirely type I diabetes, the ascertainment-cOlTected
annual incidence is presumed to represent all diabetic patients. Although ascertainment did
not take place in the Dutch Sentinel Practice Network, prevalence and incidence figures
are likely to be reliable because in the Netherlands every person has a general practitioner
who records the patient's health problems, whether the patient will be treated by the
general practitioner or by another doctor. Nevertheless, it is possible that newly diagnosed
patients who will be treated by a specialist have not yet been recorded by the general
practitioner. In that case the recorded incidence will be underestimated only slightly,
because nearly 70% of all patients will be diagnosed by the general practitioner (II).
The reduction of life expectancy for diabetic patients is presented in Figure 3 (14). These
data, which have been used in the dynamic model, are taken from three longitudinal
studies of diabetic patients (15-17). Comparing these data and mortality data of the Dutch
population from the Central Bureau of Statistics, we estimated the reduction of life
expectancy for diabetic patients at onset to be 20% to 35%, depending on age at onset.
For patients whose age at onset was in the category birth through 19 years and for those
older than 79 years, the reduction of life expectancy was 30% to 35%; for patient aged
68
Forecasting diabetes mellitus
20 through 79 years the reduction of life expectancy was 20% to 30%, decreasing with
age.
Prevalence per 10,000
800
.men
1m women
600
400
200
o~~~~~~~~~~
0-4 5-9 10-1415-1920-2425-3435-4445-5455-64 65+
Age Category
Figllre I: Prevalellce of diabetes lIlellitlls per 10,000 illhabitallts ill the Netherlallds ill 1980, by
age and sex.
Incidence per 10,000
60
.men
50
iii women
40
30
20
10
OL-____
0-4
~GL~~EmJB§L
5-9 10-14 15-1920-2425-3435-4445-5455-64 65+
Age Category
Figllre 2: Illcidellce of diabetes lIlellitlls per 10,000 illhabitallts ill the Netherlallds ill 1980, by age
and sex.
69
Chapter 5
30
25
.e'
"
~
0
--'"
20
".,
0
".,
c: u
c
15
.~
0
10
U U
:>
"'" '"x'"
0.
5
u:
0
0
10
Age at onset
Figure 3: Reduction of life e.\pectallcy for diabetic patieJIIs, by age at ol/sel.
Note: The shaded area indicates the region between the highest and lowest estimates.
Source: Reprinted with permission from van der Veen (14). Copyright 1988. The MacMillan Press Ltd.
An expected increase in incidence of 8% is assumed for the period 1980 through 2005 in
the second variant of the dynamic model. This assumption is based on the average
increase expected by 33 experts on diabetes mellitus in the Netherlands, one of the resuits
of a Delphi investigation (18). The demographic data come from the Dutch Central Bureau
of Statistics. The assumptions of the middle variant of the Central Bureau of Statistics
population forecasts are used to estimate the future number of inhabitants in the
Netherlands (19).
Historic validation and sensitivity analysis
Two validation procedures were performed to analyze the stability of the model, that is,
whether the data on incidence, prevalence, and reduction of life expectancy due to diabetes
mellitus and the assumption of no remission result in a state of relative equilibrium of the
dynamic model. The first validation procedure was a historic simulation of the prevalence
between 1955 (specific demographic data before 1955 are lacking) and 1980, assuming
time-independent relative incidence and reduction of life expectancy to forecast the 1980
70
Forecasting diabetes mellitus
absolute prevalence. We compared the calculated prevalence with the 1980 data. This
historic simulation also made it possible to subdivide the prevalence for those older than
64 years into the more specific age categories 65 through 79 years and 80 years and older.
The actual data gave just one prevalence for .all those older than 64 years.
The second validation procedure was a sensitivity analysis. We analyzed the impact on
the forecast prevalence in 2005 of variations in some main model parameters (Le., the
1980 prevalence, incidence and reduction of life expectancy data for diabetic patients). For
the incidence and prevalence, two variants were used: a 5% increase and a 5% decrease
in each age and sex-specific category compared with the actual data for 1980. For the
reduction of life expectancy, a 25% increase and 25% decrease were used. Also, one
variant with a linear increase in incidence with age for those older than 64 years was used.
For this age category the available empirical data yielded jllst one value for the incidence
for men and one for women. These values were used in the dynamic models but may not
be in accord with reality. The literature provides evidence that the incidence increases with
age for those older than 64 years (20). Therefore, in one variant a linear increase in
incidence with age was assllmed for those older than 64 years.
RESULTS
Expected number of patients predicted by the two models
The static model predicted an increase from 191,000 patients (1.35% of the population)
in 1980 to 268,000 (1.65% of the population) in 2005, an increase of nearly 41 %. Growth
and aging of the Dutch population are responsible for increases of 15% and 25%,
respectively. The dynamic model resulted in an increase to 339,000 patients (2.1 % of the
population) in 2005, that is, a total increase of 78% between 1980 and 2005. The extra
increase of 37% over the prediction of the static model is the result of the disequilibrium
between the influx and efflux of patients. The incidence exceeds the mortality. The second
variant of the dynamic model resulted in an increase to 355,000 patients (2.2% of the
population) in 2005, that is, an additional increase from 78% to 86%. In this variant the
influx exceeds the efflux of patients even more.
The absolute increase in the number of diabetic patients in the period 1980 through 2005
according to the two models is presented in Figure 4. The estimated number of patients
in 1990 predicted by the static model is 220,000 (1.5% of the population), compared with
71
Chapter 5
242,000 (1.6% of the population) and 244,000 (1.6% of the population) predicted by the
first and second variants of the dynamic model, respectively.
Diabetic Patients (x 1,000)
400
->it"'E>- •• '
350
" .. ·0 .... ·
static
dynamic variant 1
dynamic variant 2
300
250
1980
1985
1990
1995
2000
2005
Years
Figllre 4: E.lpected illcrease ill the Illimber of diabetic patiellls ill the Netherlallds betweell 1980
(Jnd 2005, accordil1g 10 the static model and l'arimlt I alld 2 of the dynamic model.
Diabetic Patients (x 1,000)
160
140
120
1112005 stalic
o 2005 dynamic variant 1
Ii 2005 dynamic variant 2
100
80
60
40
20
o L...-=",,0·19
20·44
65·79
80+
Figure 5: Age-specific prevalence/or diabetes mellitus in 1980 (Jnd 2005, according to the static
model alld variant 1 and 2 of the dynamic model.
72
Forecasting diabetes mellitus
Age-specific analysis reveals that the expected absolute rise in the period 1980 through
2005 is most prominent in the age category 45 through 64 years (Figure 5). This applies
to men as well as to women. Relatively, the most pronounced increase was found for the
age category 80 years and older (in the dynamic model the number of patients in this
category in 2005 is about 3.0 to 3.5 times the number in 1980, for both men and women).
Historic validation and sensitivity analysis
The historic simulation showed a 10% higher prevalence in 1980 than the empirical
numbers. This difference is statistically significant (P<.OO I). The calculations also showed
that the prevalence for the oldest category (80+ years) in 1980 (2.8% for men, 5.8% for
women) was lower than that for the 65 through 79 year-old category (6.6% for men, 7.7%
for women).
Table J,' Projected percentage cJzange,Sl in the number of patients ill 2005, according to the
dY/lamic model.
Change in the number of diabetic patients in 2005
Age and sex-specific
change
Prevalence
+5
-5
Incidence
+5
-5
Men
Women
Total
+0.2
-0.2
+0.4
-0.4
+D.3
+4.8
-4.8
+4.6
-4.6
+4.7
-4.7
+2.3
-9.0
+11.0
-3.7
+7.1
-6.1
-0.3
Life expectancy
+25
-25
a: as a result of variation in age and sex-specific prevalence, incidence, and life expectance in 1980.
The results of the sensitivity analysis for the dynamic model (first variant) are presented
in Table I. A 5% change in the age and sex-specific prevalence in 1980 changes the total
number of patients in 2005 by less than 1%. For the incidence, a 5% change in each age
and sex-specific category results in a similar 5% change in the total number of patients
in 2005. When the reduction of life expectancy is changed by 25%, the total number of
73
Chapter 5
patients in 2005 changes by 6% to 7%. For men, a decrease in life expectancy influenced
the results more than an increase; for women, the opposite applied. If instead of one value
for all those older than 64 years, a linear increase in incidence is used, a decrease of 9,000
patients (2.7%) is found in 2005 (not presented in Table I).
DISCUSSION
Two models were used to compute the projected number of diabetic patients: a static and
a dynamic model. The static model forecasts 268,000 patients in 2005; the dynamic model
(second variant) 355,000. These estimates include all classes of diabetes mellitus (I). Type
II diabetes represents about 80% to 90% and type I diabetes about 10% to 20% of all
diabetic patients (21,22).
Of the two models, the dynamic model is considered to be the more valid. For the static
or equilibrium model, the assumption was made that the age and sex-specific prevalence
remains constant over time. This type of model for chronic diseases can be used only if
the age and sex- specific incidence and life expectancy of diabetic patients are constant
during a long period. For diabetes this is unlikely, because an increasing incidence has
been reported in the literature (4-8). Our historic simulation procedure supports this
observation. It appeared that the forecasted prevalence for 1980 was about 10% higher
than the actual registered prevalence in 1980.
In the dynamic or disequilibrium model, it appeared that the influx of new patients was
higher than the efflux of known patients, particularly in the second variant, in which an
increasing incidence was assumed. There is no reason to assume that the past increase in
incidence, as reported in the literature, has stopped. Furthermore, there are no indications
of a significant change in life expectancy for the majority of the diabetic population (type
II patients). Therefore, this parameter was kept constant. Consequently, the second variant,
which resulted in a total increase of 86%, is viewed as the more valid. On the other hand,
it is quite obvious that the increase in incidence contributes relatively little to this total
increase in the number of diabetic patients in 2005 (8%). In contrast, changes in
demography (static model) and the disequilibrium between influx and efflux in the first
variant of the dynamic model caused increases of 41 % and 37%, respectively.
As stated earlier, the dynamic model was validated by a historic simulation procedure.
This procedure resulted in a Iewer prevalence for those persons aged 80 years and older
74
Forecasting diabetes mellitus
in 1980 than for persons aged 65 through 79 years. This finding may be a consequence
of the use of just one incidence for those older than 64 years. Empirical age-specific
incidence data for those older than 64 years were nonexistent. Therefore we might have
underestimated the incidence for those aged 80 years and older and, as a consequence, we
may have underestimated the prevalence for this age category in 1980. In the second
place, the estimated number of patients dying in this age category may be too high (the
reduction of life expectancy has been overestimated), thereby underestimating the
prevalence for those aged 80 years and older in 1980. On the other hand, the lower
prevalence in the oldest age category has been confirmed by several empirical studies
(8,23-25).
The sensitivity analysis revealed that the dynamic model is most sensitive to variations in
incidence and is relatively insensitive to variations in prevalence. (This applies to all
diseases characteIized by a prevalence that increases with age.) The majority of the
diabetic patients in 1980 were older than 64 years of age and most of them will not
survive until 2005; almost all of the diabetic patients in 2005 will represent incident
patients diagnosed in the period 1980 through 2005. The validity of the prevalence and
incidence data used is considered in the Methods section.
The dynamic model was moderately sensitive to changes in life expectancy for diabetic
patients. It is striking that for men a decrease in life expectancy influenced the results
more than an increase. For women the opposite applied. This is probably due to the cutoff
point: the year 2005. The explanation may be that the age of onset of diabetes is relatively
lower for men than for women (Figure 2). Men diagnosed at the age of 45 years in 1980
will still be alive in 2005 if life expectancy remains unchanged or increases, but they may
be dead in the event of an extra reduction of life expectancy. On the other hand, women
diagnosed at the age of 65 years in 1980 will probably be dead in 2005 if life expectancy
remains unchanged
Of
decreases, but they may be alive in the event of an increase in life
expectancy.
When a linearly increasing incidence for those older than 64 years of age was used,
instead of one incidence, the predicted number of patients in 2005 decreased by 9,000
(2.7%). The explanation is simple. First, a higher incidence in the oldest category results
in a shorter duration of the disease. Those in the oldest category will die earlier. In the
second place, the countervailing lower incidence for the age category 65 through 79 years,
which is a larger group (denominator), will result in a larger absolute decrease in the
number of diabetic patients.
75
Chapter 5
The results presented in this paper relate only to the number of patients diagnosed. In the
second National Health and Nutrition Examination Survey (1976-1980) in the United
States, it was found that diagnosed patients represent only 50% of all diabetic patients
(26). Also, in the Netherlands it appears that many individuals suffer from undiagnosed
disturbances in glucose metabolism (27-29). Preliminary results of a cross-sectional study
in the Netherlands of 2,800 persons aged 50 through 74 years revealed that the prevalence
of previously diagnosed diabetes was 4.8%. Diabetes was newly diagnosed by means of
a glucose tolerance test in 5.3%." Assuming that those results (roughly 50% undiagnosed
patients) apply to the whole Dutch population and that all hitherto undiagnosed cases are
diagnosed in 2005, an extra 355,000 diabetic patients can be expected in 2005. On the
other hand, it is not unlikely that these patients represent a subcategory needing less
intensive medical care.
Successful planning of future health care for diabetic patients depends on the availability
of valid epidemiological data on trends in the incidence, prevalence, remission, and
mortality or life expectancy for patients with this condition. Both the diabetes study of the
Dutch Sentinel Practice Network and the registry of type I diabetes by all Dutch
paediatricians and internists will be repeated - the fanner in the period 1990 through 1992,
the latter retrospectively for the period 1988 through 1990. Therefore it will be possible
to partly validate the models. Although the present computations concern diabetes mellitus
in the Dutch population, the method is also relevant for other chronic diseases and other
countries. The main restriction is the availability of valid data. It is therefore highly
recommended that registries for diabetes mellitus and other chronic diseases be started
and/or improved.
ACKNOWLEDGEMENT
This research was supported by the Dutch Steering Committee on Future Health Scenarios
(Ministry of Welfare, Health, and Cultural Affairs).
76
Forecasting diabetes mellitus
APPENDIX A - Static and dynamic models for computing the number of patients expected in tbe future.
Dynamic model
Static model
Purpose:
Definition:
Estimating changes in absolute
Estimating changes in absolute
prevalence numbers due to
prevalence nwnbcrs due to
demographic changes
demographic and epidemiologic
changes
Equilibrium: lime-independent
Disequilibrium: time-dependent
age and sex-specific relative
prevalence figures
age and sex-specific relative
prevalence figures
Assumptions
Influx equals efflux
(a) no remission (both variants)
(b) time-independent reduction of
(age and sex
specifically):
life expectancy (both variants)
(c) time-independent relative incidence
figures (first variant)
(d) time-dependent relative incidence
figures (second variant)
Fonnlllas:
(1),(2)
(3)-(6)
The static model projects actual relative prevalence figures on future population numbers.
Fonnulas:
(1)
= E... prev~~....
(2)
where yr = time moment, s = sex, a = age, prev = absolute prevaJence numbers, POP = absolute population
numbers, and PREVFR = relative prevalence figures. Names in capitals arc dala, names in lowen:ase letters arc
mcxlel results. The interpretation of the formulas is as follows:
(1) The absolute prevalence number in each age and sex category in a certain year is the prcxlucl of the timeindependent relative prevalence figure and the time-dependent popUlation number in that specifiC category.
(2) The total prevalence number in a certain year is the sum of all age and sex-specific prevalence numbers
in that year.
The dynrunic model is a Markov model. The Markov assumption is that the futurc behaviour of an individual
depends only on his or her actual slate, not on his or her (disease) history. TIle main stales being distinguished
are health status - and. if being a patient. remaining life expectancy - and age and sex.
FonnuJas:
rll)~....
= le)~..... - LERED~~.....
(3)
incp ....
= POP)~..... INC~~....
(4)
prevjH1 ....d
= prcv~~...d.1 + E. inc)~....
prev},...o
=0
,( d
= rlej~"")
(5)
(6)
77
Chapter 5
where (sec also above): rle = remaining life expectancy, Ie = Ilonnallife expectancy, LERED = reduction of life
expectancy on age at onsel, inc = absolute incidence numbers, INCFR = relative incidence figures, and d ::::
remaining life expectancy. The interpretation of the formulas is as follows:
(3) The remaining life expectancy is the difference between Ihe nannal and the reduction of life expectancy.
(4) The absolute incidence number in each age and sex category in a certain year is the product of the timc4
independent relative incidence figure and the time·depcndent population number in that specific category.
(5) The absolute prevalence number at the end of each year is the sum of the prevalence number at the slart
of the year and the incidence number during the year. TIle remaining life expectancy of the prevalent cases
who survive decreases with I year.
(6) Patients die when their remaining life expectancy is zero.
APPENDIX D - Procedure for estimating age and sex-specific distribution of patients over years of
remaining life expectancy.
To slart the Markovian dynamic modeJ, we had to estimate the age and sex·specific distribution of the prevalence
numbers over the remaining life expectancy in the frrst year of the simulation period (1980). We made two
assumptions for this precalculation procedure. First. the absolute incidence number decreases with a constant
multiplicative factor back in time. We had to make this assumption because of the lack of specific demogmphic
data for the ye.'1fS before 1955. Second. the remaining life expectancy on the age at onset of diabetes is constant
over time. TIle main formulas of this precalculation are as follows:
(\)
::; 0.98 . inc}<+L....
prev l980......J
::::
~~<1980.... inc)T....... (r1e l980......
(2)
> [l980·yr}) .
(a·a* = 1980·yr) • (rle l980...... ·d = 1980·yr)
(3)
Prcvl980...J
where (see also Appendix A): a = age in the year of prevalence (1980), a* = age in the yeM of incidence. TIle
intel]uetation of the fonnulas is as follows:
(I) The absolute incidence number in each age and sex category in 1980 is the product of the relative incidence
figure in 1980 and the population nwnber in that specific category in 1980.
(2) The absolute incidence number in each age and sex category in a cenain year before 1980 is the absolute
incidence number in Omt specific category in the next yeM multiplied by 0.98.
(3) The absolute prevalence number in 1980 is the sum of the incidence numbers during the years before 1980
that are still alive in 1980. The three multiplicative factors show the three conditions on the incidence
numbers: survival to 1980. aging from age a* in the year of incidence to age a in 1980, and reduction of
the life expectancy on the age at onset equal to the difference between the ye.1f of incidence and 1980.
(4) The calculated prevalence numbers are scaled so that the age and sex·specific numbers agree with the 1980
data. A sensitivity <Ulalysis showed that the prognostic model results (Appendix A) are not sensitive to the
exact implementation of the precalculation procedure, that is, the choice of the multiplicative factor in
equation (2), when the simulation period is about 25 years. A 5% change in this multiplicative factor
changes the total number of patients in 2005 by less than 1%.
78
Forecasting diabetes mellitus
REFERENCES
1.
WHO Study Group on Diabetes Mellitus. WHO Technical Report Series 727. Geneva: World Health
Organization, 1985.
2. Davidsc \Y, Water HPA van dc, Vaandrager GJ. Selectie van chronische pati~nlen voor scenario-onderLOck.
Melhodologische keuzeproblemen. Medisch Contact 1988; 43: 493-494.
3. Verkleij H, Casparie AF, Ruwaard D. Kromhoul D, Veldc UK van der. Toekomstscenario-studie van slart
gegaan. Diabetes mellitus, CARA. reumatoJde arthritis. Mcdisch Contact 1989; 44: 438-440,
4. Drykoningen CEM. Mulder ALM, Vaandragcr GJ, LaPorte REt Bruining GJ, TIle incidence of male
childhood Type I (insulin-dependent) diabetes mellitus is rising rapidly in the Netherlands. Diabelologia
1992; 35: 139-142.
5. Melton ill. U, Palumbo Pl, Chu CPo Incidence of diabetes mellitus by clinical type. Diabetes Care 1983:
6: 75-86.
6. Rewers M, LaPorte RE, King H, Tuomilehto J. Trends in the prevalence and incidence of diabetes:
insulin-dependent diabetes mellitus in Childhood. Wid Hllh Statist Quart 1988; 41: 179-189.
7. King H, Zimmet P. Trends in the prevalence and incidence of diabetes: nOll-insulin-dependent diabetes
mellitus. Wid Hlth Statist Quart 1988; 41: 190-196.
8. White MC, Selvin S, Menil DW. A study of multiple causes of death in California: 1955 and 1980. J CUn
Epidemiol 1989; 42: 355-365.
9. Fuller JH. Causes of dcath in diabetes mellitus. Hom Mctab Res 1985; 15(suppl): 3-9.
10. Hoogenveen RT. Ruwaard D. Velde UK van der, Verkleij H. lncidentie, prevalentie en zieJ...1eduur. Ben
dynamisch~ beschrijving. Report no. 958606002. BiltJlOven, the Netherlands: National Instiute of Public
Health and Environmental Protection. 1989.
II. Continuous Morbidity Registration, Sentinel Stations. the Netherlands. Annual Reports. Utrecht. the
Netherlands: NIVEL. 1980-1983.
12. Vaandrager GJ, Bruining GJ, Veenhof FJ. Drayer NM. Incidence of childhood diabetes in the Netherlands:
a decrease from north to south over North-Western Europe? Diabctologia 1984; 27: 203-206.
13. Bishop Yi\1M, Fineberg SE, Holland PW. Estimating the size of a closed population. In: Bishop YMM,
Fineberg SE, Holland PW (eds). Discrete multivariate analysis, 5th ed. Cambridge, Mass: Massachusetts
Institue of Technology, 1978: 229-256.
14. Veen EA van dcr. Epidemiology of diabetes mellitus and risk factors for end-organ disease. Postgrad Med
J 1988; 64(suppl 3): 5-9.
15. Marks lll-I, Krall LP. Onset, course, prognosis and mortality in diabetes mellitus. In: Marble A, White P,
Bradley RF, Kmll LP (cds). Joslin's Diabetes mellitus, 11th edn. Philadelphia, Pa: Lea and Feabiger, 1971;
209·254.
16. Gootlkin G. Mortality factors in diabetes. J Occup Med 1975; 17: 716·721.
17. Panzram G. Mortality and surviva1 in type 2 (non-insulin-dependent) diabetes mellitus. Diabctoiogia 1987:
30: 123·l3l.
18. Steering Committee on Future Health Scenarios. Chronic Diseases in the year 2005. Volume 1: Scenarios
on Diabetes Mellitus 1990-2005. Kluwer Academic Publishers, Dordrechf/Boston/London, 1991.
19 Cruijsen H. Bcyolkingsprognose 1989: meer buitcnlandsc migratie. Mnd Stat Bevolk 1990; I: 6-38.
20. Reunanen A. Prevalence and incidence of type 2 diabetes in Finland. Acla Endocrinol 1984; 262(suppl):
31-35.
2l. Herman WH, Sinnock P, BrennerE, Brimberry JL, Langford D, Nakashima A, Sepe SJ, Teutsch SM, Mazze
RS. An epidemiologic model for diabetes mellitus: incidence, prevalence and mortality. Diabetes Care 1984;
7: 367-371.
22. Laakso M, PyoriUti. K. Age of onset and type of diabetes. Diabetes Care 1985; 8: 114·117.
23. O'Sullivan JB, Williams RF, McDonald GW. The prevalence of diabetes mellitus and related variables - a
population study in Sudbury, Massachusetts. J Clrron Dis 1967; 20: 535-543.
24. Sartor G. PrcvaJence of type 2 diabetes in Sweden. Acla Endocrinol 1984; 262(suppl): 27-29.
79
Chapter 5
25. Mheen P van de. Prevalcntic van diabetes mellitus in bejaarrlenoonJen. Report no. 528904002. Bilthoven,
the Netherlands: Nationallnstiule of Public Health and Environmental Protection, Bilthovcn, 1989.
26. Harris ?\.1J, Hadden we, Knowler we, Bennett PH. Prevalence of diabetes mellitus and impaired glucose
tolerance and plasma glucose levels in U.S. population aged 20·74 yr. Diabetes 1987: 36: 523-534.
27. Cromme PVM. Glucose tolemnce in a typical Dutch community. Thesis. Amsterdam. the Netherlands: Free
University Amsterdam, 1991.
28. Fcskens ElM, Bowles CH, Kromhout D. Intra- and inlcrindividual variability of glucose tolerance in an
elderly population. J Clio Epidemiol 1991; 44: 947-953.
29. Mooy JM, Grootenhuis P, Vries H de, Heine ill, Valkenburg HA. Diabetes mellitus and impaired glucose
tolerance in a general caucasian population. In: Proceedings of the 3200 Dutch Federation Meeting. Utrecht.
the Netherlands: Federation of Medical Scientific Societies. 1991: 149.
so
CHAPTER 6
Forecasting the number of diabetic
patients in the Netherlands in 2005:
an update
Submitted as:
Dirk Ruwaardi , Rudolf T. Hoogen)'eell i , ROllald Gijsel/, Harry Verklei/, Anton F.
Casparie', Daall KromhoutY Forecastillg the number of diabetic patiellts ill the
Netherlallds ill 2005: all update.
1. National Institute of Public Health and the Enviromnent. Bilthoven. the Netherlands
2. Department of Health Policy and Management, Erasmus University Rotterdam, Rotterdam, the
Netherlands
3. Department of Epidemiology and Public Health. Agricultural University Wagcningen, Ule
Netherlands
Forecasting diabetes mellitus: an update
ABSTRACT
Objectives - To partly validate and update our forecasts of the number of diabetic
patients in the Netherlands during the period 1980-2005 with new and more detailed
data.
Methods - Two studies that recorded the prevalence and incidence of diabetes around
1980 were repeated around 1990, using a similar design. The actual prevalence in 1990
was compared with that in 1980 and with the estimated prevalences for 1990 according
to our previolls forecasts. The actual prevalence and incidence figures from around
1990 were lIsed to compute several variants for the expected number of patients until
2005.
Results - A decline of II % in the number of patients was observed between 1980 and
1990. Compared with the estimated prevalences for 1990, the actual prevalence in 1990
(1.1%) was 23-30% lower. The incidence increased by 12% between 1980 and 1990.
The update yielded an increase in the number of patients in the period 1990-2005 of
23-127%, depending on the variant chosen.
COllc/usiolls - Between 1980 and 1990, the prevalence decreased whereas the incidence
increased, resulting in an increase in the number of diabetic patients in the foreseeable
future. Diabetes mellitus is a growing public health problem and therefore important in
planning future health care.
83
Chapter 6
INTRODUCTION
During the past century, developed countries world-wide have experienced a
considerable increase in life expectancy. It was noticed that a reduction in fatal
communicable diseases was accompanied by an increase in non-communicable
diseases. This epidemiologic transition changed diabetes from a rare disease into a
disease that is now recognised as a major global public health problem. Diabetes
mellitus is expected to become even more important as developing countries get more
westernized (I).
Crude estimates indicate that well over 100 million people are affected by diabetes
mellitus around the world (2). As it often results in substantial morbidity and mortality,
the disease has a significant health and socia-economic impact. The proportion of
estimated total health care costs related to diabetes is 3.6% in the United States of
America and 4-5% in the United Kingdom (3). Besides, recent data suggest that in the
United States of America the costs of diabetes care are rising rapidly (4-6).
The recognition of diabetes as a major public health problem and the notion that it will
become even more prominent in the future makes the disease important in tenus of
planning future health care. From this point of view, one of the first questions that
needs to be answered is how the occurrence of diabetes is likely to develop in the
foreseeable future.
In order to answer that question we constructed a static and a dynamic model in which
incidence, prevalence, and life expectancy data can be used and alternative assumptions
about future trends in these parameters can be incorporated. In 1993, we published
forecasts of the number of diabetic patients in the Netherlands for the period 19802005, based on incidence and prevalence data from two studies conducted around 1980
(7). As both studies were repeated around 1990, it became possible to partly validate
the previous model results and to update the forecasts for the period up to 2005.
METHODS
Descl'iption of the two models
In our previolls forecasting study, two distinct models were used to compute the
number of patients expected in the period 1980-2005. The first is called a 'static' or
84
Forecasting diabetes mellitus: an update
'equilibrium' model. In this model the assumption is made that the age and
sex-specific prevalence of diabetes mellitus will remain constant over time. Apart from
the prevalence, the only parameter of importance is demography (the changing quantity
and composition of the Dutch population until 2005). A simple multiplication of the
age and sex-specific prevalence by the population estimates at a certain moment in
time thus yields the expected number of diabetic patients. This model makes it possible
to detennine the influence of demographic changes on the number of patients expected.
The second model is called a 'dynamic' or 'disequilibrium' model. In this model the
age and sex-specific prevalence is not presumed to be stable over time. Additional data
on the age and sex-specific influx of new patients into and efflux of known patients
out of the pool of diabetic patients are needed. The incidence represents the influx. The
efflux is the sum of diabetic patients who die or who recover from their disease. The
models have been described in more detail in a previous publication (7).
Data used in the previous forecasting study
In our previous forecasting study, the prevalence represented known diabetic patients
registered in 1980 in the Dutch Sentinel Practice Network of general practitioners,
distributed all over the country and covering about 1% of the Dutch population (8).
The incidence for age categories older than 19 years was recorded in the same sentinel
network (1980-1983), whereas for the age category 0-19 years the nation-wide
retrospective study (1978-1980) among all children aged under 20 years was used (9).
The demographic data corresponded to the data from Statistics Netherlands (10), and
life expectancy data were adapted from three longitudinal studies among diabetic
patients (II). Although recovery is likely, it is said to be mostly temporary, whereas
medical care in terms of blood glucose and body weight control is still recommended.
Therefore the assumption was made that recovery does not occur.
In addition to the forecasts based on the static model, two variants were used in the
dynamic model. In the first one the incidence remained constant over time, while in the
second one a regular age and sex-specific increase in the incidence of 8% for the
period 1980-2005 (i.e. 0.3 % per year) is taken into account. This increase was
expected by 33 experts on diabetes mellitus in the Netherlands according to the Delphi
investigation we conducted (12). In order to start the dynamic model, we had to
estimate the (age and sex-specific) years of remaining life expectancy among the
prevalent cases in the first year of the simulation period, as the data on life expectancy
85
Chapter 6
are related to the reduction of life expectancy by age at onset (i.e. amongst incident
cases), A more detailed description of this precalculation procedure is available in a
previous pUblication (7).
Data used in the present study
Employing a similar design, the Dutch Sentinel Practice Network recorded the clinically known diabetic patients in order to estimate the prevalence in 1990 (see Figure
I). All those who consulted the general practitioner in 1990-1991 and who were
diagnosed with diabetes before I January 1990 were recorded as prevalent cases,
in-espective of the reason for consultation. In addition to the first study. a
supplementary questionnaire was filled in for each recorded patient by the general
practitioner to gather information about the diagnostic approach, the treatment given
and the presence of complications. In order to avoid missing cases because of this
extra workload, it was decided to register the prevalent cases for two years (19901991) instead of one (1980) as in the former study.
Prevalence per 10,000
700
III men
600
mwomen
500
400
300
200
100
o '---'--_
.....0·19
20·44
45·64
65·79
Age CategOlY
total
Figure 1: Prevalellce of diabetes mel/fillS per 10,000 inhabitants ill the Netherlands ;/1 1990,
by age and sex.
S6
Forecasting diabetes mellitus: an update
Regarding the incidence, new data were obtained in a way similar to both former
studies. For the age category 0-19 the nation-wide retrospective study among all those
under 20 was repeated for the period 1988-1990 (13). For those over 19, the same
Dutch Sentinel Practice Network was used, which again recorded the incidence of
diabetes in the period 1990-1992 (14). Figure 2 shows the incidence according to age
and sex around 1990 based on these studies.
Incidence per 10,000
60
III men
50
mJ women
40
30
20
10
o L-_",-_
0-19
20-44
45-64
65-79
Age Category
80+
total
Figure 2: illcidence of diabetes mellitus per 10,000 illhabitams ill the Netherlallds ill 1990, by
age alld sex.
Table 1 presents the lO-year changes in prevalence and incidence according to age. To
adjust for changes caused by demographic developments, all data was standardized (by
5-year age category and sex) to the Dutch population of 1990. Because the fIrst study
in the Dutch Sentinel Practice Network did not distinguish between men and women
(at that time we obtained sex-specific figures by extrapolating age-specific sex
differences from another Dutch registry in general practice (15», we pooled these
figures in the present study. This was also done with the age categories above 65
years. Changes in prevalence and incidence were then calculated for the age categories
0-19, 20-44, 45-64 and over 64. Statistical significance was tested with the z-test to
compare two proportions (P<0.05). [n addition, the 95 percent confidence intervals (Cl)
of the differences (in prevalence and incidence, respectively) were estimated using the
87
Chapter 6
normal approximation for the binomial distribution. The method used to estimate the
incidence and the differences for the age category 0-19 years with its confidence limits
is based on the capture-recapture census as described by Hook et aI. (16).
Table 1: Age-specific challges (per 10,000) ill prevalence alld incidence over a tell-year
period (beMeen arol/nd 1980 alld 1990), standardized to tile Dwell popl/lation of 1990.
Age
0-19
Prevalence decline (95% CO
1.4
10.3
49.1
20-44
45-64
65+
178.4
Total
37.7
-2.1- 4.8)
4.1- 16.6)
( 27.3- 70.9)
(126.9- 229.8)
( 29.6- 45.7)
Incidence increase (95% CI)
0.2
0.1
6.9
0.1
1.6
( 0.2- 0.3)
(-0.9- 1.2)
( 2.5-11.3)
(-8.0- 8.2)
( 0.1- 3.0)
Validation and update
With the availability of new empirically based data, it became possible to partly
validate the previous model results and to update the forecasts for the period up to
2005.
The actual prevalence in 1990 was compared with that in 1980 and with the estimated
prevalences in 1990 (as a result of the static model and the two variants of the
dynamic model according to our previous forecasting study). The actual incidence
changes were compared with the assumptions made in the two variants of the dynamic
model according to the previous study (Le. no increase and an increase of 0.3% per
year, respectively).
We updated the model for the period 1990-2005. In agreement with our previous
forecasting study, the model-technical characteristics as well as the data on
demography, life expectancy and recovery were kept identical in this 'update'. However, the new prevalence and incidence figures from around 1990 were used. In addition
to the forecast based on the static model (adopted as the basic scenario), three variants
were computed according to the dynamic model. The first and second variants
correspond with the two variants llsed in our previous forecasting study (i.e. a constant
incidence and a regular age and sex-specific increase in the incidence of 0.3 % per
year for the period 1990-2005, respectively). In addition, in a third variant the actual
88
Forecasting diabetes mellitus: an update
age-specific incidence increase found in the ten-year period was assumed to continue
regularly (yearly) during the period 1990-2005.
RESULTS
The number of prevalent cases in 1990 amounted to 170,000 diabetic patients (1.1%).
This represents a decline of 11% since 1980 (191,000). However, the previous
forecasting model estimated 220,000, 242,000 and 244,000 diabetic patients in 1990
according to the static model and the two variants of the dynamic model, respectively.
The actual prevalence in 1990 was 23-30% lower compared with the estimated
prevalences. Standardized to the Dutch population of 1990, there was no significant
change in the age category 0-19 years, in contrast to the age categories 20-44, 45-64
and 65+ which yielded relative declines of 29%, 23% and 27%, respectively (see also
Table I).
A constant incidence and an expected increase of 8% (Le. a yearly increase of 0.3%
per age category) in the period 1980-2005 were assumed in the first and second
variants of the dynamic model in our previous forecasting study. Empirically, between
around 1980 and 1990 an overall significant incidence increase of 1.6 per 10,000 (Le.
12%) was found (see Table 1). This overall increase can largely be attributed to a
statistically significant increase in the age category 45-64 years (31 %). However, the
increase was also significant for the youngest age category (23%). Although the actual
increase is based on two points in time without information about the period in
between, this increase corresponds to a calculated annual rise of 1.1 %, which is higher
than the increase assumed in the second variant of the previous forecasting study
(0.3%).
According to the static model, the 'update' of the forecast for the period 1990-2005
resulted in an increase from 170,000 patients in 1990 to 208,000 patients in 2005 (i.e.
23%), whereas the former static model predicted an increase from 191,000 in 1980 to
220,000 in 1990 and 268,000 in 2005 (i.e. 22% in the period 1990-2005). In this
'update', the consecutive three dynamic variants showed an increase to 346,000
(104%), 353,000 (108%) and 385,000 (127%) diabetic patients, respectively. The first
and second dynamic variants of the previous forecasting study resulted in 339,000 and
355,000 patients. This is in accordance with the findings of both corresponding variants
in the 'update', although the input parameters (starting incidence and prevalence data)
were different (see Table 2).
89
Chapter 6
Table 2: E.,pecled lIumber of diabetic patients ill the Netherlands ill 2005 according to the
previolls forecasting study alld the preselll 'update',
Variants
Assumptions regarding
the actual prevalence and
incidence figures l
Number of diabetic
patients in 2005
Previous studyb
Present
'update,e
Static model: age and sex-specific prevalence constant
Dynamic model:
age and sex-specific incidence constant
- variant I
- variant 2
age and sex-specific increase in
incidence of 0.3% per year
- variant 3
sustained increase in incidence of 1.1 %
per year (age-dependent) in 1990-2005
based on the actual increase in 1980-1990
(see Table I)
268,000
208,000
339,000
355,000
346,000
353,000
385,000
a: In the previous forecasting study and the present 'update' the model-technical characteristics as ,vell as the data on
demography, life expectancy and recovery were identical. Source: Ruwaard et a1. (7).
b: Based on actual prevalence and incidence figures from around 1980. The number of patients in 1980 amounted to
191,000. Source: Ruwaard et at (7).
c: Based on actual prevalence and incidence figures from around 1990 (see Figure 1 and 2). The number of patients
in 1990 amounted to 170,000.
Age-specific analysis revealed that the expected absolute and relative increase in the
number of patients in the period 1990-2005 is most prominent in the age category
45-64 years (Figure 3), This applies to men as well as to women. In the dynamic
model the number of patients for this age category in 2005 is about 2.6-3,2 times that
of 1990 (depending on the variant chosen).
90
Forecasting diabetes mellitus: an update
Diabetic Patients (x 1,000)
200
.,990
150
112005 sIalic
02005 dynamic variant 1
2005 dynamic variant 2
m
Q 2005 dynamic variant 3
100
50
o L....:"':'."",,-0-19
20-44
65-79
45·64
Age Category
80+
Figure 3: Age-specific prel'alellce of diabetes mellillis ill 1990 alld 2005, accordillg to the
static model and three variants of the dynamic model.
DISCUSSION
The objective of this study was two-fold: (I) to validate the previous model results and
(2) to update the forecasts for the period up to 2005. The number of diabetic patients
decreased from 191,000 in 1980 to 170,000 in 1990, while the previous study predicted
an increase. This considerable discrepancy could not be explained by changes in the
design of the Dutch Sentinel Practice Network. The network has been in operation
since 1970 and the general practitioners who participate in it are not only highly
motivated but also experienced in recording health problems. Most general practitioners
have been participating for many years and about 66% of them are still involved after
a ten year period (17). Unfortunately, the opportunity to correct for an undercount of
cases by using an independent secondary source, as described by Bishop et al. (18) and
by LaPorte et al. (19), was not available in the Dutch Sentinel Practice Network for
both recording periods. On the other hand, in the most recent study all prevalent cases
were recorded over a two-year period and the general practitioner received reviews of
the patients recorded in order to avoid missing cases. Nevertheless a decrease in
prevalence was found.
91
Chapter 6
If the decline in prevalence is real then a decrease in incidence or in life expectancy or
a considerable chance of recovery (or a combination of these) seem to be conceivable
explanations. However, no decrease in incidence was found between around 1980 and
1990; rather there was an increase (13,14). A decrease in life expectancy as an
explanation of the fall in prevalence during this period is not credible either, as there
are no indications to support this. On the contrary, there is more reason to believe that
life expectancy is on the increase, because deaths from cardiovascular diseases (an
important cause of death among diabetic patients) are declining in the Netherlands
(20,21). However, studies in England and Wales indicate that diabetic patients are not
experiencing the same fan in cardiovascular mortality as that experienced by the
general population (22).
In our previous forecasting study, the assumption was made that recovery does not
occur. However, several recent studies in which an audit of diabetic care in general
practice took place (23-27), indicated that a considerable number of previouslydiagnosed diabetic patients did not meet the diagnostic criteria for diabetes fonnulated
in 1985 by the World Health Organization (28). For instance, in the Continuous
Morbidity Registration Nijmegen in the Netherlands, a total number of 427 patients
were registered between 1967 and 1989 as newly-diagnosed diabetic patients by their
general practitioner. In 1I 1 cases (26%) the diagnosis could not be confirmed
according to the 1985 WHO criteria (27). Cromme also established in Dutch general
practice that in nearly 27% of the previously-diagnosed diabetic patients aged over 64
years and treated with diet or oral hypoglycaemic drugs, the diagnosis could not be
reconfinned with an oral glucose tolerance test when applying the 1985 WHO criteria.
It were patients put on diet in particular, who were responsible for this considerable
change (80%)(25). In addition, in a population-based survey of glucose intolerance
among 2 468 subjects aged 50-74 years living in the Dutch town Hoom, the diagnosis
could not be reconfirmed in about 15% of all self-reported previously-known cases,
reducing the prevalence in this group from 4.2% to 3.6%. The cases in whom the
diagnosis could not be confirmed all used diet only and represented 57% of the total
diet-treated group (29).
In order to find out whether these observations also contribute to the observed decrease
in prevalence in the Dutch Sentinel Practice Network, we performed a questionnaire
survey among the general practitioners. It appeared that from those who remembered
their diagnostic behaviour (78%), the majority (62%) diagnosed diabetes less frequently
after 1980 than before 1980. Of the general practitioners, 24% rechecked all known
prevalent cases registered in 1980 resulting partly in 'recovery', whereas the prevalent
92
Forecasting diabetes mellitus: an update
cases in 1990 were all diagnosed according to the 1985 criteria. Because both in our
recent and former study in the Dutch Sentinel Practice Network the way of treatment
was recorded, we were able to check whether a decline in diet therapy could also be
found. From those treated with diet or oral hypoglycaemic dmgs, 43% and 25% used
diet in 1980 and 1990, respectively.
The audit of diabetic care in general practice apparently resulted in a clearance of the
diabetic register, in particular with regard to those treated with diet only. This
clearance may be the result of a real recovery (e.g. by means of changes in lifestyle
with consequent loss of weight) or may be due to changing diagnostic criteria.
Although it is not possible to quantify the separate contributions of these possible
causes, the effect of the changes in diagnostic criteria may have been largely
responsible for the observed decrease in prevalence. The first time the World Health
Organization (WHO) fonnulated diagnostic criteria was in 1965 (30), with subsequent
revisions in 1980 (31) and 1985 (28). Although the 1980 and 1985 criteria do not
differ much from each other, the diagnosis 'diabetes mellitus' based on the 1965
criteria is supposed to correspond roughly to the diagnosis 'diabetes mellitus including
impaired glucose tolerance' according to the 1985 criteria (32,33). As many people in
the Netherlands have impaired glucose tolerance (25,33-35), the changes in diagnostic
criteria may have important consequences where the diabetic register contains a
considerable number of these people. This kind of effect can be excluded for the
observed change in incidence between 1980-1983 and 1990-1992, as the incidence
figures in both studies are based on the 1980 and 1985 criteria, respectively (14).
The second objective of this study was to update the forecasts for the period up to
2005. As the 1990 prevalence was lower and the incidence higher compared with the
1980 data, the dynamic model is preferable to the static model to compute the expected
number of patients (i.e. assuming no constant age and/or sex-specific prevalence in the
period 1990-2005). The sharper increase in the number of patients in the first variant of
the dynamic model in this 'update' reflects the greater discrepancy (disequilibrium)
between the starting prevalence and incidence figures in 1990 compared to the 1980
figures used in our previous study. resulting in a quite similar number of patients in
2005 (see Table 2).
Tllis first variant of the dynamic model assumes the 1990 age and sex-specific
incidence to be constant over the period 1990-2005. However, there are no indications
that the increase in incidence for type I diabetes found in the age category 0-19 years
will stop, although the etiology remains obscure (13,36). The marked increase in
93
Chapter 6
incidence in the age category 45-64 is probably not caused by a real rise due to
changes in exposure to risk factors, but due to an earlier recognition of symptoms and
signs of diabetes followed by blood glucose measurements and/or to more intensive
case finding in general practice (14). This is plausible because in line with the findings
in the United States of America (37), it appeared that many individuals in the
Netherlands also have undiagnosed diabetes (25,33-35). It is uncertain whether
activities in general practice relating to diabetic care will continue at the 1990 level or
will be further intensified. On the other hand, a continuous increase in incidence as a
result of these activities is precluded because eventually most subclinical cases will
have been diagnosed. Hence, the third variant may result in an overestimation over the
years.
In view of the existence of a considerable number of undiagnosed diabetic patients (not
only confined to the 45-64 year age category), the real number of known diabetic
patients in the future very much depends on the policy for detecting the undiagnosed
patients (e.g. by case finding or population-based mass screening programmes)(37).
According to both most recently conducted population-based surveys of glucose
intolerance in the Netherlands (around 1990) at least 50% of all diabetic patients aged
50 years and older are currently undiagnosed (34,35). Assuming that this is also
applicable to those aged 20-49 (37), ,md all subclinical patients will have been
diagnosed by 2005, then the estimate could be double that of the first variant, i.e.
700,000 patients. This seems unlikely however, because so far there is insufficient
evidence that widespread population screening is cost-effective (38-40).
In addition, two other factors need to be mentioned that might influence the expected
number of patients. Firstly, as a result of the changes in diagnostic criteria for diabetes,
the life expectancy that is used in both forecasting studies is probably too optimistic
because the data stem from three longitudinal studies conducted before 1980 (41-43).
According to Panzram, mortality studies on diabetic populations initiated in the sixties
and seventies have included a considerable percentage of patients with impaired
glucose tolerance (44), who are expected to have a more favourable life expectancy.
On the other hand, as already mentioned, the decline in death from cardiovascular
diseases (an important cause of death among diabetic patients) in the Netherlands
might also be beneficial to diabetic patients (20,21). Secondly, recent data indicate that
the prevalence of obesity has been increasing among the Dutch population in the late
eighties and early nineties with as a possible consequence a real increase in the
incidence of diabetes over the years (45).
94
Forecasting diabetes mellitus: an update
In order to successfully plan future health care for diabetic patients, one of the
questions that needs to be answered is what the expected number of patients will be in
the foreseeable future, Therefore, a model and valid epidemiological data on trends in
the incidence, prevalence, recovery, and life expectancy are essential. In addition,
awareness of possible developments in health care (such as changes in diagnostic
criteria and case finding activities) and of changes in exposure to risk factors is also a
necessary prerequisite to obtain reliable forecasts.
In conclusion, the number of known diabetic patients recorded in the Dutch Sentinel
Practice Network decreased by II % between 1980 and 1990. This is probably due to a
clearance of the diabetic register in general practice. The incidence increased by about
12% between 1980 and 1990. All projections forecasted an increase in the number of
known diabetic patients in the period 1990-2005 (23%-127%, depending on the variant
chosen). However, the real extent of increase will very much depend on the policy for
detecting undiagnosed patients. As diabetes is a disease with a major and increasing
health and socia-economic impact, it is strongly recommended to update the
projections on a regular basis as new figures (preferably validated by a secondary
source) become available.
ACKNOWLEDGMENT
The authors gratefully acknowledge Dr A.I.M. Bartelds, project manager of the
Continuous Morbidity Registration Sentinel Stations, Netherlands Institute for Primary
Health Care, Utrecht, and Dr R.A. Hirasing, project leader of the type I nation-wide
incidence study, TNO Institute for Prevention and Health, Leiden, for providing the
prevalence and incidence data. We also wish to thank Dr EJ.M. Feskens for helpful
comments.
1.
2.
3.
Zimmet P. Challenges in diabetes epidemiology - from West to the res!. Diabetes Care 1992; 15: 232252.
McCarty D, Zimmet P. Diabetes 1994-2010: global estimates and projections. Melbourne. Australia:
International Diabetes Institute, 1994.
Laing W. Williams R. Diabetes, a model for health care management. Paper no. 92.Londen: Office of
Health Economics, 1989.
95
Chapter 6
4.
Huse OM, Oster G. Killen AR, Lacey MJ, Golditz GA. TIle economic costs of non-insulin dependent
diabetes mellitus. JAMA 1989; 262: 2708·2713.
5.
6.
Songer T. The economic costs of NIDDM. Diabetes/metabolism reviews 1992; 8: 389-404.
Songer T. 111e economics of diabetes care. In: Alberti KG11M et aI., cds. International textbook of
diabetes mellitus. London: John Wiley. 1992: 1643-1654.
Ruwaard D, Hoogenveen RT, Verkleij H, Kromhollt D, Casparie AF, Veen EA van der. Forecasting the
number of diabetic patients in the Netherlands in 2005. Am J Public Health 1993; 83: 989-995.
Continuous Morbidity Registmtion, Dutch Sentinel Pmctice Network. Annual reports. Uuccht, the
7.
8.
Netherlands: Netherlands Institute of Primary Health Care. 1980-1983.
Vaandmger GJ, Bmining GJ, Veenhof FJ, Drayer Ni\1. Incidence of childhood diabetes in the Nether·
lands: a decrease from north to south ovcr North-Western Europe? Diabctologia 1984; 27: 203·206.
10. Cruijsen H. Bevolkingsprognose 1989: meer buitenlandse migratie. Mnd Stat Bevolk 1990; I: 6-38.
11. Veen EA van der. Epidemiology of diabetes mellitus and risk factors for end·organ disease. Postgrad
9.
Med J 1988; 64(suppl 3): 5-9.
12. Steering Committee on Future Health Scenarios. Chronic Diseases in the year 2005. Vol. 1. Scenarios on
Diabetes Mellitus 1990-2005. DordrechtfBoston/London: Kluwer Academic Publishers. 1991.
13. Ruwaard D, Hirasing RA, Reeser HM, Buuren S van, Bakker K, Heine RJ, Oeerdink RA, Bruining GJ,
Vaandrager GJ. Verloove-Vanhorick SP. Increasing incidence of type I diabetes in the Netherlands: the
second nationwide study among children under 20 years of age. Diabetes Care 1994; 17: 599-601.
14. Ruwaard D, Gijsen R, Bartelds AIM, Hirasing RA, Verkleij H, Kromhout D. Is the incidence of diabetes
incrensing in all age groups in the Netherlands? Results of the second study in the Dutch Sentinel
Practice NetwOIK. Diabetes Care 1996; 19: 214·218.
15. Hoogenveen RT, Ruwaard D. Velde LJK van der, Verkleij H. Incidentie, prevalentic en ziekteduur. Een
dynrunische beschrijving. Report no. 958606002. Bilthovcn, the Netherlands: National Institute of Public
Heath and Environmental Protection, 1990.
16. Hook EB, Regal RR. The value of capture-rccapture methods even for apparent exhaustive surveys. Am
J Epidcmiol 1992; 135: 1060-1067.
17. Bartelds AIM, Fracheboud J, Zee J van der. TIle Dutch Sentinel Practice Network; relevance for public
health policy. Utrecht, the Netherlands: Netherlands Institute of Primary Health Care, 1989.
18. Bishop YMM, Fineberg SE, Holland PW. Estimating the size of a closed population. In: Bishop YN1M,
Fineberg SE, Holland PW, cds. Discrete multivariate analysis. 5th cd. Cambridge, Mass: Massachusetts
Institute of Technology, 1978: 229-256.
19. LaPorte RE, McCarthy D. Bruno G, Tajima N, Baba S. Counting diabetes mellitus in thc next
millennium: application of capture-recapture technology. Diabetes Care 1993; 16: 528-534.
20. Jansen J, Kromhout D, Verkleij H. Coronaire hartziekten. In: Ruwaard D, Kramcrs PON, eds.
Volksgezondheid Toekomst Verkelilling. De gezondheidstoestand van de Nedcrlandse bevolking 19502010. National Institute of Public Health and Environmental Protection. Den Haag, the Netherlands: Sdu
Uitgeverij; 1993: 379-386.
21. Jansen J, Keli S, Kromhout D. Cerebrovasculaire aandocningen. In: Ruwaard D, Kramers PON, cds.
Volksgezondheid Toekomst Vcrkenning. De gezondheidstoesland van de Nederlandse bevolking 1950·
2010. National Institute of Public Health and Environmental Protection. Den Haag, the Netherlands: Sdu
Uitgeverij; 1993: 387·392.
22. Fuller JH. Mortality trends and causes of death in diabetic patients. Diabctc & Melabolisme 1993; 19:
96-99.
96
Forecasting diabetes mellitus: an update
23. Verhoeven S. Behandeling. conlrole en metabole insleUing van patiE!ntcll mel diabetes mellitus type II en
de prcvalcnlic Vrul latc complicalics bij dele paticnlcn. Procfschrifl. Rotterdam: Erasmus Univcrsitcit
Rotterdam. 1989.
24. Reenders K, Rutten GEHM. Nobel E de. HoDgen liM van den, Weel C van. Mel de slandaard rus
maatslaf. Diagnosliek en behandeling van diabetes mellitus in 19 huisartspraktijkcn. Huisarts Wet 1990;
33: 379-383.
25. Cromme PVM. Glucose tolerance in a typical Dutch community. Thesis. Amsterdam. the Netherlands:
Free University Amsterdam. 1991.
26. Patchett P, Roberts D. Diabetic patients who do not have diabetes: investigation of register of diabetic
patients in general practice. nro.1] 1994; 308: 1225·1226.
27. Weel C van. Validating long tenn morbidity re<:ording. J Epidcmiol Community Health 1995; 49(suppl
I): 29-32.
28. Diabetes mellitus: Report of a WHO Study Group. Technical report series 727. Geneva, Switzerland:
World Health Organization, 1985.
29. Mooy 1M. Non-insulin-dependent diabetes mellitus in a general Caucasian population. The Room StUdy.
Thesis. Amsterdrun, the Netherlands: Free University Amsterdam. 1995.
30. Diabetes mellitus: Report of a WHO Expert Committee. Technical report series 310. Geneva,
Switzerland: World Health Organization, 1965.
31. Diabetes mellitus: Report of a WHO Expert Committee. Technical report series 646. Geneva,
Switzerland: World Health Organization. 1980.
32. West KM. Epidemiology of diabete.1; and ill; vascular lesions. New York: ElsevierlNorth-Holland. Inc,
1978.
33. Feskens EJM, Bowles CH, Kromhout D. Intra- and interindividual variability of glucose tolemnce in an
elderly population. 1 Clin Epidemiol 1991; 44: 947-953.
34. Mooy JM. Grootenhuis PA. Vries H de. Valkenburg HA, Bouter LM, Kostense PI, Heine RJ. Prevalence
and determinants of glucose intolerance in a Dutch Caucasian population. TIle Hoom Study. Diabetes
Care 1995: 18: 1270-1273.
35. Stolk RP, Grobbee DE. De prcvalentic van diabetes mellitus en hyperinsulinemie bij oudercn. Het
ERGO ondcrzoek. T Soc Gezondheidsz 1994; 72 (middenkatem): 49.
36. Drykoningen CEM. Mulder ALM, Vaandrager OJ, L"lPortc RE. Bruining GJ. The incidence of male
childhood type 1 (insulin-dependent) diabetes mellitus is rising rapidly in the Netherlands. Diabetologia
1992: 35: 139-142.
37. Harris MI, Hadden WC. Knowler we, Bennett PH. Prevalence of diabetes mellitus and inlpaircd
glucose tolerance and plasma glucose levels in U.S. population aged 20-74 yr. Diabetes 1987; 36:
523-534.
38. Harris MI. Madan M. Screening for NIDDM. Why is there no national program? Diabetes Care 1994;
17: 440-444.
39. Knowler. we. Screening for NIDDM. Opportunities for detection, treatment, and prevention. Diabetes
Care 1994; 17: 445450.
40. Nathan DM. Inferences and implications. Do results from the Diabetes Control and Complications Trial
apply in NIDDM? Diabetes Care 1995; 18: 251-257.
41. Marks RH, Krall LP. Onset, course, prognosis and mortality in diabetes mellitus. In: Marble A, White P.
Bradley RF, Krall LP. (eds.). Joslin's Diabetes mellitus. 11th edn. Philadelphia, Pa: Lea & Feabiger,
1971: 209-254.
42. Goodkin G. Mortality factors in diabetes. 1 Occup Med 1975; 17: 716-721.
97
Chapler 6
43. Panzram G, Zabcl-Langhcnnig R. Prognosis of diabetes mellitus in a geographically defined population.
Diabetologia 1981; 20: 587-591.
44. Panzmm G, Mortality and survival in type 2 (non-insulin-dc(X!ndcnt) diabetes mellitus. Diabeloiogia
1987; 30: 123-t3I.
45. Seidell Je. Obesity in Europe: scaling an epidemic. lnt JObes 1995; 19(suppl3): SI-S4.
98
CHAPTER 7
Changing occurrence of diabetes mellitus
and implications for health policy:
from a global to a national perspective
Submitted as:
Dirk Ruwaard', Edith J.M. Feskens', Hany Verkleij', Anton F. Casparie', Daan
Krom!lOut. u Changing occurrence of diabetes mellitus and implications for health
policy: from a global to a national perspective.
1. National Institute of Public Healt11 and the Enviroruncnt, Bilthoven, the Netherlands
2. Department of Health Policy and Management. Erasmus University Rotterdam, Rotterdam, the
Netherlands
3. Department of Epidemiology and Public Health, Agricultural University Wageningen, the
Netherlands
Implications for health policy
ABSTRACT
The recognition of diabetes as a major public health problem and the notion that
numbers of patients are expected to increase world-wide make the disease important
for health policy and raise questions concerning the possibilities for influencing the
expected increase by means of prevention, and the possible consequences for health
care. Based on the current knowledge, as published in the international literature, an
attempt has been made to answer these questions and to address the health policy
implications of the findings at the national level.
Up to now, primary prevention of insulin-dependent diabetes mellitus (IDDM) has been
confined to research without practical implications. To identify the determinants of
IDDM, an important contribution can be made at the national level by joining the
international network of childhood diabetes registries. Determinants of non-insulindependent diabetes mellitus (NIDDM) have been identified by epidemiological studies
(overweight, physical inactivity and unhealthy diet), but there are few empiricallybased published studies that have examined the effect of interventions on these
determinants. Because there are many similarities between both diseases, the
experience gained from the prevention of cardiovascular diseases can serve as an
example for primary preventive strategies for NIDDM at a national level. As at least
50% of all patients are undiagnosed, a national strategy based on knowledge gained
internationally is needed to formolate policy for detecting those individuals in a
community and/or clinical setting. As the number of diabetic patients is expected to
increase while increasing financial constraints are being placed on health care, more
cost-effective ways to treat diabetic patients and to reduce the complications associated
with the disease will need to be found.
In order to gain insight into the changing burden of diabetes and to determine the
feasibility and effectiveness of intervention strategies, the disease needs to be
monitored. Data from a national coordinated monitoring system (with several
subsystems) can provide the input for health models designed to forecast future
developments, which may be a valuable tool for underpinning health policy decisions.
!OI
Chapter 7
INTRODUCTION
Diabetes mellitus is a chronic metabolic disorder that represents a major public health
problem. Over time, multiple chronic complications may occur, such as cm'diovascular
diseases (myocardial infarction, stroke), circulation disorders in the legs, blindness,
kidney diseases and loss of sensitivity and/or pain in the limbs. As a consequence, both
quality of life and life expectancy are reduced. Diabetes mellitus is therefore
responsible for a substantial degree of health care utilisation (I).
During recent decades a great deal of effort has been expended to prevent and control
diabetes mellitus at various levels. At a global level, this effort resulted in the
landmark resolution on the Prevention and COlltral of Diabetes adopted in 1989 by the
World Health Assembly (2). At the European level, the meeting organized by the
World Health Organization (WHO) Regional Office for Europe and the European
Office of the International Diabetes Federation (TDF) in 1989, with representatives of
government health departments, patients' organizations and diabetes experts from all
European countries, resulted in the St Vincent Declaration Action Programme (3). This
programme, which provided the first official initiative to develop plans and policies for
the improvement of diabetes care, was adopted by the WHO Regional Committee for
Europe in 1991 (4). At the national and regional level, a number of professional and
lay organizations have joined forces and WHO has published guidelines for the
development and implementation of national diabetes programmes (5,6). Finally, at the
patients' level, daily care provided by professionals reflects that effort. It is
encouraging that many organizations and individuals are somehow involved in
activities with the ultimate goal of preventing and controlling diabetes.
Despite all these efforts it is a source of concern that diabetes is a growing public
health problem world-wide. Based on 250 prevalence studies recently conducted in
various populations, McCarty and Zimmet roughly estimated that the global number of
diabetic patients will increase from 11004 million in 1994 to 239.3 million in 20lO (7).
They emphasized that this result should be interpreted as a general indicator of diabetes
frequency and will need to be revised as new and better data become available.
Diabetes is also a growing public health problem in the Netherlands. Independent of
demographic changes, it appeared that in the period between 1980-1990 the overall
incidence increased by 12%. Age-specifically, the increase for those < 20 years and for
those 45-64 years amounted to 23% and 31%, respectively (8,9). Moreover, it is
expected that in the period 1990-2005 the number of known diabetic patients will
increase by at least 23% as a result of demographic changes. Also, taking into account
102
Implications for health policy
the initial imbalance between incidence and mortality which will become even greater
if the observed increase in incidence of 12% continues during the period [990-2005, an
increase by 127% can be expected (10).
The recognition of diabetes as a major and growing public health problem raises
questions concerning the possibilities for influencing the expected increase in the
number of patients by means of prevention, and the possible consequences of the
expected increase in terms of the burden on health care.
MKI'HODS
Prevention and health care: definitions used
In health policy, a distinction is often made between primary, secondary and tertiary
prevention on the basis of the natural history of a disease. In the literature, the
definitions used to describe the various forms of prevention are not always identical.
With respect to the prevention of diabetes we used the following definitions (II).
Primary prevention covers activities aimed at preventing a disease from occurring (Le.
reducing the incidence) in susceptible individuals or populations through modification
of environmental and behavioural risk factors/determinants, Of specific intervention for
susceptible individuals. Secondary prevention involves activities aimed at early
detection of as yet undiagnosed cases of diabetes followed by treatment with the aim
of reversing the condition and/or halting its progression. When a disease has already
been diagnosed, any action to prevent it from getting worse or leading to disablement
is a form of tertiary prevention, In this manuscript the term health care is used to
address this tertiary form of prevention, to distinguish it from both other fonns that
may have a major influence on the number of patients. Figure 1 presents the
definitions used here in relation to diabetes mellitus schematically.
It should be realised that primary prevention, secondary prevention and health care are
closely related. During the natural course of a disease there is often a gradual transition
frOlil the healthy to the unhealthy or diseased state. To make a clear distinction
between primary and secondary prevention for diabetes, blood glucose cutoffs above
which diabetes is regarded as being present can be used (I). Preventive activities aimed
at those who do not meet these criteria are basically primary preventive activities. Tn
addition, there is also a close relationship between secondary prevention and health
103
Chapter 7
care. The result of secondary preventive activities will lead to an increased burden on
health care, at least initially.
no diabetes
not yet diagnosed
diabetes
diagnosed
diabetes
deadl
binh-II
t
primary prevention
t
secondary prevention
t
health care
Figure 1.' Schematic presentatiol/ of primm)' prevellfioll, secolldmy preventioll. alld health
care related to diabetes mellitlls.
Source: adapted from Taylor ct a)., 1993 (12) according to the definitions used by WHO (11).
Note: this is not a linear lime axis.
Approach
The observed and expected increase in the number of diabetic patients raises the
question what the options are and what the policy is or might be to influence this
increase by means of primary and/or secondary prevention. Knowledge about the
epidemiology of diabetes (i.e. its determinants) is essential for effective implementation
of preventive measures. A useful paradigm for developing a public health strategy for
prevention consists of three phases: evidence from observational epidemiological
studies (cross-sectional and prospective), evidence from intervention trials and finally,
public health action (13). According to these phases, the meaning of primary and
secondary prevention has been addressed for the two most common forms of diabetes,
i.e. insulin-dependent diabetes mellitus (lDDM) and non-insulin-dependent diabetes
mellitus (NlDDM)', by reviewing the current state of knowledge as published in the
international literature. In this perspective. an attempt has been made to view the
consequences of the changing number of patients for health care. Finally, the
implications of these findings for health policy at a national level are discussed.
I
104
IODM represents about 1O~20% and NIDDM about 80~90% ofal! diabetic patients (14,15),
Implications for health policy
PREVENTION AND IDDM
Determinants of IDDM
IDDM arises in genetically susceptible individuals who are exposed to putative
environmental or exogenous triggers that may activate immunological mechanisms,
leading to a progressive loss of pancreatic islet beta cells (16-18). The excessive
glucose concentration stems from a lack of the hormone insulin (insulin deficiency),
which is produced by the beta cells. The immunological and inflammatory mechanisms
concerned have not yet been clearly defined. It is an insidious process which may
occur over many years. During the 'pre-diabetic' stage of the evolution of the disease,
individuals can often be recognized by the presence of immunological markers and a
decline in pancreatic beta-cell function (19). Studies carried out on healthy children in
the Dutch population have shown that in the presence of certain immunological
markers, 50% of the children concerned develop IDDM within eight years (20).
The fact that hereditary factors playa role is apparent from studies of identical twins,
which demonstrate a higher concordance rate for IDDM in monozygotic twins (2540%) than in dizygotic twins (5-10%) (21-23). In addition, the overall risk of IDDM
among whites in the United States of America is 0.2-0.4%, while the risk in siblings of
probands with IDDM is about 5% and in the offspring of diabetic parents 2-3% (if the
mother has the disease) and 5-6% (if the father has the disease)(23). The major genetic
predisposition is conferred by genes located on the short arm of chromosome 6, either
within or in close approximation to the major histocompatibility complex (24,25).
Viruses, such as mumps, rubella and coxsackie, are suspected exogenous determinants
(26). It is thought that incidence peaks noted by some observers in autumn and early
winter could be explained by the presence of viruses (27). In addition, there are indications that diet may also play a role. For example, breast-feeding could have a
protective effect, while a high level of consumption of protein-rich foods and carbohydrates or nitrosamine-containing foods could increase the risk of IDDM (28-30).
However, a recently published meta-analysis of 17 case-control studies showed that the
increased risk of IDDM associated with early infant diet exposures is small and may be
explained by methodological shortcomings (31). Quantification of the relative contribution of the different determinants has so far not been possible.
105
Chapter 7
Pl'imary prevention of IDDM
According to the determinants of IDDM, two ways of primary prevention can be
distinguished: (I) by reducing exposure to the determinants and (2) by means of active
intervention in the 'pre-diabetic' stage with pharmacological agents.
The results of genetic and epidemiological studies showed that at least 60% of IDDM
world-wide is environmentally determined and thus potentially avoidable (32).
However, as mentioned above, the determinants of IDDM are far from clear. The
importance of cow's milk as an environmental trigger is the subject of research in a
multinational, prospective, double-blinded trial in North America and Europe (33). In
this Trial to Reduce the Genetically at Risk (TRIGR) about 5,400 newborn fIrst-degree
relatives of IDDM patients with high-risk genes will receive either a standard cow's
milk baby formula or breast-feeding/non-antigenic protein hydrolysate until 6 months
of age with a follow-up of 10 years to detect IDDM.
The second way requires active intervention in the 'pre-diabetic stage' in order to halt
or reverse the auto-immune disease process. In particular two multinational,
prospective, intervention trials should be mentioned. Firstly, the European
Nicotinamide Diabetes Intervention Trial (ENDIT). In this trial conducted in Europe
and North America, an estimated number of over 400 first-degree relatives of IDDM
patients with high titres of islet cell antibodies will be randomly allocated to either a
nicotinamide (free radical scavenger)-treated group or a placebo-treated group. In a
second trial, the Diabetes Prevention Trial-Type I (DPT-I), over 800 relatives of IDDM
patients with prediabetes according to immunological and metabolic markers will
receive insulin in order to allow beta-cell rest and to encourage the development of
antigen tolerance or will receive a placebo (33). In addition, several immune
intervention studies in vitro, in animals and in high-risk persons or in newly diagnosed
IDDM patients have been performed or are under consideration (34).
Currently, many questions remain unanswered as regards the most appropriate interven-
tion to prevent diabetes as well as the individuals who should be targeted (on the basis
of family history, genetic markers, immunological risk markers and/or metabolic risk
markers). To identify individuals in any of the stages of prediabetes, massive screening
efforts are required. However, not all individuals will eventually develop diabetes. For
instance, first-degree relatives are at a IO-fold increased risk of developing IDDM,
while the majority (95-97%) will not develop the disease (23). Besides, 85-90% of
IDDM patients do not have a first-degree relative with IDDM. It will be some years
106
Implications for health policy
before the resuIts of ongoing trials become available, and newly emerging prevention
strategies have still to be tested in large-scale, long-term and well planned clinical
trials (11).
Secondary prevention of IDDM
The meaning of secondary prevention, as we defined it, might be of less importance in
lDDM than in NlDDM. Once the disease becomes manifest with declining insulin
production and increasing hyperglycaemia, symptoms develop rapidly and most cases
come to medical attention in a period of days or weeks. Therefore, screening at this
later stage is unnecessary as it would be impractical and as there is no sustained
benefit in achieving diagnosis a few days early (35,36). The available experience with
immunosuppressive therapy or nicotinamide therapy administered shortly after clinical
diagnosis indicates that improvement of beta-cell function is not sufficient or
longlasting. The onset of intervention comes too late in many cases, as 80-90% of the
beta-cell mass are already destroyed at onset of the disease (34).
PREVENTION AND NIDDM
Detel'lllinants of NIDDM
Primarily, lDDM and NlDDM are clinically descriptive subclasses of diabetes.
Whereas lDDM appears to be the resuIt of an auto-immune disease process (i.e. a type
I pathogenetic process), the etiology and pathogenesis of NlDDM is heterogenous.
Many patients with NlDDM and individuals with an impaired glucose tolerance (lGT)
exhibit
insulin
resistance
and
hyperinsulinaemia
in
association
with
dyslipo-
proteinaemia, central obesity and hypertension. This cluster of cardiovascular
determinants has been described by a number of names such as 'syndrome X' or
'Reaven syndrome' (37), the chronic metabolic syndrome, and the insulin resistance
syndrome (38). The extent to which this cluster of determinants represents a single
disease process is still unclear (39). Persons with a history of IGT and gestational
diabetes are at increased risk of developing NlDDM. The rate of progression from IGT
to NlDDM is about 2-3% per year in studies can-ied out in the UK and the USA, and
107
Chapter 7
the incidence of NlDDM in women with gestational diabetes is about 3-5% per year
(40,41 ).'
Whereas lDDM is most commonly encountered at a younger age « 20-30 years),
NIDDM is a form of diabetes particularly associated with advancing age. Apart from
age, heredity is currently considered as an important risk factor associated with
NlDDM, apparently playing an even greater role than it does in lDDM. In the case of
identical twins, a concordance rate of 95-100% for NIDDM can be found (21). The
risk of developing NlDDM in individuals where one or both parents have NlDDM is
almost three times as great as in those whose parents are free of the disease (43).
Irrespective of the presence or absence of the disease in the family, overweight
increases the risk of the occurrence of NIDDM by a factor of two to three (44). There
are indications that people must be overweight for some time before it becomes a risk
factor for diabetes (45,46). Also the distribution offat around the body is an important
factor. Abdominally localised body fat (a 'paunch') imposes an additional risk (47).
Table 1: Relative risks of established detem/inalllS for NlDDM.
Detenninant
Endogenous:
genetic factors
body weight
Exogenous:
physical activity
Indicator
RR
Reference
1 or 2 parents
with diabetes
2.9
43
BM! >25kglm'
waist/hip ratio> 1
2.5
2.5
44
47
energy expenditure
1.7
43
0.7
44
<2,000 kcaVweek
nutrition
Guidelines for a
healthy diet
BMI: Body Mass Index (weight (kg)/heighI1 (m»
Note: for the risk of NIDDl-.-f in those with IGT and gestational diabetes, see lext.
In addition, there are indications that 10-20% of NIDDM is the result of an auto-immune disease process as found
in lDDM (Latent Auto-immune Diabetes in Adults) and that 2-4% is the result of spedfie gene mutations (42).
108
Implications for health policy
lt has been demonstrated that physical inactivity promotes the development of NIDDM
(40,43). The composition of the diet is also a risk factor. In line with the 'Guidelines
for a healthy diet' published by the Netherlands Nutrition Council in 1986 (48), the
consumption of high-fibre foods and unsaturated fatty acids at the expense of foods
rich in saturated fatty acids is encouraged to prevent NIDDM (44). To what extent
alcohol lise and smoking are independent determinants for NIDDM is still unclear. The
results of several studies on alcohol use (49-54) as well as smoking (52-57) are
contradictory. Table 1 summarizes the relative risks of established determinants for
NTDDM.
P"imal'Y prevention of NIDDM
As with IDDM, two ways of primary prevention can be distinguished: (I) by reducing
exposure to the determinants and (2) by means of pharmacological intervention in
those with IGT (which can be viewed as a 'pre-diabetic' stage).
As overweight, physical inactivity and the composition of the diet are established
detenninants, behavioural interventions such as restricting caloric intake, reducing
dietary saturated fat and increasing physical activity may be beneficial in diabetes
prevention. These measures are probably most usefully applied to high-risk individuals
(those with a positive family history, with elements of the chronic metabolic syndrome,
with a history of gestational glucose disturbances, changing from traditional to
Westernized lifestyles. from rural to urban societies or from active to sedentary
lifestyles). In addition to the high-risk approach, a population-wide approach which
aims to modify lifestyles by introducing health education programmes is strongly
recommended in societies susceptible to NlDDM (41).
Based on the results of II cohort studies on overweight and the incidence of NTDDM,
it can be estimated theoretically that prevention of overweight as indicated by BM! <:25
kg/m' could result in a reduction of 10-60% of the incidence of NlDDM, with an intermediate value of about 30% (44). As regards physical activity, it has been estimated
that 25% of NIDDM in the USA can be attributed to physical inactivity (58). As the
majority of NlDDM cases appear to be lifestyle related, it is potentially preventable
through the pursuit of a healthy way of life. In order to prove this, intervention studies
are needed.
109
Chapter 7
No large-scale studies have yet been published that prove that a healthy lifestyle
intervention can prevent NIDDM (42). The feasibility of behavioral interventions for
the prevention of NIDDM has been demonstrated in Malmo, Sweden, in a study of two
groups of middle aged men with lOT (59). Those in the treated group (n=161) were
given physical training and advised to reduce sugar and fat, increase complex
carbohydrate and fibre in the diet, and lose weight if they were overweight. The treated
group showed a significant weight loss, most of which was maintained for 5 years,
whereas the reference group did not (n=56). After 5 years, 11 % of the treated persons
and 21 % of the reference group had developed diabetes according to the WHO criteria.
This study demonstrated the feasibility of a 5-year diet exercise programme. However,
the effect of treatment remains uncertain because those involved were not assigned
randomly to the treated or reference group. Fortunately, the preliminary results of a
successful diet and exercise intervention have recently been reported (60). A total of
530 subjects with lOT were randomized into four groups; a control group without
intervention, a diet-only intervention group, an exercise-only intervention group and a
group receiving diet and exercise treatment. The cumulative incidences after a follow-
up of 6 years were 15.7%, 10,0%, 8.3% and 9.6% in those four groups, respectively.
The results indicate that diet and/or exercise intervention treatment are effective
methods for reducing the incidence of NTDDM.
It should be realised that these positive results are based on a trial among people with
lOT. Hopefully, these results will be confirmed in other intelvention studies and also
apply to such high-risk groups outside of a trial or even in the general population.
However, it is well known that (sustained) weight reduction is hard to achieve. Despite
activities to promote health in the general population (regarding weight control,
physical activity and healthy diet) it is a matter of concern that an increase in the mean
BMl and the prevalence of obesity (BM! ;"30kg/m2) has been observed in the
Netherlands in the late eighties and early nineties, and for a longer period in several
other countries (61).
The second way of primary prevention is by means of oral antidiabetic drugs nonnally
used in the treatment of NTDDM. As these drugs are known to stimulate insulin
secretion and/or insulin action and/or to inhibit intestinal glucose absorption, the
potential for preventing NIDDM in lOT subjects had been studied in three randomised
trials (62-64). Unfortunately, the summary cumulative incidence rate ratio (active
drug/placebo) was 0.9 (95% CI: 0.6-1.5) suggesting that the active treatment had little
or no effect (65).
110
Implications for health policy
Secondary prevention of NIDDM
Several secondary preventive activities can be distinguished, such as community
screening and case finding (i.e. clinical screening), Community screening involves the
application of a screening test to individuals in the population or a part of it in order to
determine whether a certain disease might be present, prior to the appearance of
clinical signs or symptoms (66). Although case finding also targets asymptomatic
persons, it focuses on individuals rather than the population. A person who consults a
physician for a certain health problem will also be examined for the presence of other
diseases if that person is thought to be at increased risk. In addition, such individuals
might also be detected by routine examinations (67,68).
In the Netherlands there appeared to be a highly age-specific increase in the incidence
of diabetes (31%) for those aged 45-64 between 1980-1990, which is probably due to
earlier recognition of the signs and symptoms of diabetes and more intensive case
finding in general practice (9). This is plausible because several Dutch studies revealed
that many individuals have undiagnosed disturbances in their glucose metabolism (6972). In line with findings in the United States of America (73), roughly 50% of all
diabetic patients are undiagnosed.
In accordance with pronounce'ments in the literature (74), the Dutch Steering Committee on Future Health Scenarios was cautious about the establishment and administration
of large community screening programmes for NlDDM (75). In 1990 the Committee
recommended instead to explore the possibilities of case finding in general practice
among persons above 50 years of age with obesity and/or a positive family history of
NlDDM and/or the existence of complications that might be attributable to diabetes.
Opinions on the value of mass screening have changed frequently over recent decades.
In the sixties and seventies it was widely recommended. In 1978 the Diabetes
Screening Workshop in Atlanta, Georgia. proclaimed that screening to detect diabetes
should not be encouraged, except during pregnancy. Evidence of benefits did not
outweigh the evidence of adverse effects, whereas the cost of screening was not
justified (76). Are there now more valid reasons for mass screening to detect NlDDM?
To decide whether screening programmes in general are worthwhile, a number of
criteria have been formulated, which can be summarized as follows: the disease should
be an important health problem; there should be an accepted treatment and an agreed
policy on who to treat as patients; there is a recognizable latent or early symptomatic
III
Chapter 7
stage of the disease and the prognosis should be better if the disease is diagnosed and
treated early, as opposed to late or never; an appropriate screening test which is
acceptable to the population should be available; facilities for diagnosis, treatment and
follow-up should be available and the costs should be justified by the benefits, and
finally; the screening programme should be a continuing process and not a 'one-off
exercise (77).
It is established that NIDDM represents a major clinical and public health problem for
which there is an accepted treatment with an agreed policy on who should be
considered as patients according to the WHO criteria (I). As stated earlier, roughly
50% of all diabetic patients are undiagnosed. It appeared that the onset of NIDDM
may occur 9-12 years before clinical diagnosis (78). Besides, there is a great deal of
evidence that micro and macrovascular complications begin to develop before clinical
diagnosis. For instance, diabetic retinopathy is estimated to become evident
approximately 7 years before diagnosis of NTDDM (78). Retinopathy and nephropathy
appeared to be present in 10-29% and 10-37% of clinically-diagnosed NIDDM
patients, respectively (78-85). In general terms, macrovascular disease and
macrovascular determinants are as frequent in undiagnosed NIDDM as in established
NIDDM and twice as high as in non-diabetic individuals; they are even present before
the real onset of diabetes, namely in the stage of impaired glucose tolerance (86-90).
Moreover, it was observed that the adverse effect of diabetes on coronary heart disease
is enhanced disproportionately when macrovascular disease determinants (Le. risk
factors) are present (91,92).
Finally, it has been demonstrated that mortality in individuals with undiagnosed
NIDDM is at least as frequent as in those with established NTDDM and significantly
higher than in non-diabetic individuals. In the II-year follow-up of the Paris
Prospective Study, mortality was 23%, 20% and 9% in undiagnosed diabetic persons,
established diabetic patients and in non-diabetic individuals, respectively (93), and in
the IS-year follow-up of the Whitehall Study, the mortality rate per 1000 person-years
was 40, 27 and 12, respectively (94). The higher mortality in the undiagnosed
compared with established patients may be due to a lack of treatment or insufficient
treatment of associated determinants, such as hypertension and dyslipidaemia. as the
majority of patients died of macrovascular disease. For instance, it appeared from the
Hoom Study in the Netherlands that 14% of all established NIDDM patients with
hypertension were untreated, while in those cases which remained undiagnosed the
proportion was 31 % (90). It is therefore indisputable and also in line with the criteria
for screening, that NIDDM is characterized by a preclinical period with the presence of
t 12
Implications for health policy
both micro and macrovascular complications and an increased risk of morbidity and
mortality.
There was less agreement regarding the criterion that the prognosis is better if the
disease is diagnosed and treated early than rather late or never (74). The difference of
opinion focused primarily on the glucose hypothesis, i.e. whether it is possible or not
to decrease the risk of the development and progression of diabetic complications by
lowering blood glucose concentrations. In the meantime, two randomized clinical trials
have clearly demonstrated that intensive treatment improves glycaemic control and
decreases the incidence rates and rates of progression of microvascular complications
in lDDM patients (95,96). Although these trials are restricted to lDDM patients, it is
widely believed that the results support the hypothesis that hyperglycaemia is causally
related to these complications (97-102). In that case, lowering glycaemia in NIDDM
patients will also be beneficial.
Nevertheless, some investigators are still cautious about applying the results of the
lDDM trials to NIDDM. The impact of the side effects of the intensive therapies based
on oral medication or insulin necessary to achieve normoglycaemia in patients with
NlDDM (such as hypoglycaemia, weight gain, increased risk of cardiovascular events)
is unknown (103). However, there are indications that intensive therapy may be
accompanied by a very low rate of severe hypoglycaemia and without significant
weight gain (104). Besides, available clinical trials have not so far confirmed the fear
that insulin increases the risk of macrovascular events in individuals with NIDDM
(105-108). Therefore, the most important research question relating to metabolic
control in NIDDM is now whether glucose control prevents or delays macrovascuiar
disease in patients with NIDDM (109). Hopefully the Veterans Affairs Cooperative
Study (104) and the UK Prospective Diabetes Study (110) on glycaemic control and
complications in NlDDM will shed more light on this crucial issue.
On the other hand, it would be unfortunate to restrict the discussion of screening for
NlDDM to the benefits and side effects of glycaemic control only. As mentioned
earlier, micro and macrovascular complications and several dete~inants are already
evident in the preclinical period. Evidence strongly indicates that early detection of
NIDDM and intervention with diet, weight control, exercise and treatment for
hypertension and dyslipidaemia will improve the prognosis in NIDDM (99). In
addition, earlier detection of complications, such as retinopathy, followed by adequate
treatment (coagulation therapy) will certainly improve the quality of life by preventing
blindness.
113
Chapter 7
This approach will help focus the discussion more on questions such as who, where,
when, and how to screen for NIDDM (depending on the distribution of the prevalence
in the population and available resources) rather than waiting for evidence to decide
whether mass screening for NlDDM would be beneficial. Harris and Modan (99)
conclude that screening for NIDDM is an important measure for promoting health. In
community screening programmes where considerations of cost and efficiency are
important, restricting screening to individuals who are obese and/or hypertensive might
be considered. In the clinical setting, it is important to incorporate periodic screening
for diabetes into routine follow-up of at-risk patients (i.e. case finding). In addition,
Knowler believes screening is clearly indicated for asymptomatic individuals with
conditions often associated with diabetes, such as micro and macrovascular
complications and their determinants, for whom diagnosis and treatment of diabetes
would help in the management of these complications (100).
The available screening tests for NIDDM include questionnaires, biochemical screening
tests and a combination of both. The validity of these tests (sensitivity, specificity,
positive predictive value) has been reviewed by Engelgau et aI. (111). However, the
proper choice for a certain test or combination of tests will depend on the selected
approach (community screening or case finding) and the available resources. The use
of a risk-assessment questionnaire seems reasonable. In the community setting, a
capillary glucose measurement is easily performed and can produce reasonably good
results when adjusted for the postprandial period and age of the person being screened.
Screening for urine glucose, fasting glucose, and glycosylated hemoglobin tend to lack
sensitivity. In the clinical setting, postprandial venous glucose adjusted for the
postprandial period and age may be reasonable. Harris and Modan endorse the OGTT
(a single 2-hr post-challenge glucose measurement) as the primary screening method
because the complexity of this test is more than balanced by its sensitivity, specificity
and positive predictive value (99).
In conclusion, up to now there is insufficient evidence to indicate the value of mass
screening of asymptomatic individuals. However, several organizations have made
recommendations for community screening among high risk groups or in a clinical
setting (11,35,36,112-114), although the performance of the strategies has yet to be
quantitatively evaluated. Operational research is needed to define more clearly the
question as to who should be screened, when, where, and how screening should be
carried out and the effectiveness of sllch screening programmes, taking into account the
available resources. The results of that research will provide insight into possible
consequences as regards the burden on health care (see also below).
114
Implications for health policy
CONSEQUENCES FOR HEALTH
CARI~
Health care for diabetes encompasses a broad spectrum of activities (ranging from
glucose control by diet and antidiabetic medication with appropriate education to
detection of determinants and complications followed by adequate treatment) provided
by a range of health care professionals, slIch as the general practitioner, the internist,
the paediatrician, the diabetic nurse, the dietician, the podotherapist, and
ophthalmologist.
the
All these activities place a heavy burden on health care and health care costs. The
proportion of estimated total health care cost related to diabetes is 3.6% in the United
States of America and 4-5% in the United Kingdom (115). Recent data suggest that in
the United States of America the costs of diabetes care are rising rapidly (116-118). In
the Netherlands the health care costs for diabetes are estimated to be about 1%
(119,120) of total health care costs. However, this estimate only represents the costs
inculTed when diabetes is the primary diagnosis. As diabetes mellitus is often involved
in other health problems (micro and macrovascular diseases), this figure underestimates
the real health care costs (121,122), and taking this into account, the health care costs
in hospitals attributable to diabetes are about twice as high in the Netherlands (123).
The expected increase in the number of patients makes the disease important not only
for prevention but also for planning future health care and allocating health care costs.
At first sight, it seems plausible that an increase in the number of patients will increase
the burden on health care and costs by an equivalent amount. However, results from
empirically-based studies do not confIrm this. For instance, in the Netherlands it
appeared that between 1980 and 1991 the number and duration of hospital admissions
due to IDDM among those < 20 years decreased more than 30% and nearly 55%,
respectively (124), while an increase in incidence of 23% was noted in the same period
(8). Apparently, there is a shift from costly in-patient to relative inexpensive out-patient
care, probably due to improved care, diabetes education and self-management.
Therefore, these kinds of health care developments as well as the policy of reducing
budgets and capacity and increased productivity need to be taken into account and may
be responsible for changes in the burden on certain health care provisions and health
care costs (125-129).
As stated earlier, the real extent of the increase in diabetic patients expected over the
next decades will very much depend on the policy for detecting undiagnosed
individuals. This raises the question whether these earlier diagnosed patients have a
115
Chapter 7
different health status that needs more or less health care than those not detected by
screening. Determinants for complications as well as the prevalence of macrovascular
complications in undiagnosed NIDDM are very common and are as frequent as in
diagnosed N1DDM. However, some of these patients may already be receiving medical
attention because of these determinants and complications without being aware of the
existence of diabetes. Hence, the burden on health care will not increase as much as
expected. If the detected patients are treated adequately as early as possible, it is
believed that their health status will be less impaired in the long run compared with
patients who were diagnosed and treated several years after the age at onset. As a
consequence the burden on diabetes health care may eventually decrease.
In conclusion, it is not easy to foresee what the burden on health care will be in the
near future as a result of changes in the occurrence of diabetes. It is very likely that
the number of patients will increase, irrespective of developments in the field of
primary prevention. However, in addition to the number of patients, changes in the
health status (as a result of possible secondary preventive and health care activities) as
well as developments in health care (such as a shift from in-patient to out-patient
treatment, the policy of reducing budgets and capacity, and increased productivity) will
influence the burden on health care and the concomitant costs for diabetes.
IMPLICATIONS FOR HEALTH POLICY AT THE NATIONAL LEVEL
The recognition of diabetes as a major public health problem and the notion that
lDDM and NIDDM will become even more prominent in the future makes the disease
important for health policy. World-wide, a great deal of research focused on finding
tools for preventing and controlling diabetes. What are the implications of the results
of this research for health policy at the national level?
Two ways to prevent lDDM from occurring have been described: (I) by reducing
exposure to the determinants and (2) by means of active intervention in the 'prediabetic' stage with pharmacological agents. Up to now, primary prevention of lDDM
is still confined to research without practical implications. Some large multinational
trials are under way. Although it seems likely that pharmacological interventions may
contribute to the prevention of IDDM in the future, epidemiological studies designed to
search for determinants need to be encouraged. At the national level, an important
contribution can be made by joining the international network of childhood diabetes
registries (WHO Multinational Project for Childhood Diabetes (WHO DIAMOND))
116
Implications for health policy
(130). Observed differences in incidence over time and between countries will be
helpful in the search for environmental determinants for IDDM.
As regards the two ways of primary prevention for NIDDM, there is some cause for
optimism. [n contrast to IDDM, determinants for NIDDM have been identified by
epidemiological studies. Unfortunately, few empirically-based studies have been
published that examine the effect of interventions on the established determinants with
regard to the development of NIDDM. Therefore, additional clinical trials are needed
to test this possibility. Meanwhile, the experience gained from the prevention of
cardiovascular diseases can serve as an example for primary preventive strategies for
NIDDM (131). In line with the recommendations of the WHO (132), the Dutch
National Diabetes Platfonn encourages an integrated approach to the primary
prevention of NIDDM (by means of health promotion related to lifestyle), together
with
other
non-communicable
diseases
with
common
determinants,
such
as
cardiovascular diseases (133). Whether pharmacological interventions can prevent
NIDDM is unknown. The results from the few trials that have been performed are not
encouraging.
Internationally, several organizations have made recommendations for community
screening among high risk groups or in the clinical setting to detect as yet undiagnosed
patients with NIDDM (11,35,36,112-114). It is unlikely that universal guidelines on
secondary preventive strategies will be established because the distribution of high risk
groups, the resources available, and the existing health care structures differ from
country to country. It is therefore strongly recommended that a national strategy be
established, based on knowledge gained internationally. The implementation has to be
taken place in such a way that the results can be evaluated in order to derme more
clearly the who, when, where, and how of screening and the effectiveness of screening
programmes in a community and/or clinical setting. The development of all kinds of
different local programmes to detect diabetic patients without proper evaluation of
costs and health benefits, should be avoided.
Although it is not easy to foresee what the burden on health care will be in the near
future, it is very likely that the number of patients will increase while financial
constraints in health care are becoming increasingly evident. Therefore, considerable
efforts will be needed to find more cost-effective ways to treat diabetic patients and to
reduce the complications associated with the disease. It is encouraging that representatives from all European countries have agreed on the five-year 'St Vincent' targets for
reducing complications caused by diabetes (3). To evaluate progress towards these
117
Chapter 7
targets, it is necessary to assess the Cllrrent situation and to follow the rate at which
complications develop. Unfortunately, reliable data are still lacking in most countries,
including the Netherlands.
It is quite clear that several kinds of developments have influenced and will continue to
influence the disease, possibly with important consequences. In order to gain accurate
and up to date insight into the relative importance of these developments for the
necessary health care facilities and costs, and to determine the feasibility and effectiveness of intervention strategies, it is essential to improve or start monitoring of the
disease.3 Although activities in this field have started in several European countries
(136), including the Netherlands (137,138), there is a need for central coordination at
the national level to set up a coherent monitoring system based on those sub-systems
that already exist and new ones that need to be developed. As diabetes is often
accompanied by other diseases (comorbidity), an integrated approach to monitoring
activities is recommended. Data from such a monitoring system can provide the input
for health models designed to forecast future developments, which may be a valuable
tool for underpinning health policy decisions (139-147).
A great deal of effort from all parties involved in reducing the heavy burden of
diabetes is needed to make well-considered decisions with the help of monitoring and
modelling. For the Netherlands, a key role can be played by the Netherlands Diabetes
Federation, in which the parties involved in diabetes care and research have joined
together in 1995 (148). Just as the results of world-wide research activities as well as
efforts from WHO have been of great help for the development and implementation of
national programmes to prevent and control diabetes, experience gained at the national
level may be helpful to other countries wishing to combat the global threat of diabetes.
REFERENCES
I.
2.
WHO Study Group on Diabetes Mellitus. WHO Technical Report Series 727. Geneva: World Health
Organi7-<'ltioll, 1985.
World Health Organization. Prevention and control of diabetes IllClliIUS. In: Handbook of Resolutions
Vol. 3, 2nd ed. (1985-1989). Geneva; World He~'lIth Organization, 1990, p.71.
J He.alth monitoring is defined here as the continuous col!edion, analysis, interpretation and dissemination of data on
individuals or groups to detect the occurrence of cert.1in events and their put.1tive causes for the purposes of planning and
evaluating of inter.·cntions (adapted from Foege et aI., 1976 (134), and Thacker et al.,1988 (135».
liS
Implications for health policy
3.
4.
5.
6.
7.
8.
9.
lO.
II.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
Krans HMJ, Poria M, Keen H (eds.). Diabetes care and research in Europe: the St Vincent Declaration
action programme. Copenhagen: World Hcallh Organization. Regional Office for Europe. 1992.
World Health Organization. A better life for diabetics: a model for quality development. Copenhagen:
World Health Organization, Regional Office for Europe, 1991.
Reiber G. King H. Guidelines for the development of a national programme for diabetes mellitus.
Geneva: World Health Organization, Division of Noncommunicable Diseases and Health Technology,
1991.
King H, Gruber \Y, L1nder T (cds.). Implementing national diabetes programmes. Geneva: World
Health Organization, Division of Noncommunicable Diseases, 1995.
McCarty D. Zimmet P. Diabetes 1994-20lO: global estimates and projections. Melbourne. Australia:
International Diabetes Institute. 1991.
Ruwaard D, Hirasing RA, Reeser mi, Buuren S van, Bakker K, Heine R], Geerdink RA, Bruining GJ,
Vaandrager GJ, Verloove-Vanhorick SP. Increasing incidence of type I diabetes in the Netherlands: the
second nationwide study among children under 20 years of age. Diabetes Care 1994; 17: 599·601.
Ruwaard D, Gijscn R, Bartelds AIM. Hirasing RA, Verkleij H, Kromhou! D. Is the incidence of
diabctes increasing in all age groups in the Netherlands? Results of the second study in the Dutch
Sentinel Practice Network. Diabetes Care 1996; 19: 214-218.
Ruwaard D, Hoogenveen RT, Gijscn R, Verkleij H, Casparie AF, Kromhou! D. Forecasting the number
of diabetic patients in the Netherlrulds in 2005: an update (submitted/Chapter 6).
Report of a WHO Study Group. Prevention of diabetes mellitus. WHO Technical Report Series 844.
Geneva: World Health Organization, 1994.
Taylor WR. Marks JS. Livengood JR. Koplan JP. Current issues and challanges in chronic dise..1se
control. In: Brownson Re, Remington PL, Davis JR (cds.). Chronic disease epidemiology ruld control.
Washington: American Public Hea1th Association, 1993. pp. 1-18.
Stem MP. Primary prevention of type II diabetes mellitus. Diabetes Care 1991; 14: 399-410.
Hennan WH, Sinnock p, Brenner E, Brimberry JL, Langford D, Nakashima A, Sepe SJ. Teutsch SM,
Mazze RS. An epidemiologic model for diabetes mellitus: incidence, prevalcnce and mortality. Diabetc.1O
Care 1984; 7: 367·371.
Laakso M. Pyorilla K. Age of onset and type of diabetes. Diabetes Care 1985; 8: 114-117.
Eisenbarth GS. T)Te I diabetes mellitus: a chronic autoimmune disease. N Engl J Med 1986: 314:
1360·1368.
Krolewski AS, Warram JH, R1nd LI, Kahn CR. Epidemiologic approach to tile etiology of type I
diabetes mellitus and its complications. N Engl J Med 1987; 317: 1390-1398.
Atkinson I\'IA, Maclaren NK. The pathogenesis of insulin-dependent diabetes mellitus. N Engl J Mcd
1994; 331: 1428·1436.
Palmer JP. Predicting IDDM-1991. Diabetes Reviews 1993; 1: 104-115.
Bruining GJ, Grobbce DE, Scheffer e], Bruyn AMD. Molenaar JL, Hofman A. Ten-year follow-up
study of islet·cell antibodies ruld childhood diabetes mellitus. Lancet 1989; 333: 1100-1103.
Lebovitz HE. Ethiology and pathogenesis of diabetes mellitus. Paediatrics Clinics of North America
1984; 31: 521·530.
Olmos P, A'Hem R, Heaton DA, Millward BA. Risley D, Pyke DA, Leslie RDG. TIle significance of
concordance rate for type I (insulin-dependcnt) diabetes in identicaJ twins. Diabetologia 1988; 31; 747750.
Skyler J. Immune intervention in Type I diabetes mellitus. Diabetes/Melabolism Reviews 1993; 1: 15·
42.
119
Chapter 7
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
42.
43.
44.
45.
120
Nerup J, Mandrup-Poulsen T, Molvig J. TIle HLA-IDDM association: implications for etiology and
pathogenesis of IDDM. Diabetes/Metabolism Reviews 1987; 3: 779-802.
Nepom QT, Ehrlich, H. ?\'1HC Class II molecules and autoimmunity. Annual Rev hnmunol 1991~ 9:
493-525_
Gamble DR. TIle epidemiology of insulin-dependent diabetes, with particular reference to the
relationship of virus infection to its etiology. Eridemial Rev 1980; 2: 49-70.
GI~1Son RE. Kahn CB, Funk IB, Seasonal incidence of insulin-dependent diabetes (lOD11) in
Massachusetts, 1964-1973.1nl J Epidemial 1982; 11: 39-45.
Borch-Johnsen K. Joner G. Mandrup-Poulsen T. Cluisty M, Zachau-Christiansen B, Kastrup K, Nemp
J. Relation between breast-feeding and incidence of insulin.dependent diabetes mcllitus. Lancet 1984; 2:
1083-1086.
Dahlquist G, Blom LG. Persson L, Sandstrom AIM, Wall SGI. Dietary factors and the risk of
developing IDDM in childhood. Brit Med J 1990; 300: 1302-1306.
Virtancn SM, Rlisanen L, Aro A, Lindstrom J, Sippola H, Lounamaa R, Toivanen L, Tuomilehto J,
Akerblom HK. Infant feeding in Finnish children < 7 yr of age with newly diagnosed IDDM. Diabetes
C!lJC 1991; 14: 415417.
Norris JM, Scott FW. A meta-analysis of infant diet and insulin-dependent diabetes mellitus: do biases
playa role? Epidemiology 1996; 7: 87-92.
Diabetes Epidemiology Research International (DERI). Preventing insulin dependent diabetes meUitus:
the environmental challenge. BMJ 1987; 295: 479-481.
Muir A, Schatz D. Pccvention of insulin·dependent diabetes-1995. Trends Endocrinol Mctab 1995; 6:
312-317.
Schernthaner G. Progress in the immunointervention of type-l diabetes mellitus. Honn Metab Res
1995; 27: 547-554.
American Diabetes Association. Screening for diabetes (position statement). Diabetes Care 1989; 12:
588-590.
Paterson KR. Population screening for diabetes mellitus. Diabetic Med 1993; 10: 777-781.
Reaven GM. Role of insulin resistance in human disease. Diabetes 1988; 37: 1595-1607.
Zimmet P. Challenges in diabetes cpidemiology - from West to the rest. Diabctcs Care 1992; 15: 232252.
Zimmet P. TIle epidemiology of diabetes mellitus and associated disorders. In: Alberti KGMM, Krall
LP (cds.). TIle Diabetes Almua1!6. Amsterdam: Elsevier, 1991: 24: 1-19
Tuomilehto J, Knowler WC, Zimmet P. Primary prevention of non-insulin-dependent diabetes mellitus.
DiabetesIMetabolism Reviews 1992; 8: 339-353.
Tuomilehto J, Tuomilehto-Wolf E, Zimmet P, Alberti KG11M, Keen H. Primary prevention of diabetes
mellitus. In: Alberti KGMM, DeFronzo RA, Keen H, Zimmet P (eds.). International textbook of
diabetes mellitus. London: John Wiley & Sons Ltd, 1992: 1655-1673.
Zimmet P. The pathogenesis and pccvenlion of diabetes in adults: genes, autoimmunity, and
demography. Diabetes Care 1995; 18: 1050-1064.
Helmrich SP, Ragland DR, Leung RW, Paffenbarger RS. Physical activity and reduced occurrence of
non-insulin-dependent diabetes mellitus. N Engl J Med 1991; 325: 147·152.
Feskens ElM. Nutritional factors and the etiology of non-insulin-dependent diabetes mellitus: an
epidemiological overview. World Rev Nutr Diet 1992; 69: 1~39.
Modan M, Karasik A, Halkin H. Fuchs Z, Lusk.1' A, Shimt A, Modan B. Effect of past and concurrent
body mass index on prevalence of glucose intolerance and type 2 (non-insulin-dependent) diabetes and
Implications for health policy
46.
on insulin response: The Israel study of glucose intolerance, obesity and hypertension. Diabetoiogia
1986; 29: 82-89.
Everhart JE, Pettitt DJ, Siaine KR. Knowler We. Duration of obesity is a risk factor for non-insulin-
dependent diabetes mellitus. Am J Epidemiol 1986; 124: 525.
47. Ohlson L-O, Larsson B, Svardsudd K, Welin L, Eriksson H. Wilhelmsen L, Bj5mtorp P, Tibblin O.
The influence of body fat distribution on the incidence of diabetes mellitus. 13.5 years of follow-up of
the participants in the study of men rom in 1913. Diabetes 1985; 34: 1055·1058.
48. Voedingsraad. Advies Richtlijnen Goede Voeding. 's-Gravenhage: Voedingsmad, 1986.
49. Papoz L, Eschwcgc E Warne! J-M. Richard J-Lt Claude J-R. Incidence and risk factors of diabetes in
the Paris Prospective Study (GREA). In: Eschwcge E (ed.) Advances in diabetes epidemiology.
Amstcrdmn: Elsevier Biochcmical Press, 1982, 113·122.
50. Stmnpfer MJ. Colditz GA, WilIctt WC. Manson lE, Arky RA, Hennckens CH, Speizer FEA. A
prospective study of moderate alcohol drinking and risk of diabetes in women. Am J Epidemiol 1988;
128: 549-558.
51. Holbrook TI.., Barrett·Conner E, Wingard DL. A prospective population-based study of alcohol use and
non-insulin-dependent diabetes mellitus. Am J Epidemiol 1990; 132: 902-909.
52. Feskens ElM, Kromhout D. Cardiovascular risk factors and the 2S·ycar incidence of diabetes mellitus.
The Zutphen Study. Am J Epidemiol 1989; 130: 1101-II08.
53. Rimm EB, Chan J, Stampfer MJ, Coldilz GA, Willett WC. Prospective study of cigarette smoking.
alcohol use, and the risk of diabetes in men. B:MJ 1995; 310: 555-559.
54. Perry H, Wannamethee SG, Walker MK, Thomson AG, Whincup PH, Shaper AG. Prospective study of
risk factors for developing of non-insulin dependent diabctes in middle aged British men. BMJ 1995;
310: 560-564.
55. Mcdalie JH, Papie CM, Goldbourt V, Hennan JB. Major factors in the development of diabetes mellitus
in 10,000 men. Arch Intern Med 1975; 135: 811-87.
56. Wilson PWF, Anderson KM, Kannel WB. Epidcmiology of diabetes in the elderly. Am J Med 1986;
80(suppl 5A): 3-9.
57. Rimm EB, Manson JE. Stampfer MJ, Colditz GA, Willct WC, Rosner B, Hennekens CH, Speizer FE.
Cigarette smoking and the risk of diabetes in women. Am J Public Health 1993; 83: 211-214.
58. Manson JE, Nathan DM, Krolewski AS, Stampfer MI, Willett WC, Hennekens CR. A prospective
study of exercise and incidence of diabetes muong US male physicians. JAMA 1992; 268: 63·67.
59. Eriksson KF, Lindgtirde F. Preyention of type 2 (non·insulin-dependent) diabetes mellitus by diet and
physical exercise. The 6-year Malmo feasibility study. Diabetologia 1991: 34: 891-898.
60. Pan X, Li G, Hu Y. Zuoxin A, Jixing W. Bennett PH, Howard BY. Effect of dietary and/or exercise
interventions on incidence of diabetcs in 530 subjects with IGT - the Da-Qing IGT and diabetes study.
Jnpan: Proceedings of the 15th International Diabetes Federation Congress. 1994. p,489.
61. Seidell JC. Obesity in Europe: scaling an epidemic. Int JObes 1995; 19(supp13): SI·S4.
62. Jarrett RJ, Keen H, Fuller JH, McCartney M. Worsening to diabetes in men with impaired glucose
tolerance ("borderline diabetes"). Diabctologia 1979; 16:25·30.
63. Sartor G, Schersten B, Carlstrom S, Melander A, Norden A, Persson G. Ten-year follow-up of subjects
with impaired glucose tolerance: prevention of diabetes by tolbutamine and diet regulation. Diabctes
1980; 29: 41-44.
64. Keen H, Jarrett RJ, McCartney P. The lO-year follow-up of the Bedford survey (1962-1972): glucose
tolerance and diabetes. Diabetologia 1982; 22: 73-78.
t
i21
Chapler 7
65.
66.
67.
Knowler we, Narayan KMV, Hanson RL. Nelson RO. Bcnnett PH, Tuomilclho J. Schersten n, Pettitt
DJ. Preventing non-insulin-dependent diabetes. Diabetes 1995:44: 483-488.
Morrison AS. Screening in chronic diseases. 2nd ed. New York: Oxford University Press, 1992.
Boutcr LM, Dongcn MCJM van. Epidemiologisch ollderLOck. Opzel en inlerpretalie. Utrecht: Bohn.
Scheltema en Holkema, 1988.
68.
69
70.
71.
72.
73.
Collette IDA, Verbeek ALM. Secundairc prevcntie. in het bijzonder het bevolkingsondcrzoek. In:
Epidemiologic en gezondheidsbeleid. Mans PI van der, Hofman A. Dekker E (red.). Alphen aan den
rujn: Samson Staflcu, 1989, pp 254-262.
Crommc PVM. Glucose tolerance in a typical Dutch community. Thesis. Free University Amsterdam.
Amsterdam, the Netherlands: Free University Amsterdam, 1991.
Feskens ElM, Bowles CH, Kromhout D. Intra- and interindividual variability of glucose tolerance in
an elderly population. J Clin Epidemio1 1991; 44: 947·953.
Mooy 1M, Grootenhuis PA, Vries H de, Valkenburg HA. Bouter LM, Koslense PJ. Heine RJ.
Prevalence and delenninanfs of glucose intolerance in a Dutch Caucasian JXlPulation. The Hoom Study.
Diabetes Care 1995; 18: 1270-1273.
Stalk RP, Pols HAP, Lamberts SWJ, Jong PTVM de, Hofman A, Grobbce DE. Diabetes mellitus,
impaired glucose tolerance and hyperinsulinemia in an elderly population. The Rotterdam Study. Am J
Epidemiol1996; 143: (accepted).
Harris 1\'11, Hadden WC, Knowler WC, Bennett PH. Prevalence of diabetes mellitus and impaired
glucose tolerance and plasma glucose levels in U.S. population aged 20-74 yr. Diabetes 1987; 36:
523-534.
74. Sioger DE, Samet JH, Coley CM, Nathan DM: Screening for diabetes mellitus. Ann Intern Med
109:639-649. 1988
75. Steering Committee on Future Health Scenarios: Chronic Diseases in the year 2005. Volume 1;
Scenarios on Diabetes Mellitus 1990-2005. Dordrecht-Boston-London, Kluwcr Academic Publishers,
1991.
76.
Herron CA. Screening in diabetes mellitus; rCpJrt of the Atlanta Workshop. Diabetes Care 1979; 2:
357-362.
77.
78.
79.
80.
81.
82.
83.
84.
122
Wilson JHG. Jungner G. Principles and practice of screening for disease. WHO. Geneva, 1968.
Harris 1\'11, Klein RE, Welborn TA. Knuiman MW. On.<;et of NIDDM occurs at lea<;t 4-7 years before
clinical diagnosis. Diabetes Care 1992; 15: 816-822.
Klein R, Klein BEK, Moss SE. Davis i\ID, DeMets DL. TIle Wisconsin Epidemiologic Study of
Diabetic Retinopathy. I1I. Prevalence WId risk of diabetic retinopathy when age at diagnosis is 30 or
more years. Arch Ophthalmol 1984; 102: 527-532.
Dolben J, Owens DR, Young S, Vora J. Atiea J, Dean J, Luzio S. Retinopathy at presentation in t)1)e II
(non·insulin-dependent) diabetic patients. Diabetic Med 1988; 5 (Suppl. 2): 20.
AJdington SJ, Kohner EM, Nugent A. Retinopathy at entry in the United Kingdom Prospective Diabetes
Study (UKPDS) of maturity-onset diabetes. Diabetic Med 1987; 4: 355.
Haffner SM, Fang D, Stern rvw, Pugh JA, Hazuda HP, Patterson JK. Heuven WN van, Klein R.
Diabetic retinopathy in Mexican Americans and non-Hispanic whites. Diabetes 1988; 37: 878-884.
Fabre J. Balant LP. Dayer PG, Fox HM. Vernet AT. TIle kidney in maturity·onset diabetes mellitus: a
clinical study of 510 patients. Kidney Jnt 1982~ 21: 730-738.
Uusitupa M, Siitoncn 0, Penllila I, Am A, Pyorala K. Proteinuria in newly diagnosed type II diabetic
patients. Diabetes Care 1987: 10; 191-194.
Implications for health policy
85.
Haffner SM, Mitchell BD, Pugh JA, Stem :MP, Kozlowski MK. Hazuda HP, Patterson JK, Klein R.
Proteinuria in Mexican Americans and Non-Hisp..'1nic whites with NlDDM. Diabetes Care 1989; 12:
530-536.
86.
87.
88.
Harris 111. Impaired glucose tollerance in the U.S, popuJalion. Diabetes Care 1989; 12: 464-474.
Harris 1"'11. Undiagnosed NIDDM: clinical and public hca1lh issues. Diabetes Care 1993; 16: 642-652.
Haffner SM. Stem MP, Hazuda HP, Mitchell BD, Patterson JK. Cardiovascular risk factors in
confirmed prediabetic individuals: docs the clock for coronary heart disease start licking before the
onset of clinical diabetes? JAMA 1990; 263: 2893-2898,
Grootcnhuis PA. Epidemiological aspects of the insulin resislence syndrome. TIle Hoom Study. Thesis.
Free University Amsterdam. Amstcrdrun. the Netherlands: Frec University Amsterdam, 1994.
90. Heine RJ, Mooy JM. Impaired glucose tolerance and unidentified diabetes. Postgrad Med I 1996; 72:
467-471.
89.
91.
92.
93.
94.
95.
96,
97.
98,
99.
100.
101.
102.
103.
Manson JE, Coldilz GA, Stampfer l\.1J, Willett WC, Krolewski AS, Rosner B, Arky RA, Speizer FE.
Hennekens CH. A prospectivc study of maturity·onset diabetes mellitus and risk of coronary heart
disease and stroke in women. Arch Intern Med 1991; 151: 1141·1147.
Stamler J, Vaccaro 0, Neaton JD, Wentworth D. Diabetes, other risk factors, <Uld 12·year
cardiov<'lsculnr mort.'llity for men screened in thc multiple risk factor intervention trial. Diabetes Care
1993; 16: 434-444.
Eschwcge E, Richard IL, Thibuit N, Ducimeticrc P, Warnet 1M, Claudc JR, Rosselin GE. Coronary
heart disease mortality in relation with diabetes, blrnxl glucose, and plasma insulin Icvels. the Paris
Prospective Study ten years later. Honn Metab Res 1985; 15(suppl): 41-46.
Jarrett RJ, Shipley MJ. Type II (non-insulin dependent) diabetes mellitus and cardiovascular disease putative association via common antecedents; further evidence from thc Whitehall Study. Diabetologia
1988; 31: 737-740.
Reichard P, Nilsson BY, Rosenqvist U. The effect of long-tenn intensified insulin treatment on the
development of microvascular complications of diabetes mellitus. Ncw Engl I Med 1993; 329: 304-309.
The Diabetes Control and Complications Trial Reserach Group, The effect of intensivc treatment of
diabctes on the dcvelopment and progression of long·term complications in insulin-dependent diabetes
mellitus. N Engl J Med 1993; 329: 977-986.
Eastman RC, Siebert CW, Harris M, Gorden P. Implications of the Diabetes Control and Complications
Trial. I Clin Endocrinol Metab 1993; 77: 1105·1107.
American Diabetes Association. Implications of the Diabetes COlltrol and Complications Trial, Diabetes
Care 1993: 16: 1517-1520.
Harris MI. Madan M, Screening for NIDDM, Why is there no national program? Diabetes Care 1994;
17: 440·444.
Knowler WC. Screening for NlDDM, Opportunities for detection, treatment. and prevention, Diabetes
Care 1994; 17: 445-450.
Ohkubo Y, Kishikawa H. Araki E, Iv1iyata T, Isruni S, Motoyoshi S, et aI. Intensive insulin therapy
prevents the progression of diabetic microvascular complications in Japanesc patients with non-insulin·
dependent diabctes mellitus: a randomized prospectivc 6-year stuidy. Diabetes Research and Clinical
Practice 1995; 28: 103-117.
Klein R, Klein BEK, Moss SE, Relation of glycemic control to diabetic microvascular complications in
diabetes mellitus. Ann Intern Med 1996; 124: 90·96,
Nathan OM. Inferences and implications. Do results from the Diabetes Control and Complications Trial
apply in NlDDM? Diabetes Care 1995; 18: 251-257.
123
Chapter 7
104. Colwell JA. The fc..1Sibility of intensive insulin management in non-insulin-dependent diabetes mellitus.
Implications of the Veterans Affairs Cooperative Study on glycemic control and complications in
NIDDM. Ann Intern Med 1996; 124: 131-135.
105. Abraira C, Colwell lA, NuUaJl FQ, Swain CT, Nagel NJ, Comstock JP, el aI. Veterans Affairs
Cooperative Study on glycemic control and complications in type U diabetes (VACSDM). Diabetes
Care 1995; 18: 1113-1123.
106. Genuth S. Exogenous insulin administration and cardiovascular risk in non-insulin-dependent diabetes
mellitus. Ann Intern Med 1996; 124: 104-109.
107. Stem MP. Do insulin-dependent diabetes mellitus and cardiova')cular disease share common
antecedents? Ann Intern Med 1996; 124: 110-116.
108. La.1kso M. Glycemic control and the risk for coronary heart disease in patients with non-insulindcpendent diabetes mellitus. Thc Finnish studies, Ann Intern Med 1996; 124: 127-130.
109. Colwell lA. Forum two: unanswered research questions about metabolic control in non-insulindependent diabetes mellitus. Ann Intern Mcd 1996; 124: 178-179.
110. Turner R, Cull C, Holman R. United Kingdom Prospective Diabetes Study 17: a g-year update of a
randomized. controlled trial on the effect of improved metabolic control on complications in noninsulin-dependent diabetes mellitus. Ann Intern Med 1996: 124: 136-145.
111. Engelgau l>.1M, Aubert RE, Thompson TJ, Hennan WHo Screening for NIDDM in nonpregnant adults,
A review of principles, screening tests, and recommendations. Diabetes Care 1995; 18: 1606-1618,
112. Canadian Task Force on the Periodic Health E,;amination. The perlcxJic healOl examination. Can Med
Assoc J 1979; 121: 1193-1254.
113. US Preventive Services Task Force. Report of the U.S. Preventive Services Task Force: Guide to
clinical preventive services. Baltimore."MD: Williams & Wilkins. 1989.
114. Hayward RSA, Steinberg EP, Ford DE, Roizen MF, Roach KW. Preventive care guidclines: 1991. Ann
Intern Mcd 1991; 114: 758-783.
115. Laing W, Williams R. Diabetes, a model for health care managemcnt. Londen: Office of Health
Economics, 1989. paper no. 92.
116. Huse DM, Ostcr G, Killen AR, Lacey Ml, Golditz GA. The economic costs of non·insulin dependent
diabetes mellitus, JAMA 1989; 262: 2708-2713.
117. Songer T. The economic costs of NTDDM. Diabele..~etabolism Reviews 1992; 8: 389-404.
118. Songer T. The economics of diabetes care. In: Alberti KGM1\.1, DeFronw RA, Keen H, Zimmet P
(eds). International textbook of diabetes mellitus. London: 10hn Wiley, 1992, 1643-1654.
119. Koopmanschap MA, Roijen L van, Bonncux L. Kosten van ziek1cn in Nederland. Rouerdam:
iMGZ/i1\.ITA, Era<;wnus Universitcil, 1988.
120. Casparie AF. In: Steering Committee on Future Health Scenarios: Chronic Disc..1Scs in the year 2005.
Volume 1: Scenarios on Diabetes Mellitus 1990-2005. Dordrecht- Boston-London, Kluwer Academic
Publishers, 1991, 181-188.
121. Gerard K. Donaldson C, Maynard AK. The cost of diabetes. Diabetic Med 1989; 6: 164-170.
122. Jacobs 1. Sena M, Fo,; N. The cost of hospitalization for the late complications of diabetes in the united
stales. Diabetic Mcd 1991; 8: S23·S29.
123. Loy NCML van, Koppeling van epidemiologic en gezondhcidseconomie ten behoeve van de integrale
volksgezondhcidsmodellering. Maastricht: Rijksunivcrsiteit Limburg, 1994.
124. Hirasing RA, Reeser HM, Groot RRM de, RuwtL'lfd D, Buuren S van, Verloove-Vanhorick SP. Trends
in hospital admissions among children aged 0-19 years with type I diabetes in the Netherlands. Diabetes
Care 1996; 19: 431-434.
124
Implications for health policy
125. Ballcgooie E van, Offerman JJG, Casparie AF, Israel-Bultman H. Poliklinische inslelling op insuJine bij
patit!nten mel diabetes mellitus. Ned Tijdschr Geneeskd 1991; 134:68-70.
126. Casparie AF, Hoogendoom D. Effects of budgeting on health care services in DUlch hospitals. Am J
Public H"'IIth 1991; 81: 1442·1447.
127. Ministcrie van WVC, Analyse volume-ontwikkeling zorgsector. Rijswijk: !\.1inisterie van WVC, 1992.
128. Ruwaard D, Kramers PGN. Berg Jelhs A van den, Achterberg PW. Projections of health-care
utilisation. In: Public health status and forecasts: Ihe health status of t1le Dutch population over the
period 1950-2010, The Hague: Sdu Uilgcvcrij, 1994: 118-124.
129. Postma MJ, Ruwaard D, Jager le, Dekkers ALM. Projecting utilization of hospitaJ in-patient days in
the Netherlands: a time-series analysis. IMA Journal of Mathematics Applied in Medicine & Biology
1995; 12: 185·202.
130. World Health Organization DIAMOND Project Group. WHO Multinational Project for Childhood
Diabetes. piabeles Care 1990;13:1062-1068.
131. Re{X'lrt of a WHO Expert Committee. Prevention of coronary heart disease. WHO Technical Report
Series 678. Geneva: World HeaJth Organization, 1982.
132. Shigan EN. Integmtcd programme for noncommunicable diseases prevention and control (NCO). World
HeaJth Statistics Quartrely 1988; 41: 267-273.
133. Nationaal Diabetes Platfonn. Implementatie St Vincent Declaration in Nederland. Amersfoort: Nationaal
Diabetes Platfonn, 1993.
134. Foege WH, Hogan RC, Newton LH. Surveillance projects for selected diseases. Int J Epidemiol 1976;
5: 29·37.
135. Thacker SB, Berkebnan RL. Public health surveillance in the United States. Epidcmiol Rev 1988; 10:
164·190.
136. Piwernetz K, Home PD, Staehr-Johansen K. Kleinebreil L. Venneij D. Stonns GEMG. GaJlego rvtR.
Massi Benedetti M. Cowley C. Diabetes care quality network in Europe. Telematics project within the
SI Vincent declaration. Diab Nutr Melab 1995; 8: 243-249.
137. Nijdam KH. Diabetes Monitoring Sysleem: ceo verkenning van de mogelijkheden voor registratie van
diabcteszorg in Nederland. Amersfoort: Diabetes Vereniging Nederland. 1996,
138. Ruwaard D, Nijdam KH. Heine RJ, Krans HMJ. Ooijen JPG. Stonns GEMG. Diabetes monitoring
system in the Netherlands. Diabetes Research and ClinicaJ Practice 1996 (in press),
139. Manton KG. StaJlard E. Chronic diseases modelling: measurement and evaluation of the risk of chronic
disease processes. London: Charles Griffin & Company Ltd, 1988.
140. Weinstein MC, Coxson PG, Willianls LW, Pass TM. Stason WB, Goldman L. Forecasting coronary
heart disease incidence, mortality and cost: the coronary heart disease policy model. Am J Public
Health 1987; 77: 1417·1426.
141. Gunning-Schepers LJ. The health benefits of prevention, A simulation approach, Thesis. Rotterdam:
Erasmus University Rotterdam, 1988.
142. Ruwu.-ud D, Hoogenveen RT, Vcrkleij H, Kromhout D, Casparie AF. Veen EA van der, Forecasting the
number of diabetic patients in the Netherlands in 2005, Am J Public Health 1993; 83: 989-995.
143. Niesscn LW, Barendregt JJ. Bonneux L. Koudslaal PJ. Stroke trends in an ageing {X'lpulalion, Stroke
1993; 24: 931·939.
144, Crijns H, Diabetic retinopathy. A cost-effectiveness analysis of ophthalmoscopy and photocoagulation,
Thesis. Rotterdam: Erasmus University Rotterdam. 1993.
145. Bonneux L, Barendregt JJ. Meeter K. Boosel G, Maas Pl van der, Estimating clinical morbidity due to
ischemic heart disease and congestivc heart failure, Am J Public Health 1994; 84: 20-28.
125
Chapler 7
146. Boa.. OM. Scenario-analyse economische aspeclen coronaire harlzieklcn. Proefschrift. i\1aastricht:
Rijksunivcrsitcit Limburg. 1994.
147. Javiu Ie, Aiello LP. Cost-effectiveness of detecting and treating diabetic retinopathy. Ann Intern Med
1996;124: 164-169.
148. Nederlandse Diabetes Federalie. Statutell. Amersfoort: Nederlandse Diabetcs Fcdcratie. 1995.
126
CHAPTER 8
General discussion
General discussion
In this thesis the occurrence of diabetes mellitus in the Netherlands, the changes in
recent years and possible future developments have been described. Based on the
current knowledge published in the international literature, the likely implications of
these developments for health policy at a national level have been addressed.
Following a brief recapitulation of the main finding as presented in Chapters 2-7, this
chapter focuses on some methodological considerations regarding the validity of the
selected sources (to assess the occurrence of diabetes), the trend data, and the future
projections. In this light, the importance of monitoring diabetes as well as the health
status of the population in general will be discussed in order to improve the availability
and use of epidemiological data for health policy purposes. Finally, conclusions will be
drawn.
MAIN FINDINGS
The incidence among 0-19 year-aIds was based on a questionnaire survey conducted
among all Dutch paediatricians and internists, while the Dutch Sentinel Practice
Network was used to estimate the incidence from age 20 onwards and the prevalence
for all age groups. As both studies were conducted around 1980 and repeated around
1990, changes over a ten-year period could be obtained. It appeared that the overall
incidence of diabetes mellitus had increased significantly. Age-specifically, this
increase could be attributed to a statistically significant increase in the age groups 0-19
and 45-64 years, respectively. Although the causes for the increase in the youngest age
group are unknown, the increase in the older age group is probably due to earlier
recognition of the signs and symptoms of diabetes followed by blood glucose
measurements and/or to more intensive case finding in general practice. However, the
prevalence of diabetes for those 20 years and older had decreased significantly during
the same period. Apparently, the audit of diabetic care in general practice may have
resulted in a clearance of the diabetic register which in particular may be the result of
changing diagnostic criteria.
Two forecasting studies examined the possible future developments in the occurrence
of diabetes mellitus until 2005. The first study forecasted the number of patients during
the period 1980-2005 based on the incidence and prevalence data from around 1980,
whereas in the second study the data from around 1990 were used for the period 19902005. Both studies revealed that the number of diabetic patients will increase
considerably. However, in view of the existence of a considerable number of
undiagnosed diabetic patients (at least 50%), the real number of known diabetic
129
Chapter 8
patients in the future very much depends on the policy for detecting these patients (e.g.
by case finding Of population-based screening programmes),
The recognition of diabetes as a major public health problem and the notion that it will
become even more prominent in the future makes the disease important for health
policy. Up to now, primary prevention of IDDM has been confined to research without
practical implications. To identify the determinants of IDDM, an important contribution
can be made at the national level by joining the international network of childhood
diabetes registries. Determinants of NTDDM have been identified by epidemiological
studies (overweight, physical inactivity and unhealthy diet), but there are few
empirically-based published studies that have examined the effect of interventions on
these determinants. Because there are many similarities between both diseases, the
experience gained from the prevention of cardiovascular diseases can serve as an
example for primary preventive strategies for NIDDM at a national level. As at least
50% of all patients are undiagnosed, a national strategy based on knowledge gained
internationally is needed to formulate policy for detecting those individuals in a
community and/or clinical setting. As the number of diabetic patients is expected to
increase while increasing financial constraints are being placed on health care, more
cost-effective ways to treat diabetic patients and to reduce the complications associated
with the disease will need to be found. In order to gain insight into the changing
burden of diabetes and to determine the feasibility and effectiveness of intervention
strategies, the disease needs to be monitored.
METHODOLOGICAL CONSIDERATIONS
The validity of the selected sources fot" assessing incidence, prevalence, remission
and mortality
Incidence and prevalence
Two surveys were chosen from 18 wherein the incidence and/or prevalence of diabetes
mellitus had been examined during the period 1971-1987 (Chapter 2). Both have since
been repeated, which made it possible to observe changes over time. As we were
particularly interested in the burden on health care in the Netherlands, the following
basic principles were applied to choose the most appropriate sources: the data should
represent clinically-known patients and not those as yet undiagnosed, and they should
be representative for the Dutch population as a whole in terms of age, gender, degree
of urbanization and geographical variation.
130
General discussion
The incidence among 0-19 year-aids was assessed twice retrospectively (in 1978-1980
and in 1988-1990) by a questiollllaire survey collducted all/ollg all Dutch
paediatricians and internists (Chapters 2 and 3). To obtain incidence estimates that are
less prone to chance, the survey was not only nation-wide but also lasted three years.
To conect for an undercount of cases, the Dutch Diabetes Association was used as a
secondary source for ascertainment. The overall rate of ascertainment (defined as the
proportion of responding patients from the Dutch Diabetes Association who were also
reported by the specialists) was 88% and 81 % in the two studies, respectively.
However, age-specific analysis revealed that the rate of ascertainment was relatively
low for the age category 15-19 (79% and 55% in the fIrst and second survey,
respectively) in contrast with the age category 0-14 years (89% and 90%). Apparently,
the primary source (the specialists) used to assess the incidence for those aged 15-19
years was rather poor and a substantial conection of incidence is needed because of
ascertainment bias. As the majority of all patients aged 15-19 had been recorded by
internists and the rate of ascertainment for intemists was worse than that of
paediatricians (a total of 75% and 94% in the fIrst versus a total of 54% and 91% in
the second study, respectively), the internists in particular seemed to be less reliable as
a primary source. Therefore, the application of the capture-recapture census method by
using a secondary source for ascertainment is strongly recommended in these kinds of
studies (1-4).
The incidence (especially from age 20 onwards) and the prevalence of clinically-known
diabetes were also assessed twice (around 1980 and 1990) in the Dutch Selltillel
Practice Network (Chapters 2, 4 and 6). [n the Netherlands, general practices are a
very useful source for gaining insight into the morbidity patterns of the population. In
the Dutch health care system everyone has their own general practitioner, who operates
as a 'gatekeeper'. This implies that health problems will fIrst be presented to the
general practitioner and that no patient will visit a specialist without being refelTed by
his or her general practitioner. In addition, the speciaHst infonns the general
practitioner about clinical or policlinical findings (such as diagnosis and laboratory
results). However, it should be emphasized that the morbidity patterns registered in
general practice specifically reflect the health problems presented by those who make
an appeal to the health care system. Hence, those who have undiagnosed diabetes will
be missed and people living in institutions for a long time (e.g. nursing homes) and
relying on institutional doctors will also be missed. This may result in an
underestimation of the incidence and prevalence when diabetes is relatively more
prominent among such people.
131
Chapter 8
To obtain valid incidence estimates which are less prone to chance, a sufficiently large
population is needed. The Dutch Sentinel Practice Network, consisting of about I % of
the Dutch population, has the largest denominator of all Dutch continuous morbidity
registrations in primary care. Nevertheless, the incidence has been recorded for several
years in both periods to increase the denominator even more. Although the incidence
data found in the Dutch Sentinel Practice Network are likely to be reliable (the
network has been in operation for a long period of time and the general practitioners
who participate in it are not only highly motivated but also experienced in recording
health problems), a few cases might have been missed at the end of the recording
period because of a delay in transferring information from the specialist to the general
practitioner.
In contrast with the nation-wide retrospective study to assess the incidence among 0-19
year-olds, a secondary source for validating the incidence and prevalence in the Dutch
Sentinel Practice Network was lacking. To make use of the Dutch Diabetes Association
as a secondary source would be problematic. The Sentinel Practice Network represents
only I % of the total population instead of 100% in the nation-wide retrospective study,
and the majority of the members of the Dutch Diabetes Association represent young
IDDM patients while in general practice most patients suffer from NIDDM. It would
be more efficient to send a questionnaire to all those belonging to the general practices
to gain information about self-reported incidence and prevalence. However, a basic
premise of the capture-recapture census method is that the secondary source must be
independent of the first one. This is not possible as the general practitioners would
need to be involved in such an activity at the same time as they were recording the
patients. A third possibility that has been considered is to make use of prescription data
from pharmacies. As a nation-wide central drug database of pharmacy dispensing
histories is not available in the Netherlands (5), the Sentinel Stations are distributed all
over the country, and several phannacies are involved per station, we rejected this
option. However, if such a nation-wide central database were available and valid, it
would be very cost-effective to validate morbidity registers in cases of diseases for
which drug prescription is applicable. Such a central database could even be used as
the primary source. With regard to diabetes, this applies to IDDM patients using
insulin and NlDDM patients using insulin or oral hypoglycaemic drugs, but not to
NTDDM patients treated with diet only. Based on the results of the second study in the
Dutch Sentinel Practice Network, approximately 20% of all prevalent patients were
treated with diet only.
132
General discussion
When the most recently obtained incidence (1990-1992) and prevalence data (1990)
from the Dutch Sentinel Practice Network are compared with those obtained from other
studies conducted in general practice among all ages around 1990, the incidence as
well as the prevalence is low in the Practice Network (fable I).
Table 1: Incidence and prevalellce of diabetes mellitus (per 1,000 for all ages) according to
studies COlldlicted ill general practice around 1990.
Study
Period
DSPN'
1990-92
RNH-Limburg
1994
CMR-Nijmegen 1987-91
Transition Project'1985-88
National Survey' 1987-88
sub-project NS
1988
Reenders
1987-88
Verhoeven
1987
Incidenceb Prevalenceb
Population3
Location
140,000
63,000
12,000
41,000
83,000
24,000
42,000
12,000
Netherlands
1.5
Limburg
1.9
Nijmegen
2.2
Neth./Antilles
2.4
Netherlands
3.5
S-E Netherlands
Hoogeveen
Heerde
11.3
24.9
19.4d
11.2
12.3'
14.5
19.0
Source
thesis
6
7
8,9
10
11
12
13
a: rounded numbers.
b: except for the sub-project of the National Survey and Reenders. aU incidence and prevalence figures are
standardized to the Dutch population of 1990.
c: Dutch Sentinel Practice Network.
d: Continuous Morbidity Registration Nijmegen; prevalence is the mean of the known cases recorded per year during
the period 1987-199l.
e: recording period per general practice is one year.
f: 332,000 persons were follO\ved for 3 months.
g: sub-project of the National Survey; prevalence is 7.0 when patients are removed who did not meet the criteria for
diabetes or for whom the diagnosis was not clear.
In general, it is known that differences in morbidity rates between studies conducted in
general practices can be ascribed to discrepancies in the objectives, the design, the
definition of the numerator (such as diagnostic criteria), the extent and definition of the
denominator and the length of the recording period (14). For instance: the Dutch
Sentinel Practice Network is specially designed to obtain incidence and prevalence
figures in primary practice, while other registrations focus more on recording medical
consumption and/or include uncertain diagnoses. Despite the fact that most studies are
too small to obtain reliable incidence data for diabetes, some factors may be
particularly responsible for the observed differences in the incidence and prevalence of
diabetes. The general practitioner's behaviour for detecting as yet undiagnosed diabetic
patients may be of great influence on the recorded incidences, as over 50% of all
133
Chapter 8
patients have undiagnosed diabetes (15-18). Another important factor that may exert
considerable influence on the recorded prevalences is the number of false positive
cases, In some studies, an audit of diabetic care even resulted in a clearance of the
diabetic register by around 25% (15,19)(see also under the heading 'Remission'
below).
As a consequence, a high detection rate for undiagnosed patients by means of case
finding in a general practice in which an audit did not take place will result in a much
higher prevalence compared with the prevalence assessed in a general practice which
has been audited and has a low detection rate for undiagnosed patients. Unfortunately,
these kinds of developments could not be disentangled quantitatively in the Dutch
Sentinel Practice Network. As the contribution of these influencing factors may vary
between different studies (and even within one study on different occasions), it is very
difficult to validate the results of one study with those of others without having
quantitative information about these influencing factors (see also the section on 'Trend
data' below).
Remission
At the time we conducted the background study (Chapter 2), it was decided to assume
that no remission takes place. Infonnation about the chance of remission was lacking
and in so far as real remission takes place (e.g. by means of losing weight) these
'patients' will remain under a certain degree of medical supervision. However, recent
studies indicate that as a result of an audit in general practice a considerable number of
diabetic patients do not meet the 1985 WHO criteria for diabetes mellitus (Chapter 6).
This clearance of the diabetic register may be the result of a real recovery (e.g. by
means of changing lifestyle with loss of weight) or may be due to changing diagnostic
criteria. Although it is not possible to quantify the separate contributions of these
possible factors, the effect of changing diagnostic criteria may have been largely
responsible for the observed decrease in prevalence in the Dutch Sentinel Practice
Network (see the section on 'Trend data' below).
Mortality
For the purposes of making future projections, it was decided that the reduction of life
expectancy as identified in studies from other countries provides a more reliable
measure of the outflow from the diabetic population than the Dutch causes of death
statistics (Chapter 2). Besides, data on life expectancy provides additional information
when compared with mortality data, because the duration of diabetes is also included.
However, it is not clear to what extent these life expectancy data also apply to the
134
General discussion
Dutch diabetes population. In addition, as a result of the changes in diagnostic criteria
for diabetes, the life expectancy that is used is probably too optimistic because the data
stem from three longitudinal studies conducted before 1980 (20-22). According to
Panzram, mortality studies on diabetic populations initiated in the sixties and seventies
have included a considerable percentage of patients with impaired glucose tolerance
(23), who are expected to have a more favourable life expectancy. On the other hand,
the decline in death from cardiovascular diseases (an important cause of death among
diabetic patients) in the Netherlands (24,25) might also have benefitted to diabetic
patients in terms of life expectancy.
Trend data
In order to gain insight into possible changes in the occurrence of diabetes mellitus,
both the survey among all Dutch paediatricians and internists as well as the survey in
the Dutch Sentinel Practice Network have been repeated with a similar design 10 years
later. However, it should be understood that neither survey recorded data on a
continuous basis, which means that two points in time are compared whereas data
regarding the period in between are lacking. It is therefore not possible (0 conclude
that the incidence of IDDM among the age category 0-19 is rising continuously.
However, from combining the incidence studies with the previous study that reported
the increase of IDDM in the 1960-1970 birth cohorts of 18 year-old male anny
conscripts, a sustained increase of IDDM in the Netherlands is suspected (26).
The same uncertainty applies to the significant increase in incidence observed in the
age category 45-64 when comparing the first (1980-1983) and second study (19901992) in the Dutch Sentinel Practice Network. It cannot be rnled out that the increase
in incidence may have started in the late eighties or early nineties. In fact, this seems
rather plausible, as the level of incidence very much depends on the behaviour of the
general practitioners
for detecting the undiagnosed persons.
Recently,
general
practitioners have become very aware of diabetes mellitus as a public health problem
for several reasons:
1. in 1988 the Dutch College of General Practitioners published its standard for
diabetes mellitus type II (i.e. NIDDM)(27);
2. in 1988-1990 the Steering Committee on Future Health Scenarios emphasized the
phenomenon of underreporting diabetes mellitus and the importance of the disease
as a major and growing cause of prolonged ill health and premature mortality and
recommended exploring the possibilities of case finding in general practice among
135
Chapter 8
people over 50 with obesity and/or a positive family history of NIDDM and/or the
existence of complications that might be attributable to diabetes (28);
3. since the early nineties evidence from several Dutch studies has been accumulating
that many individuals in the Netherlands also have undiagnosed disturbances in
glucose metabolism (15-18,29), and in view of these reasons;
4. the study by itself may have influenced the general practitioners' behaviour with
regard to detecting as yet undiagnosed patients, although this does not guarantee a
sustained behavioural change.
To interpret the decreased prevalence observed between 1980 and 1990 numerous
developments need to be taken into account that influence incidence, life expectancy
and remission. Changes in exposure to determinants for diabetes, changes in diagnostic
criteria or the accuracy of canfioning the diagnosis as well as changes in policies for
detecting as yet undiagnosed patients may have influenced the incidence. Changes in
diagnostic criteria as well as improvements in health care may have influenced life
expectancy, whereas changes in diagnostic criteria as well as an audit in general
practice may have influenced the extent of recovery. Unfortunately, these different
kinds of developments could not be disentangled quantitatively in the Dutch Sentinel
Practice Network (see also under the heading 'Incidence and prevalence' above).
Future pl'ojections
It is beyond dispute that valid data are needed to make future projections. However, as
mentioned earlier, there is variation in the incidence and prevalence data when
comparing different studies and reliable Dutch mortality data are lacking. This should
be kept in mind when interpreting the future projections. They do not predict the
number of patients in the true meaning of prediction, but explore possible future
developments according to the assumptions made.
Historic validation and sensitivity analysis
Two validation procedures have been performed to analyze the stability of the dynamic
model (Chapter 5). That is, whether the data on incidence, prevalence and reduction of
life expectancy due to diabetes mellitus and the assumption of no remission result in a
state of relative equilibrium of the dynamic model.
The first validation procedure was a historic simulation of the number of prevalent
cases between 1955 (specific demographic data before 1955 are lacking) and 1980,
136
General discussion
assuming time-independent (Le. constant) relative incidence and reduction of life
expectancy to forecast the 1980 absolute numbers. We compared the calculated
prevalent number of cases with the 1980 data. The historic simulation showed a 10%
higher prevalence in 1980 than the empirical numbers. This difference is statistically
significant and seems plausible; the incidence in 1980-1983 which has been used
throughout the whole simulation period may be too high, as an increase in incidence
has been reported in the literature (26,30-33). Real trend data on incidence, life
expectancy and remission are needed to validate the model characteristics, but these
data are lacking.
The second validation procedure was a sensitivity analysis. We analyzed the impact on
the forecasted prevalent number of cases in 2005 of variations in some main model
parameters, Le. the 1980 prevalence, incidence or reduction of life expectancy data for
diabetic patients (Chapter 5). The sensitivity analysis revealed that the dynamic model
is most sensitive to variations in incidence and moderately sensitive to variations in life
expectancy. The dynamic model was relatively insensitive to variations in prevalence.
This is obvious as all prevalent patients at the onset will have died in the long run.
This finding has been confirmed in the forecasting 'update' study (Chapter 6). Despite
a lower starting prevalence in 1990 in the 'update' study, the number of diabetic
patients in 2005 is in accordance with the findings of both corresponding variants in
the first forecasting study as a result of a higher starting incidence in 1990. This
implies that in particular the incidence data (as well as the life expectancy data) needs
to be valid to make future projections. The validity of these data has been discussed
earlier (see the section on 'The validity of the selected sources .... ' above).
Limitations of the dynamic model used
The dynamic model used focuses on making future projections regarding the number of
diabetic patients. However, this is only a first step towards a useful model for health
policy purposes. Extensions of the model are necessary in two directions. Firstly,
determinants for diabetes need to be incorporated to simulate possible effects on the
incidence of diabetes. Secondly, disease stages with their characteristics (such as
concomitant exposure to determinants and complications) need to be incorporated to
simulate the 'natural' course of diabetes and the possible effects of clinical
interventions to enhance quality of life and life expectancy. The outcomes of such an
extended model may be of great help in planning future health care, as it not only
takes into account the number of patients anticipated, but also takes account of their
health status.
137
Chapter 8
So far, the proposed extensions of the model are limited to diabetes as a major public
health problem per se. However, diabetes mellitus cannot be regarded separately,
because it shares common determinants with other diseases such as coronary heart
disease (e.g. unhealthy diet, physical inactivity and overweight) whereas diabetes itself
is accompanied by determinants (such as hyperglycaemia, hypertension,
dyslipoproteinaemia and overweight) for macrovascular diseases, such as coronary
heart disease, cerebrovascular accidents and peripheral vascular disease and for
microvascular diseases (such as nephropathy and retinopathy, including neuropathy).
Therefore, an integrated approach is needed to judge the benefits of interventions when
there are competing causes of morbidity and mortality. Work has started on developing
such a model. However, up to now the most important limitation has been the lack of
valid data.
FROM EPIDEMIOLOGY TO HEALTH POLICY
As described in the Introduction of this thesis, epidemiology has two main functions in
the policy cycle:
l. to provide data which can be used in the preparation of new policy, and;
2. to provide data which can be used to evaluate current policy (see Figure I in
Chapter I).
Despite the fact that the choices that need to be made in health policy are social or
political in nature, the availability of data from epidemiological research should
actually play a cmcial role in underpinning decision~making. In addition, a
mathematical model in which the data can be incorporated may serve as an important
tool for predicting possible future developments and assessing the effects of health
interventions.
Health policy is defined here as the actions of government, doctors and other players
who aim to maintain and improve the state of health of individuals and the popUlation.
This ultimate goal can be subdivided according to the form of intervention. A widely
used distinction is that of primary and secondary prevention and health care which is
known as tertiary prevention (see Chapter 7 for definitions). As not everyone working
in the field of prevention and health care will be involved in all these activities it
should be realized that in order to make the data useful for health policy purposes they
should be made available to those responsible for the different kinds of intervention.
For instance, health care prov'ders will be more interested in the health status of their
patients in order to control the quality of care or to evaluate health care interventions,
138
General discussion
while community workers will focus more on the health status of the total population
and group interventions. Besides, health care planners will be more interested in
possible changes in the need for health care.
Diabetes mellitus: from epidemiology to health policy
The results of this thesis indicate that diabetes mellitus is a growing public health
problem in the Netherlands. It appeared that in the period between 1980 and 1990 the
overall incidence increased significantly for IDDM as well as for NIDDM. Besides, it
is expected that in the period 1990-2005 the number of known diabetic patients will
increase considerably. These epidemiological findings raise questions for health policy
concerning the possibilities for influencing the expected increase in the number of
patients by means of prevention, and the possible consequences of the expected
increase in terms of the burden on health care. Answers to these questions are
presented in Chapter 7.
With regard to health care policy, it is worth mentioning that representatives of
government health departments and patients' organizations from all European countries
met with diabetes experts under the aegis of the WHO Regional Office for Europe and
the International Diabetes Federation European Region in St Vincent, Italy in 1989.
They unanimously agreed on general goals and five-year targets aimed at creating
conditions in which a significant reduction in this heavy burden on health can be
achieved (34). With respect to complications, the five-year targets are:
I. to reduce new blindness due to diabetes by one third or more;
2. to reduce new end-stage diabetic renal failure by at least one third;
3. to reduce by one half the rate of limb amputation for diabetic gangrene;
4. to cut morbidity and mortality from coronary heart disease in diabetic patients by
vigorous programmes of risk factor reduction and;
5. to achieve pregnancy outcomes in diabetic women that approximates that of nondiabetic women.
However, to evaluate progress towards the targets of the St Vincent Declaration, it is
necessary to assess the current situation and to follow the natural course of the disease
and the rate at which complications develop. Unfortunately, reliable data are still
lacking.
Because of the lack of valid data, it was decided for this thesis to restrict the subject to
the occurrence of diabetes in the Netherlands and its developments over time. The
139
Chapter 8
dynamic model used focuses on future projections regarding the number of diabetic
patients. However, this is only a first step if it is to be useful for health policy
purposes. Monitoring the disease in its different aspects is urgently needed to make
proper decisions and to evaluate the effect of policy decisions and interventions (see
below).
Monitoring diabetes mellitus
It is quite clear that several kinds of developments have influenced and will continue to
influence diabetes mellitus, possibly with major consequences. However, in order to
gain accurate and up to date insight into the relative importance of these developments
for the necessary health care facilities and costs, and to detennine the feasibility and
effectiveness of intervention strategies, it is essential to improve or start monitoring of
the disease. Health monitoring is defined here as the continuous collection, analysis,
interpretation and dissemination of data on individuals or groups to detect the
occurrence of certain events and their putative causes for the purposes of planning (i.e.
policy preparation with subsequent policy development and implementation) and
evaluating of interventions (35,36)(see Figure I in Chapter I).
To ensure that the data gathered is brought to the attention of those responsible for
making health policy decisions, the different aspects to be monitored can be subdivided
chronologically according to the various forms of health policy interventions, i.e.
primary prevention, secondary prevention, and health care (see Figure I in Chapter 7).
For instance, monitoring determinants and disease incidence are related to primary
prevention, whereas monitoring the ratio diagnosed/undiagnosed diabetes is related to
secondary prevention and monitoring complications is related to health care. To
disentangle the influence of several kinds of developments, a coherent monitoring
system (with several subsystems) needs to be set up, which links outcome-specific data
with prevention and control programmes. Attention should not only be paid to
sustaining or improving current monitoring of the incidence, prevalence and mortality
of diabetes, but also to setting up new monitoring systems to gain insight into the
changing consequences of the disease for health status (such as the existence of
complications). A few elements (or subsystems) that need to be considered as essential
components of such a coherent monitoring system for diabetes are briefly summarized
below.
140
General disc liS sian
Monitoring determinants Jor diabetes mellitlls
In order to gain insight into possible future changes in the incidence of diseases,
monitoring determinants in the population may be very helpful. With regard to
NIDDM, it is essential to monitor changes in weight, physical activity and the
composition of the diet. An example of such a monitoring system is the project
'Monitoring Risk Factors and Health in the Netherlands', in which a combined health
interview survey and a health examination survey is conducted annually among 5,000
persons aged 20-59 (37). For those under 20, the 'Child Health Monitoring System'
which is also conducted annually among 5,000 individuals in the Netherlands, is a
useful source in this respect (38). Unfortunately, such a monitoring system for those
above 60 years of age is lacking. Monitori ng the entire age range including an
interview survey combined with a health examination survey would be of great value.
However, it should be realized that additional information (such as the relative risk and
time-lag) from national and international epidemiological research is needed to
transcribe the impact of changes in 'exposure' to these determinants on the incidence
of N1DDM for modelling purposes. Examples of such population-based cohort studies
in the Netherlands are the Hoorn Study (17,39) and the Rotterdam Study (40). In
addition to these cohort studies, follow-up studies linked to cllrrent monitoring systems
are also useful for obtaining additional longitudinal-based information (41).
Monitoring the il/cidel/ce al/d preva/el/ce oj diabetes mellitlls
To obtain reliable incidence figures that are not prone to chance, the incidence of
IDDM among youngsters needs to be examined in the Netherlands at a national level.
Sample surveys would be less appropriate because of their lack of statistical power.
Both incidentally perfonned retrospective studies among all paediatricians and
internists have been replaced since 1993 by a continuous prospective registry among all
Dutch paediatricians to assess the incidence of IDDM among children (42). This
registry will eventually also provide the prevalence data (cumulative incidence), as the
disease is chronic and mortality is low (and can be corrected for). Joining the
international network of childhood diabetes registries (WHO Multinational Project for
Childhood Diabetes (WHO DIAMOND)) is worthwhile (43). Observed differences in
incidence over time and between countries may be helpful in the search for
environmental determinants for IDDM.
In the Netherlands, but also in other comparable countries, the incidence of diabetes
above 14 or 19 years can best be recorded by the general practitioner. General
practices here provide a very useful source for gaining insight into the morbidity
141
Chapter 8
patterns of the population for reasons which have been mentioned earlier (see the
section 'The validity of the selected sources ... .'). Attention should be paid to the
sample size. In our two studies of general practice several years were combined to
assess the incidence more reliably. In order to disentangle the causes of possible
changes in incidence "in general practice. it is recommended to record the reason for
encounter when the diagnosis of diabetes is made (such as symptoms, case finding or
routine examination). However, it should be emphasized that the morbidity patterns
registered in general practice specifically reflect the health problems presented by those
who make an appeal to the health care system. In order to find out the changing
diagnosed/undiagnosed ratio, part of a continuous morbidity registration (physiciandiagnosed cases) should be linked both in time and on an individual level with
intermittently performed population-based (screening) surveys. In addition, it is
recommended to audit the register intermittently to find out whether or not patients
'recovered' from their disease. The prevalence can be assessed intermittently, but
eventually the prevalence can be derived from the cumulative incidence minus those
who recovered from diabetes and those who died.
The importance of a secondary source for ascertainment to correct for an undercount of
cases and the significance of using a central drug database as a primary source to
assess the occurrence of diabetes has been discussed earlier (see the section 'The
validity of the selected sources ... .').
Monitoring the cOllrse of diabetes mellitlls
In order to judge whether certain targets (such as those agreed on in the St Vincent
Declaration with regard to reduction in complications) will be achieved by means of
adequate health care, it is necessary to be informed about the existence of
complications among diabetic patients and their appearance over time. From this
perspective it should be mentioned that the changing occurrence of complications may
not only be dependent on the health care provided. Besides, it is likely that the effects
of activities in the field of secondary prevention will not be confined to changes in the
number of patients, but that the health stalus of the patient may also alter (see Chapter
7). Hence, monitoring characteristics of the diabetic population, such as the mode of
treatment, the glucose levels, and the complication status is necessary.
Although this information can be obtained from a relative small number of patients
involved in population-based cohort studies (such as the Hoorn Study (39) and the
Rotterdam Study (40)), the Dutch National Diabetes Platform started a project with the
aim of exploring the possibilities for setting up a longitudinal and standardized diabetes
142
General discussion
monitoring system in the Netherlands (44,45), with the WHO/IDF-Diabcare basic
information sheet as its point of departure (34,46). According to this sheet, data such as
therapy, glucose levels, the presence of determinants, symptoms and complications
need to be gathered. Such a registry, linked with a registry to monitor the occurrence
of diabetes, will be of great value to guard and improve the quality of care and will
offer possibilities for research activities (intervention and evaluation studies). Activities
aimed at gathering data at a local level, sending them to a central node for analysis,
and providing feedback have been started at both a European and a national level (47).
Monitoring death and diabetes mellitus
As mentioned earlier, the national mortality statistics are not suitable for assessing
mortality among diabetic patients. For that purpose we decided in our study that the
reduction of life expectancy as identified in studies from other countries provides a
more reliable measure of the outflow from the diabetic population. As life expectancy
is not expected to be stable over time and may vary between countries, such studies
are still needed at a national level. However, national data from causes of death
statistics could provide the necessary information if a 'multiple-cause' encoding
procedure were to be introduced. This implies that the presence or absence of diabetes
should be recorded on all death certificates, as has also been recommended for
monitoring the targets of the St Vincent Declaration Action Programme for Europe
(34).
Although some activities in the field of monitoring diabetes have been started in the
Netherlands, there is a need for central coordination to set up a coherent monitoring
system based on several (sub )systems which already exist and others which need to be
developed. A key role can be played by the Netherlands Diabetes Federation, in which
the parties involved in diabetes care and research have joined together in 1995 (48).
From diabetes to an integrated approach: monitoring health in the population
For several reasons diabetes has to be put in a broader perspective. Diabetes shares
common detenninants with other diseases and is itself a determinant for other diseases.
Taking into account these interrelationships, the effect of interventions can be
estimated more accurately. For instance, losing weight in overweight individuals is not
only beneficial for diabetes but also for coronary heart disease. Therefore the total
public health effect will be underestimated if the outcome (decrease in incidence) is
confined to diabetes.
143
Chapter 8
However, it should be realized that overweight is just one determinant for both
diseases, with the consequence that by decreasing weight the onset of disease may only
be postponed. As life expectancy is likely to increase because of preventive activities,
people will be exposed to other diseases which might otherwise not have occurred.
Hence, the phenomenon of competing morbidity and mortality also needs to be taken
into account (49). Therefore, not only the effect of certain interventions for a specific
disease itself, but also for related diseases as well as for diseases which are not
primarily related, need to be taken into account in order to make well-considered
decisions. The epidemiological transition clearly showed a rapid fall in infectious
diseases with an increase in life expectancy unmasking chronic diseases (50).
Therefore, a so-called 'episystems' approach which investigates the processes and
patterns of several diseases instead of one could lead to important insights and will
certainly be important for forecasting future diseases in populations (51,52). This
approach was also the reason that in successive monitoring projects in the Netherlands,
the number of determinants and the number of chronic diseases to be monitored have
been increased (37).
In addition, diabetes needs also to be pnt in a broader perspective from a financial
point of view. In recent years it has become increasingly evident that financial
constraints dictate choices in health policy. The crucial question is then how and in
what areas the greatest health gain (Le. to live lives which are as long and as healthy
as possible) can be achieved with the available resources. This ultimate aim of health
policy distinguishes two elements: extending life expectancy ('adding years to life')
and improving the quality of life or health expectancy ('adding life to years'). The
relationship between life expectancy and health expectancy can be illustrated with the
help of Figure I.
This Figure presents a survival curve (calculated for Dutch males for 1990) in which
the fraction of the relevant birth cohort still alive is given at each age (the outer curve).
The inner curve ('health curve') gives the percentage of the cohort which is in good
health at each age; the difference between the curves (the shaded area) gives the
hypothetical percentage which is in poor health. For females the survival curve lies
further to the right, since women live on average more than 6 years longer than men
(in 1990 the life expectancy of males and females was 73.8 and 80.1 years,
respectively). On the other hand, the inner curve probably does not vary significantly
from that for males. The benefit that females have with respect to total life expectancy
is thus spent primarily in ill-health (53). When life expectancy is extended, the survival
144
General discussion
curve becomes more rectangular (compression of mortality), whereas narrowing of the
shaded area between the curves results in a compression of morbidity.
mates ('%)
q.m
10°I-....
80
60
40
20
°0L---~2~0-----4~0~---760~--~8~0~~10~0~
age
Figure 1: Sun·ival curve in /990 (calculated) alld 'health curve' (hypothetical) for Dutch
males according to age.
Source: Ruwaard el aI., 1993 (53); adapted from ManIon and Soldo, 1985 (54).
From this perspective, four questions are relevant for health policy to make wellconsidered decisions from an integrative viewpoint:
I. Which specific health problems largely detennine life expectancy (survival curve)
and the unhealthy period within it (shaded area)?
2. Which determinants are responsible for these health problems?
3. What are the expectations in this regard in the years ahead?
4. What are the possibilities for improving both life expectancy and the unhealthy
period?
In order to answer these questions a great deal of infonnation is needed. An attempt to
do so for Dutch society has been made in the project 'Public Health Status and
Forecasts. The health status of the Dutch population over the period 1950-2010'. The
basic conceptual model of this project is illustrated in Chapter I (Figure 2). In Figures
2 and 3 below the basic model is shown in more detail for the indicators of health
status and for the determinants, respectively. A more detailed description of these
145
Chapter 8
elaborated submodels with their different layers is given in a report on this project
(53).
Indicators 01 health status
I
I
I
diseases and disorders
functiolling and
I
qualityollife
~
I
I
I
IOOrtaJijy
health and
Ii, expectancy
I
Figure 2: The conceptual model elaborated for the indicators of health status.
Source: Ruwaard el a!., 1993 (53).
A large number of experts contributed to this project in order to quantify health
problems and health risks by using all kinds of existing health monitoring (including
surveillance) and health information systems. Ultimately, these contributions have been
used as building blocks to integrate the material (Le. looking at 'things in context') in
order to answer the questions listed above and to be helpful for health policy purposes.
It appeared that despite the fact that a great deal of information is available in the field
of public health in the Netherlands, there are also particular shortcomings.
146
General disclission
health polIcy
health-eare policy
pre-.enoon polIcy
intersectoral po/;Cy
delecmlnsnts
health status
Figure 3: The cOl/ceptual model elaborated for the determ;l/allfs.
Source: Ruwaard el aI., 1993 (53).
147
Chapter 8
On the basis of the questions raised and the conceptual model used, five levels of
information needs can be distinguished. These levels differ in that they require more
and more coherent llse of data, which means that increasing demands are made on the
quality and structure of data. The levels are:
I. recordillg the situatioll at a specific poillt ill time for illdividual indicators of health
status and determinants;
2. recording trends over time for individual indicators of health status and
determinants (comparability of data over time);
3. simultalleously recordillg series of di/ierellt indicators or determinants (direct
comparison and assessment of combinations);
4. describing relationships between various indicators, between indicators and
determinants, and between variolls determinants Oinking different layers of the
conceptual model);
5. makillg forecasts with the help of modelling (integration of data both within and
between the layers and also over time).
With regard to (I) it can be stated that, for most indicators and detenninants, data is
available on the current situation. This information often falls short, however, as far as
its representativeness of the Dutch population as a whole is concerned. Data on the
incidence and prevalence of diseases has for the most part not been collected at the
population level. Exceptions are the data on IDDM and cancer. The data for incidence
and prevalence has mostly been approximated on the basis of data obtained from
general practitioners or other 'encounter' records, which only contain the presented
morbidity. This means that for some diseases the figures will be underestimations (e.g
for NlDDM). Various information sources also fan short with respect to both the
youngest and the oldest sections of the population, and also where the residents of
institutions are concerned. For the indicators and determinants discussed in the
document 'Public Health Status and Forecasts', for various reasons insufficient data
was available. This is the case, for example, with mental disorders and their
determinants. A number of determinants were also dealt with as a broad group, because
of a lack of an easily measured unit (e.g. exposure to physical factors, the state of the
immune system, or intrinsic ageing).
As far as (2) is concerned, quantitative data on trends over time for individual
indicators alld determillallts is only available for mortality over the whole period
1950-1990. For certain indicators and determinants, information is indeed available
over a shorter period. This trend data comes from regional sources more often than
148
General discussion
data used to describe the situation at a specific point in time. The reason for this
situation is the limited number of monitoring systems which have been in operation for
a fairly long time. Recently, an increase in regularly repeated or continuously recorded
data has been noticed. Where monitoring has taken place, the data collected has
sometimes proved to be of limited use because of changes in the definitions and
measuring instruments used. Such changes may be prompted by new insights, which
does not alter the fact that, for the sake of continuity, efforts must be made towards
standardisation and the comparability of procedures and definitions over time.
With regard to (3), comparison of data on individual indicators or determinants
certainly appears to be possible where only a single source is involved, for example
general practitioners' records, surveys, or specific investigations in which several
indicators and/or determinants are measured simultaneously. In this way, the
occurrence of combinations of indicators and determinants at the personal level can
also be investigated. It often proves difficult to compare data from different sources
because of (sometimes minor) differences in the design of the study or the definitions
used.
As regards (4), information on relationships between diseases and their determinants
does not generally come from recording systems but from cohort studies aimed at
identifying the causes of diseases. There is still a considerable lack of clarity here, for
example with respect to the detenninants of many chronic, non-life threatening
disorders of a somatic or mental nature. There is also too little data available on the
consequences of diseases and disorders for functioning and the quality of life in the
somatic, psychological or social sense (,impact'), and likewise on the non-disease
specific relationships between determinants and quality of life. This means that the link
between the occurrence of diseases and the significance of these for the 'unhealthy
years of life' can hardly be made. Only for the indicators 'invalidity' and 'sickness
leave' and, to a limited extent, for self-rated health is any insight into the share of
underlying (groups of) diseases obtained from records and surveys.
Finally, as regards (5): through modelling of al1 the types of data mentioned (incidence/
prevalence of diseases and disorders, quality of life, mortality, past trends, trends in the
determinants concerned, and interrelationships), this document essentially strives to
make forecasts by estimating expected changes in health status and the effect on these
of possible interventions. A modest start has been made with this in the project 'Public
Health Status and Forecasts'. It will be clear that all the shortcomings in the areas
discussed earlier accumulate here and reinforce each other.
149
Chapter 8
To sum up, it may be said that the ideal monitoring system for providing information
for a forecast should have the following features:
detenninants and indicators of health status are recorded for all layers of the
conceptual model;
recording takes place in a nationally representative way, which also offers insight
into the distribution over various sections of the population;
the recording of data is regularly repeated over time;
linking of individual data is attempted as far as possible.
As long as the wide-scale standardized application of smart cards in health care
remains a dream for the future (55,56), a significant step towards achieving a more
appropriate health information system may be accomplished by improving the links
between existing health interview surveys, epidemiological monitoring programmes,
general practitioners' records, nation-wide health care and insurance systems, and
statistics on causes of death. I However, it would be disappointing if privacy legislation
would hinder research activities and the optimal use of data in order to get insight into
the possibilities to improve life expectancy and the unhealthy part of it (57).
A stepwise approach for tuning monitoring and infonnation systems in order to answer
the above-mentioned questions is urgently needed. After it has been decided what kind
of information is necessary (58, amongst others), an extensive inventory of the existing
monitoring and information systems needs to be available (59). Validation of these
systems according to the five levels of information requirements is important to judge
whether they are appropriate or need to be adapted.
As a general practitioner operates as a 'gatekeeper' in the Netherlands, general practice
is potentially a very useful source for gaining insight into the morbidity patterns of the
population. A continuous morbidity registration of sufficient power in general practices
distributed all over the country, combined with intermittent health interviews and
health examination surveys and linked with other nation-wide health care and insurance
systems would be of great value. This would substantially meet the requirements of an
ideal monitoring and information system mentioned above. For instance, this wil1
provide information on a continuous or regular basis about the Dutch population in
In most cases the information needed to provide a finn basis for aetiological relationships and to enable
model simulations of developments over time to be made cannot be obtained from registration systems; here, cohort
studies are needed, which invol\'e the follow-up of groups of people over time. In addition to these cohort studies,
follow-up studies linked to current monitoring systems are also useful for obtaining additional longitudinal-based
information.
150
General discussion
terms of comorbidity as well as self-perceived health while the physician-based health
status is also known. The flexibility of the system required to anticipate current health
policy issues can be guaranteed by surveys performed intermittently. The National
Information Network of Primary Care which is currently being developed, may be used
for that purpose (60). As regards NIDDM, the essential elements to be monitored, as
mentioned earlier, can be incorporated in tltis approach. Although such a core system
would be of great value, other systems will still be necessary (e.g. registry for lDDM
(42) and cancer (61). However, possibilities for harmonising these systems need to be
considered.
To be successful in setting up a coherent monitoring and information system that will
be useful for health policy, special attention should be paid to the following issues:
All parties involved (i.e. the data suppliers as well as the users) need to be
convinced of the necessity of such an approach. Thacker and Stroup stated that the
major barriers to a successful comprehensive, nation-wide, integrated public health
surveillance and information system are a lack of appreciation for the value of highquality provisional surveillance data and a weak societal commitment to public
health (62).
Major prerequisites if the system is to be useful are that it (I) is flexible and offers
an infrastructure for incorporating new important health policy questions and (2)
reports all requested findings to those involved in making health policy decisions in
a timely manner. Therefore, a well-organized method of data collection, central
analysis as well as dissemination of the results by telecommunication systems are
all necessary. An example of such a surveillance system is the recently developed
Infectious Diseases Surveillance Information System (ISIS) for the Netherlands at
the National Institute of Public Health and the Environment (63).
As regards data collection, standardized measuring methods for each data element
are essential to facilitate analysis and comparison with data collected in other
systems. Although additional assurances of confidentiality and privacy
considerations will be required, the ability to link data to other systems will be of
great value for an integrative approach.
The principles and practice of public health surveillance and health information
systems in the United States are described in detail in the literature (36,62,64).
It should be realized that a great deal of effort from all parties involved is needed to
answer the crucial question as to how and in what areas the greatest health gain (Le. to
live lives which are as long and as healthy as possible) can be achieved with the
available resources. A prerequisite is the existence of a coherent monitoring and
lSI
Chapter 8
information system. This implies national cooperation and the willingness to cut across
disease boundaries. International exchange of experiences in this field (e.g with the
Centers for Disease Control (36,62,64), and the World Health Organization (65,66)) is
strongly recommended. Outside the direct area of health, such as in the field of
environmental outlooks (67,68) and weather forecasts (69), exchange of experience
regarding monitoring and forecasting at a local, national and international level can
also be beneficial.
CONCLUSION
The results described in this thesis indicate that a considerable increase in the number
of diagnosed diabetic patients can be expected. This notion makes the disease
important for health policy. For several reasons (e.g. regarding the validity and
availability of data, and possible developments in preventive and health care policy) it
is not easy to foresee what the exact number of patients and the concomitant burden on
health care and costs in the near future will be. In order to make more accurate
forecasts and to support policy preparation as well as policy evaluation, the disease and
its determinants need to be monitored.
However, it would be inappropriate to confine monitoring to diabetes alone, as changes
in the occurrence or course of diabetes may influence other related and even unrelated
diseases (competing morbidity and mortality). In addition, the available resources are
limited and need to be allocated where the greatest health gain can be achieved. TIns
also applies to resources for monitoring. Therefore, an integrated approach to setting up
a coherent monitoring and information system is strongly recommended.
Data from such a monitoring and information system can provide the input for health
models designed to forecast future developments, which may be a valuable tool that
can be used to underpin health policy decisions. However, a great deal of effort is
required from all parties involved to make well-considered decisions backed up by
sound monitoring and accurate health modelling.
152
General discussion
REFERENCES
I.
2.
Bishop YMM, Fineberg SE, Holland PW. Estimating the size of a closed population. In: Bishop YM:M,
Fineberg SE, Holland PW (eds). Discrete multivariate analysis. 5th cd. Can1bridge. Mass: Massachusetts
Institute of Technology; 1978: 229-256.
Hook EB, Regal RR. TIle value of capture-recapture methods even for apparent exhaustive surveys. Am
J Epidemiol 1992; 135: 1060-1067.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
LaPorte RE, McCarthy D. Bruno G, Tajirna N, Baba S. Counting diabetes mellitus in the next
millennium: application of capture-recapture technology. Diabetes Care 1993; 16: 528-534.
International Working Group for Disease Monitoring and Forecasting. Capture-recapture and multiplerecord systems estimation II: applications in human diseases. Am J Epidemiol 1995: 142: 1059-1068.
Herings R.M:C. PHARMO. A record linkage system for postmarketing surveillance of prescription drugs
in lhe Netherlands. Thesis. Utrecht University. Utrecht: University Utrecht, 1993.
RNH, Registfatienet Huisartspraktijken. Gezondheidsproblemen en diagnosen in de huisartspraktijk,
rapportage nr. I. jaargang 5_ Maastricht: Rijksuniversiteit Limburg, 1993 (en gegevens verkregen op
verzoek).
Lisdonk EH van dc, Bosch \VJHM, Huygcn PJA. Lagro-Jansen ALM. Ziekten in de huisartspraktijk
(Diseases in general practice). Utrecht: Wctcnschappelijke Uitgeverij Bunge. 1990 (en gegevens
verkregcn op verzoek).
Lamberls H. Brouwer ill, Mohrs J. Reason for encounter- & episode- & process-oriented standard
output from the Transition Project. Part 1 and 2. Amsterdam: Department of General Practice, 1991.
OskaIn SK, Brouwer HJ, Molus J. TRANS. an interactional access program for standard output of the
Transition Project. Amsterdam: Department of General Practice, 1994.
Yelden J van der, Bakker DH de. Claes.<;ens AAMC. Schellevis FG. Een nationale studie naar ziek1en en
verrichlingen in de huisartspraktijk. Basisrapport: Morbiditeit in de huis.'lI1spraktijk. Utrecht: NIVEL.
1991 (en gcgevens verkregen op verzoek).
Schellevis FG. Lisdonk EH van de, Velden J van der, Eijk J111M van. Weel C van. Validity of
diagnosis of chronic diseases in general practice. J Clin Epidcmiol 1993; 46: 461·468.
Reenders K, Rutten GEH1\.1, Nobel E de. Hoogen fUM van den. Weel C van. Met de standUs'lfd als
maatstaf. Diagnostiek en behandeling van diabetes mellitus in 19 huisartspraktijken. Huisarts Wet 1990;
33: 379-383.
13. Verhoeven S. Behandeling, controle en metaoole instelling van patienten mel diabetes mellitus type II en
de prevalentie van late complicatics bij deze patienten. Proefschrift. Erasmus Universitei( Rotterdam.
Rotterdam: Erasmus Universitcit, 1989.
14. Meyooom·de long B: Morbidity registration in general practice in the Netherlands. Huisarts &
We.enschap 1993; 36(suppl): 49-53.
15. Cromme PVM. Glucose tolerance in a typical Dutch communily. Thesis. Free University Amsterdam.
Amsterdam. the Netherlands: Free University Amsterdam. 1991.
16. Feskens EJM, Bowles CH, Kromhout D. Glucose tolerance in the elderly. Proceedings of the 32nd
Dutch Federation meeting. Federation of Medical Scientific Societies, 1991: 95.
17. Mooy JM, Grootenhuis PA, Vries H de, Valkenburg HA, Bouler LM. Kostensc PJ, Heine RJ. Prevalence
and detenninallts of glucose intolerance in a Dutch Caucasian population. The Hoom Study. Diabetes
Care 1995; 18: 1270-1273.
153
Chapter 8
18. Stolk RP, Pols HAP, Lamberts SWJ, Jong PTVM de, Hofman A, Grobbee DE. Diabetes mellitus.
impaired glucose tolerance and hypcrinsulinemia in an elderly population. The Rotterdam Study. Am J
Epidemiol 1996; 143: (accepted),
19. Weel C van. Validating 10llg lenn morbidity recording. J Epidemiol Community Health 1995; 49 (suppl
1): 29-32.
20 Marks HH & Krall LP. Onsel, course, prognosis and mortality in diabetes mellitus. In: Marble A. White
P, Bradley RF & Krall LP (cds). Joslin's Diabetes mellitus, 11th cdn. Lea and Feabiger. Philadelphia
1971: 209-254.
21. Goo<lkin G, Mortality factors in diabetes. J Occup Med 1975; 17: 716-721.
22. Panzram G, Zabel-L'Ulghcnnig R. Prognosis of diabetes mellitus in a geographically defined population.
Diabclologia 1981; 20: 587-591.
23. Panzram G. Mortality and survival in type 2 (non-insulin-dependent) diabetes mellitus, Diabetologia
1987; 30: 123-131.
24. Jansen J, Kromhout D, Verkleij H. Coronaire hartziekten. In: Ruwaard D, Kramers PGN. eds,
Volksgezondheid Toekomst Verkenning, De gezondheidstocstand van de Nederlandse hevolking 19502010, National Institute of Public Health and Environmental Proteclion, The Hague, the Netherlands: Sdu
Uitgevenj, 1993: 379-386.
25, Jansen J, Keli S, Kromhout D, Cerebrovasculairc aandoeningen, In: Ruwa..'lfd D, Kramers PGN, eds,
Volksgezondheid Toekomsl Verkenning, De gezondheidstoestand van de Nederlandse bevolking 19502010. National Institute of Public Health and Environmental Protection. The Hague, the Netherlands: Sdu
Uitgeverij, 1993: 387-392,
26, Drykoningen CEM, Mulder ALM, Vaandrager GJ, LaPorte RE. Bruining GJ: The incidence of male
childhood Type 1 (insulin-dependent) diabetes mellitus is rising rapidly in The Netherlands, Diabetologia
1992; 35: 139-142.
27. Nederlands Huism1sen Genootschap (Dutch College of General Practitioners). NHG-Standaard diabetes
mellitus type II. Utrecht: Nederlands Huisartsen Genootschap. 1988.
28. Steering Committee on Future Health Scenarios, Chronic diseases in the year 2005. Volume I. Scenarios
on diabetes mellitus 1990-2005. Dordrccht/Boslon/London: KIuwer Academic Publishers. 1991.
29, Feskens ElM, Bowles CH. Kromhout D, Intra- and interindividuai variability of glucose tolerance in an
elderly population. J Clin Epidemiol1991; 44: 947-953.
30. Melton III. U. Palumbo PJ, Chu CP, Incidence of diabetes mellitus by clinical type, Diabetes Care
1983; 6: 75-86.
31. Rewers M, LaPorte RE. King H, Tuomilehto ], Trends in the prevalence and incidence of diabetes:
insulin·dependent diabetes mellitus in childllOOd. Wid Hlth Statist Quart 1988; 41: 179~189,
32, King H. Zimmet p, Trends in the prevalence and incidence of diabetes: non-insulindependent diabetes
mellitus. Wid Hlth Statist Quart 1988; 41: 190-196,
33, White Me, Selvin S. Menil DW. A study of multiple causes of death in California: 1955 and 1980. ]
Clill Epidcmiol 1989; 42: 355-365,
34. Krans H1\.11. Porta M, Keen H (cds,), Diabetes care and research in Europe: the St Vincent Declaration
action programme. Copenhagen: World Health Organization, Regional Office for Europe, 1992,Krans
35. Foege WH, Hogan RC, Newton LH. Surveillance projects for selected diseases, lnt J Epidemiol 1976; 5:
29-37.
36, Thacker SB, Bcrkelman RL. Public health surveillance in the United slates. Epidemiol Rev 1988; 10:
164-190.
154
General discussion
37. Smit HA, Verschuren Wl\Uo,1, Bueno de Mesquita HB, Scidcll Je. Monitoring van risicofactorcn en
gezondheid in Nederland (MORGEN-project): doelstellingen en werkwijze. Rapportnr. 263200001.
Bilthoven: RlVM, 1994.
38. Bmgman E, Meulmccster JF, Spec-van der Wckke J, Beuker RJ, Radder JJ. Peilingen in de
jeugdgezondheidszorg. PGO-Peiling 1993/1994. 1NO-PG-publicalienummer 95.061. Leiden: "INO
Preventie en Gezondheid, 1995.
39. Beks PJ, Mackaay AJC, Neeling JDN, de, Vries H de, Bouter LM, Heine HI. Peripheral arterial disease
in relation 10 glycaemic level in an elderly CaucasiM population: the Room Study. Diabetologia 1995;
38: 86·96.
40. Hofman A, Grobbee DE. long PT de. Ouweland FA van den. Detenninants of disease and disabilities in
the elderly: the Rotterdam Elderly Study. Eur J Epidemiol 1991; 7: 403-422.
41. Verschuren Wlvl1\1, Kromhout D. Total cholcstcrol and mortality at a relatively young age: do men and
women differ? TIle Netherlands Consultation Bureau Heart Project. BM] 1995; 311: 779-783.
42. Hirasing RA. Jaarverslag Nederlands Signaleringscentrum Kindergeneeskunde 1992 en 1993. "INO-PGpublicatienummer 94.052. Leiden: "INO Prevenlie en Gezondheid, 1994.
43. World Health Organization DIAMOND Project Group. WHO Multinational Project for Childhocx.l
Diabetes. Diabetes Care 1990; 13: 1062-1068.
44. Nijdam KH. Diabetes Monitoring Systeem: een verkclUling van de mogeJijkheden voor registratie van
diabeteszorg in Nederland. Amersfoort: Diabetes Vercniging Ncderland, 1996.
45. Ruwaard D. Nijdam KH. Heine RJ. Kr<U1S HMJ. Ooijen JPG, Stonns GEMG. Diabetes monitoring
system in the Netherlands. Diabetes Research and Clinical Practice 1996 (in prcss).
46. Piwemetz K, Home PD, Snorgaard O. Antsiferov M, Staehr-Johansen K. Krans HMJ. Monitoring (he
largets of the St Vindent declaration and the implementation of quality management in diabetes care: the
DiabCare initiative. Diabetic Moo 1993; 10: 371-377.
47. Piwcmctz K, Home PO, Staehr-Johansen K, Kleinebrcil L, Vemlcij D, SlOnns GEMG. Gallego IvfR,
Massi Benedetti M, Cowlcy C. Diabetes care quality network in Europe. Telematics project within the St
Vincent declaration. Diab Nutr MCL1b 1995; 8: 243-249.
48. Statutcn Nederlandse Diabetes Federatie. Amersfoort: Nederlandsc Diabetes Federatie, 1995.
49. Hoogenveen RT, Verkleij H, Jager IC. Verkennende studie bctreffcnde concurrererende en vervangcnde
doodsoorzaken in verband met kwantitatievc gezondheidsdoelen. Rapportnr. 958606004. Bilthoven:
RIVM.1994.
50. Omran AR. TIle epidemiologic transition - a thcory of the epidemiology of population change. Milbank:
Memorial Fund Quarterly 1971; 49:509-538.
51. LaPortc RE, Sauer F, Marlet E, Gamboa C, Akazawa S, Gooch T. Health and climate change. Lancet
1994: 343: 302·303.
52. Laporte RE. Patterns of disease: diabetes mellitus and Ihe rest. BMJ 1995; 310: 545-546.
53. Ruwaard D, Kramers PGN. Berg leths A van den, Achterberg PW (eds.). Public health status and
forecasts. The health status of the Dutch population over the period 1950-2010. National Institute of
Public Health and Environmental Protcction, Bilthoven. TIle Hague, the Netherlands: Sdu Uitgeverij,
1994.
54. Manton KG, Soldo BJ. Dynamics of health change in the oldest old: new perspectivcs and evidence.
Milbank Memorial Fund Quarterly 1985; 63: 206-285.
55. BrunninkllUis AEJ. Kerkhoff AHM (red). Kaarten of passen? Smmcards in de Nederlandse
gezondheidszorg. Den Haag: Sdu Uitgeverij, 1994.
155
Chapter 8
56. Engelbrecht R, Hildebrand C, lung E. Brenner G, Corcoy R. Hirsch S. Gapp C, Klcpscr G, Leiva de A,
Massi Benedelli M, Mcchller R, Piwemelz K, Sembrilzki J, 111icry J. Diabcard - a smart card for
patients with chronic discases. In: Laires MF. Ladeira }l.1], Christensen IP (eds), Health in the new
communications age: health care lelematics for Ihe 21st century. Amsterdam: IDS Press, 1995: 107-114.
57. Coobergh JWW. Observationeel wctcnschappelijk onderzoek en wctgeving. Mcdisch Contact 1994; 49;
835-836.
58. Minisleric van WVC. Infonnatiebeleid in cen ycranderend stelsel van zorg. Tweede kamer. vergaderjaar
1991-1992,22540, nrs. 1-2. 's-Gravenhage: Sdu Uitgeverij. 1992.
59, Ministerie van WVc. Tellen en meten. Een overzichl van landclijkc rcgislmties in de ZOfg. Rijswijk:
Ministerie van WVC. 1993.
60. Velden J van der. Orol RPTM, Stob: LJ. Naar cen landeJijk infonnatienetwerk huis.1rtsenzorg. Utrecht:
NIYEL. 1993.
61. Coebergh JWW. Heijden LH van der. Janssen-Heijnen MLO (cds.). Cancer incidence and survival in the
Southeast of the Netherlands 1955-1994. Einhhovcn. the Netherlands: Comprehensive Cancer Centre
South. 1995.
62. Thacker SB, Stroup DF. Future directions for comprehensive public health surveillance and health
infonnation systems in the United States. Am J Epidemiol 1994; 140: 383-397.
63. Sprenger l\1JW. Pelt W van. Infectieziekten Surveillance lnfonnatie Systeem. Rapportnr. 214670001.
Bilthovcn: RlVM, 1994.
64. Teutsch SM, Churchill RE (cds.). Principles and practice of public health surveillance. New
York/Oxford: Oxford University Press, 1994.
65. Directors of the World He..1.Ith Organization. Non-Communicable Disease CoUaborating Centers and Key
Officials. Needed: Universal Monitoring of all serious dise..'lSes of global imporlance. Am J Public Health
1993; 83:941-943.
66. WHO Directors of Non-Communicable Disease CoUaborating Centers and Key Officials. Shanghai
Declaration on non-communicable diseases. BMJ 1993; 306: 588.
67. Langeweg F (cd). Concern for tomorrow. A national environmental survey 1985-2010. Bilthoven, the
Netherlands: National Institute of Public Health and Environmental Protection, 1989.
68. National Environmental Outlook 2: 1990-2010. Follow up of Concern for tomorrow. Bilthoven, the
Netherlands: National Institute of Public Health and Environmental Protection. 1992.
69. LaPorte RE. How to improve monitoring and forecasting of disease patterns. BMJ 1993; 307: 1573-
1574.
156
SUMMARY
This thesis deals with the interrelationship between epidemiology and health policy as
regards diabetes mellitus (Chapter I). Diabetes mellitus has been selected because this
chronic metabolic disorder represents a major public health problem. Over time,
multiple chronic complications may occur, such as myocardial infarction, stroke,
circulation disorders in the legs, blindness, kidney diseases, and loss of sensitivity
and/or pain in the limbs. As a consequence, both quality of life and life expectancy are
reduced. Diabetes mellitus is responsible for a substantial degree of health care
utilisation. The objectives of the studies described in this thesis are to investigate the
occurrence of diabetes mellitus in the Netherlands, the changes that have taken place in
recent years as well as possible future developments. In addition, the implications of
these epidemiological developments for health policy have been addressed.
The points of departure are dealt with in Chapter 2, which gives a general description
of the disease and the concepts used. In a background study, which was completed in
1989, an inventory was made of the available data regarding incidence, prevalence,
remission, and mortality of diabetes in the Netherlands. The incidence alldlor preva·
lence of diabetes mellitus had been estimated in 18 surveys during the period 19711987. As we were particularly interested in the burden on health care in the Netherlands, the following basic principles were applied to choose the most appropriate
sources: the data should represent clinically-known patients and not those as yet
undiagnosed, and they should be representative for the Dutch population as a whole in
terms of age, gender, degree of urbanization and geographical variation. The incidence
among 0-19 year-olds was based on a questionnaire survey conducted among all Dutch
paediatricians and internists in the period 1978-1980. The Dutch Sentinel Practice
Network, which represents about one percent of the Dutch population, was used to
estimate the incidence from age 20 onwards (recorded in the period 1980-1983) and
the prevalence for all age groups (recorded in 1980). The annual number' of newlydiagnosed patients amounted to 17,300, which corresponded to an incidence of 0.12%
in 1980. The number of clinically-known diabetic patients appeared to be 191,000 in
1980, which corresponded to a prevalence of 1.35% of the population. Both the
incidence and prevalence increased with age.
Reliable information on remission and mortality among diabetic patients in the
Netherlands was not available. As regards remission, it was assumed that no recovery
takes place. Despite the possibility of remission, the assumption lVas made that these
'patients' will remain under a certain degree of medical supervision. Reduction of life
157
Summary
expectancy as observed in studies from other countries was chosen to assess the
outflow of diabetic patients as an alternative to mortality statistics for diabetes. From
these studies it appeared that the younger the age at which diabetes is diagnosed the
greater the reduction of life expectancy will be.
The two studies that examined the incidence around 1980 were repeated around 1990
using a similar design (two 'trend' studies). Comparing the results of the first (19781980) and second (1988-1990) nation-wide retrospective studies among children under
20 years of age, it appeared that the incidence had increased by 23% (Chapter 3). This
suggests a sustained increase of insulin-dependent diabetes mellitus (IDDM) in the
Netherlands, as it was found that the cumulative incidence previously studied in the
1960-1970 birth cohorts of male anny conscripts at 18 years of age had also risen. The
causes of the increasing incidence of IDDM, also observed in several other countries,
are unknown.
Comparing the results from the first (1980-1983) and second (1990-1992) incidence
studies in the Dutch Sentinel Practice Network, it appeared that the overall incidence
of diabetes mellitus increased by about 12% (Chapter 4). This overall increase can
largely be ato'ibuted to a statistically significant increase in the 45-64 age group (31 %).
This selective increase of non-insulin-dependent diabetes mellitus (NIDDM) is
probably not caused by a real rise due to changes in exposure to determinants. It is
more likely that the increase is due to earlier recognition of the signs and symptoms of
diabetes followed by blood glucose measurements and/or to more intensive case
finding in general practice. Although not statistically significant, the 36% increase of
diabetes mellitus in the 0-19 age group is in accordance with the increase of IDDM
based on the first and second retrospective studies covering the total Dutch population.
When age and gender-specific prevalences are not stable over time, a static model that
only takes into account demographic changes is unable to correctly forecast the
expected number of diabetic patients. We therefore developed a dynamic model in
which actual incidence, prevalence, and life expectancy data are used and in which
alternative assumptions about future trends in these parameters can be incorporated.
Two forecasting studies examined possible future developments in the occurrence of
diabetes mellitus. The first study (Chapter 5) estimated the number of patients during
the period 1980-2005, based on the information provided by the background study
described in Chapter 2. In the second study (Chapter 6), the forecasts were updated for
the period 1990-2005 by using the incidence data from around 1990 based on the tlVO
'trend' studies described in Chapters 3 and 4. Furthennore, new prevalence data were
158
Summary
used, as the prevalence study of diabetes in the Dutch Sentinel Practice Network was
repeated in 1990. It appeared that the prevalence of diabetes for those 20 years and
older had decreased significantly between 1980 and 1990. A decrease of 11 % (from
191,000 to 170,000 diabetic patients) was observed. Apparently, the audit of diabetic
care in general practice may have resulted in a clearance of the diabetic register which
in particular may be the result of changing diagnostic criteria.
Although the input parameters (starting incidence and prevalence data) were different
in each of the forecasting studies, the number of diabetic patients to be expected in
2005 were quite similar, amounting to 340,000-355,000 (2.1-2.2% of the total
population). This increase is in particular the result of an imbalance between the
incidence and life expectancy data (a higher inflow and lower outflow, respectively).
This imbalance was even greater in the second forecasting study because of a higher
incidence at baseline. Therefore the initial lower prevalence resulted in a similar
number of diabetic patients in 2005 to that found in the first forecasting study.
Assuming that the increase in incidence as observed in the 'trend' studies continues
during the period 1990-2005, the estimated number of patients in 2005 will rise even
further to 385,000. However, in view of the existence of a considerable number of
currently undiagnosed diabetic patients, the real number of known diabetic patients in
the future will probably be substantially higher (see below).
The recognition of diabetes as a major and growing public health problem raises
questions concerning the possibilities for influencing the expected increase by means of
prevention, and the possible consequences for health care. Based on the current
knowledge published in the international literature, an attempt has been made to
answer these questions and to address the health policy implications of the findings at
the national level (Chapter 7). Up to now, primary prevention of IDDM has been
confined to research without practical implications. To identify the determinants of
IDDM, an important contribution can be made at the national level by joining the
international network of childhood diabetes registries. Determinants of NIDDM have
been identified by epidemiological studies (overweight, physical inactivity and
unhealthy diet), but there are few empirically-based published studies that have
examined the effect of interventions on these determinants. Because there are many
similarities between both diseases, the experience gained from the prevention of
cardiovascular diseases can serve as an example for primary preventive strategies for
NIDDM at a national level.
159
Summary
Epidemiological studies suggest that at least 50% of all NIDDM patients may be
undiagnosed. The implications for secondary prevention have been discussed. Up to
now there is insufficient evidence for mass screening of asymptomatic individuals with
diabetes. Operational research is needed to define more clearly the different aspects of
screening and the effectiveness of screening in a community or clinical setting. It is
unlikely that universal guidelines on secondary preventive strategies will be established
because the distribution of high risk groups, the resources available, and the existing
health care stmctures differ from country to country. It is therefore strongly
recommended that a national strategy be established, based on knowledge gained
internationally.
It is not easy to foresee what the burden on health care will be in the near future as a
result of changes in the occurrence of diabetes. It is very likely that the number of
patients will increase, irrespecti~e of developments in the field of primary prevention.
However, in addition to the number of patients, changes in the health status (as a result
of possible secondary preventive and health care activities) as well as developments in
health care (such as a shift from in-patient to out-patient treatment, the policy of
reducing budgets and capacity, and increased productivity) will influence the future
burden on health care and the concomitant costs for diabetes. As the number of
diabetic patients is expected to increase while financial constraints in health care
become increasingly tighter, considerable efforts will need to be made to find more
cost-effective ways to treat diabetic patients and to reduce the complications associated
with the disease.
In the general discussion (Chapter 8) some methodological issues regarding the validity
of the sel€cted sources for assessing the occurrence of diabetes, the 'trend' data, and
the future projections, have been addressed. It is indisputable that valid data are needed
to make future projections. However, for various reasons there is a discrepancy
between the incidence and between the prevalence data when comparing different
Dutch studies, and reliable mortality data on diabetes are lacking. In addition, it should
be realised that neither study described in this thesis recorded data on a continuous
basis - they only compared two points in time. Data regarding the period in between
are lacking, which might produce a less reliable estimate of the trend over the whole
period. When interpreting the future projections, it should be borne in mind that they
do not predict the number of patients in the true meaning of prediction. The projections
explore possible future developments according to the assumptions made. In this light,
the importance of monitoring diabetes as well as the health status of the population in
160
Summary
general has been discussed in order to improve the availability and use of
epidemiological data for health policy purposes.
In conclusion, the results described in this thesis indicate that diabetes mellitus is a
serious and growing public health problem. The disease is therefore important for
health policy. For several reasons (e.g. regarding the validity and availability of data,
possible developments in preventive and health care policy), it is not easy to foresee
what the exact number of patients and the concomitant burden on health care and costs
in the near future will be. In order to make more accurate forecasts and to support
policy preparation as well as policy evaluation, the disease and its determinants need to
be monitored. However, it would be inappropriate to confine monitoring to diabetes
alone, as diabetes shares common determinants with other diseases and is itself a
determinant for other diseases (such as cardiovascular diseases). Changes in the
occurrence and course of diabetes may thus influence other related and even unrelated
diseases (competing morbidity and mortality). Therefore, an integrated approach to
setting up a coherent monitoring and information system is strongly recommended. A
significant step towards achieving a more appropriate health information system may
be accomplished by improving the links between existing health interview surveys,
epidemiological monitoring programmes, general practitioners' records, nation-wide
health care and insurance systems, and statistics on causes of death. Data from such a
monitoring and information system can provide the input for health models designed to
forecast future developments, which may be a valuable tool that can be used to
underpin health policy decisions.
161
SAMKNV ATTING
Dit proefschrift stelt de interactie aan de orde tussen epidemiologie en
gezondheidsbeleid aan de hand van diabetes mellitus (hoofdstuk I). Diabetes is
gekozen omdat deze chronische stofwisselingsziekte cen belangrijk volksgezondheids-
probleem vormt. Op den duur kunnen bij diabetespatienten meerdere chronische
complicaties optreden, zoals het hartinfarct, beroerte, circulatiestoornissen in de benen,
blindheid, nierziekten, gevoelsstoornissen en/of pijn in de ledematen. Als gevolg
hiervan nemen de kwaliteit van leven en de levensverwachting af. De ziekte vereist
dan ook de nodige aandacht van de gezondheidszorg. De doelstellingen van de in dit'
proefschrift beschreven onderzoeken zijn het bestuderen van het voorkomen van
diabetes mellitus in Nederland, de veranderingen hierin in de afgelopen jaren alsmede
de mogelijke toekomstige ontwikkeiingen. Tevens zijn de implicaties van deze
epidemiologische ontwikkelingen voor het gezondheidsbeleid beschreven.
De uitgangspunten staan vermeld in hoofdstuk 2. Dit hoofdstuk geeft een algemene
beschrijving van de ziekte en de gehanteerde concepten en definities. Een
achlel'grondsllldie, die in mei 1989 werd afgesloten, gaf inzicht in de beschikbare
infonnatie over de incidentie, prevalentie, herstel en sterfte aan diabetes in Nederland.
De illcidelltie elliof prevalelltie was in 18 onderzoeken bestudeerd in de periode 19711987. Aangezien we in het bijzonder ge[nteresseerd waren in het beslag op de
gezondheidszorg in Nederland, werden de volgende uitgangspunten gehanteerd om tot
een selectie van de meest geschikte bronnen te kamen; de cijfers over het voorkomen
van diabetes dienen betrekking te hebben op de klinisch gediagnostiseerde patienten en
niet op degenen die ongediagnostiseerd zijn, en de cijfers dienen representatief te zijn
voor de totale Nederlandse bevolking naar leeftijd, geslacht, urbanisatiegraad en
geografische spreiding. De incidentie onder 0-19 jarigen werd gebaseerd op een
vragenlijst-onderzoek onder aIle Nederlandse kinderartsen en internisten in de peri ode
1978-1980. De Continue Morbiditeits Registratie Peilstations Nederland, die circa een
procent van de Nederlandse bevolking vertegenwoordigt, werd geselecteerd om de
incidentie vanaf 20 jaar (gemeten in de periode 1980-1983) en de prevalentie voor aile
leeftijden (gemeten in 1980) te schatten. Het jaarlijks aantal gediagnostiseerde
diabetespatienten bedroeg 17.300, hetgeen overeenkomt met een incidentie van 0,12%
in 1980. Het aantal klinisch bekende patienten bedroeg 191.000 in 1980, hetgeen
overeenkomt met een prevalentie van 1,35% van de bevolking. Zowel de incidentie als
prevalentie stijgen met de leeftijd.
163
Samenvatting
Betrouwbare informatie over herstel en sle/fte bij patienten met diabetes zijn niet
beschikbaar in Nederland. Ondanks de mogelijkheid van hers tel is aangenomen dat dit
niet optreedt, omdat deze 'patienten' over het aigemeen toch onder medische contrale
blijven. De reductic in levensverwachting, zoals gevonden in buitenlandse onderzoeken,
is gekozen om de uitstroOTll van diabetespatienten te bepaJen als alternatief voar
gegevens afkamstig van de sterftestatistiek. Vit deze onderzoeken bleek dat hoe jonger
de diagnose diabetes wordt gesteld, des te groter de reductie in levensverwachting.
De twee onderzoeken die de incidentie bepaalden rond 1980 zijn rond 1990 herhaald
met eenzelfde opzet (Iwee 'Irend' onderzoeken). Bij vergelijking van de resultaten
van het eerste (1978-1980) en tweede (1988-1990) landelijke relrospectieve onderzoek
onder kinderen tot 20 jaar, bIijkt dat de incidentie met 23% is gestegen (hoofdstuk 3).
Dit suggereert een blijvende toename in insuline-afhankeIijke diabetes mellitus (IADM)
in Nederland, aangezien de cumulatieve incidentie in de 1960-1970 geboorten-cohorten
van mannelijke IS-jarige dienstplichtigen ook bleek te stijgen. De oorzaken van de
stijgende incidentie van IADM, die oak in verschillende andere landen is gevondell,
zijn niet bekend.
Vit vergeIijking van de resultaten van het eerste (1980-1983) en tweede (1990-1992)
incidentie-onderzoek van de Peilstations Nederland blijkt dat de incidentie van diabetes
met 12% is gestegen (hoofdstuk 4). Deze totale toename is in belangrijke mate het
gevolg van een significante taename in de leeftijdsgroep 45-64 jaar (31%). Deze
selectieve toe name van niet-insuline-afhankelijke diabetes mellitus (N1ADM) is
waarschijnlijk niet veroorzaakt door een verhoogde blootstelling aan determinanten
(risicofactoren). Het is waarschijnlijker dat deze taename het gevolg is van een eerdere
herkenning van klachten en kenmerken passend bij de ziekte gevolgd door
bloedglucose bepalingen en/of intensievere case finding in de huisartspraktijk.
Alhoewel niet statistisch significant, de 36% toename van diabetes in de leeftijdsgroep
0-19 jaar is in overeenstemming met de toename van IADM gebaseerd op de
bevindingen uit het eerste en tweede landelijke retrospectieve onderzoek.
Ais de leeftijd- en geslachtspecifieke prevalenties niet constant zijn in de tijd, is een
statisch model, dat aileen veranderingen in demagrafie in rekening brengt, niet geschikt
om het toekomstig aantal patienten met diabetes mellitus te berekenen. Daarom
ontwikkelden we een dynamisch model waarin de feitelijke gegevens over de
incidentie. prevalentie en levensverwachting z!jn gebruikt en alternatieve aannames
over toekomstige trends in deze parameters kunnen worden verwerkt. Twee scenario-
sludies onderzochten mogelijke taekomstige ontwikkelingen in het voorkomen van
164
Samenvatting
diabetes mellitus. De eerste studie (hoofdstuk 5) berekende het aantal patienten voor de
periode 1980-2005, gebaseerd op de infonnatie uit de achtergrondstudie beschreven in
hoofdstuk 2. In de tweede studie (hoofdstuk 6) zijn deze berekeningen bijgesteld voor
de periode 1990-2005 met gebruik van de incidenties rond 1990, zoa1s gevonden in de
twee 'trend' onderzoeken die zijn beschreven in hoofdstuk 3 en 4. Tevens zijn nieuwe
gegevens over de prevalentie gebl'llikt, aangezien in de Peilstations Nederland de
prevalentie opnieuw is onderzocht in 1990. Hieruit bleek dat de prevalentie van
diabetes vanaf 20 jaar significant is gedaald tussen 1980-1990. We constateerden een
daling van 11% (van 191.000 naar 170.000 diabetespatienten). Veranderde diagnosecriteria zijn mogelijk in belangrijke mate verantwoordelijk voor een opschoning van het
diabetesbestand in de huisartspraktijk.
Alhoewel de invoergegevens (start incidentie en prevalentie) verschillend zijn in elk
van de scenariostudies, is het aantal te verwachten patienten met diabetes in 2005
vergelijkbaar en wordt geschat op 340.000-355.000 (2.1 %-2.2% van de totale
bevolking). Deze toename is in het bijzonder het gevolg van het niet in even wicht zijn
van de incidentie en de levensverwachting (respectievelijk een hogere instroom en een
lagere uitstroom). Dit speelt een nog grotere rol in de tweede studie vanwege een
hogere ineidentie bij aanvang. Daarom resulteert de initiele lagere prevalentie in de
tweede studie in een vergelijkbaar aantal diabetespatienten in 2005 als in de eerste
studie. Ais de stijging in incidentie zoals gevonden in beide 'trend' onderzoeken zich
voortzet in de periode 1990-2005, dan zal het aantal te verwachten patienten in 2005
zelfs stijgen tot 385.000. Indien echter rekening gehouden wordt met het gegeven dat
een aanzienlijk aantal diabetespatienten momenteel niet als zodanig gediagnostiseerd is,
zal het werkelijke aantal bekende patienten in de toekomst waarschijnlijk aanzienlijk
hoger zijn (zie hierna).
De her kenning van diabetes mellitus ais een belangrijk en toenemend volksgezondheidsprobleem roept vragen op wat de mogelijkheden zijn om de te verwachten
toename te beInvioeden door middel van preventie en wat de mogelijke consequenties
zijn voor de gezondheidszorg. Gebaseerd op de huidige kennis zoals gepubliceerd in de
internationale literatuur, is getracht deze vragen te beantwoorden en de implicaties
het gezondheidsbeleid op nationaal niveau te adresseren (hoofdstuk 7). Tot nu
toe blijkt primaire preventie van IADM zich te beperken tot onderzoek, vooralsnog
zonder toepassing in de dagelijkse praktijk. Om de detenninanten van IADM op te
sporen kan op nationaal niveau een belangrijke bijdrage geleverd worden door zich aan
VOOI'
te sluiten bij het internationale netwerk van diabetesregistraties voor kinderen.
Determinanten
van
NIADM
zijn
opgespoord
in
epidemiologisch
onderzoek
165
Samenvatting
(overgewicht, lichamelijke inactiviteit en ongezonde voeding), maar er zijn slechts
weinig onderzoeken gepuhliceerd die het effect van interventies op deze determinanten
beschrijven. Aangezien er vele raakvlakken zijn tussell beide ziekten, kan de ervaring
die is verkregen met de preventie van hart- en vaatziekten als een voorbeeld dienen
voor primaire preventie-strategieen voor NlADM op nationaal niveau.
Epidemiologische onderzoeken wijzen uit dat tenminste 50% van aile pam~nten met
NlADM als zodanig niet gediagnostiseerd is. De implicaties voor secundaire preventie
zijn besproken. Tot nu toe is er onvoldoende bewijs dat massale screening in de
bevolking van asymptomatische personen met diabetes rechtvaardigt. Toegepast
onderzoek is noodzakelijk om meer inzicht te krijgen in de verschillende aspecten van
screening en de effectiviteit van screening in bevolkingsonderzoek of klinische praktijk.
Het is onwaarschijnlijk dat universele richtlijnen voor secundaire preventie-strategieen
zullen worden vastgesteld, omdat de verdeling naar risicogroepen, de beschikbare
middelen en de bestaande zorgstructuren van land tot land verschillen. Een nationale
strategie op basis van internationaal verworven kennis is daarom sterk aan te bevelen.
Het is niet eenvoudig te voorzien wat in de nabije toekomst het beslag op de
gezondheidszorg zal zijn ais een gevoig van de veranderingen in het voorkomen van
diabetes mellitus. Het is zeer waarschijnlijk dat het aantal patienten zal toenemen,
onafl13nkelijk van ontwikkelingen op het terrein van primaire preventie. Tevens zullen
veranderingen in de gezondheidstoestand (als gevoig van mogelijk activiteiten op het
terrein van secundaire preventie en therapeutische mogelijkheden) alsmede
ontwikkelingen in de gezondheidszorg (zoals een verschuiving van klinische naar
poliklinische zorg, bezuinigingen en verhoogde produktiviteit) de toekomstige last op
de gezondheidszorg en de daaruit voortkomende kosten voor diabetes bei'nvloeden. De
te verwachten toename van het aantal patienten met diabetes onder omstandigheden
van verminderde beschikbaarheid van financiele middelen in de gezondheidszorg,
impliceert dat belangrijke inspanningen geleverd moeten worden om de
diabetespatienten effectiever te behandelen en complicaties te voorkomen.
In de algemene discussie (hoofdstuk 8) is aandacht besteed aan enkele methodologische
aspecten van de onderzoeken, zoals de validiteit van de geselecteerde bronnen voor het
bepalen van het voorkomen van diabetes, de 'trend' gegevens en de toekomstprojecties.
Het staat buiten kijf d.t betrouwbare gegevens nodig zijn om toekomstprojecties uit te
voeren. Echter, vanwege uiteenlopende redenen zijn er discrepanties waar te nemen
tussen incidentie- en prevalentiegegevens van verschillende Nederlandse onderzoeken.
Ook ontbreken betrouwbare sterftecijfers voor diabetes in Nederland. Daarnaast dient
166
Samenvatting
men zich te realiseren dat de beschreven 'trend' onderzoeken slechts twee punten in de
tijd met elkaar vergelijken. Gegevens over de tussenliggende periode ontbreken.
Hierdoor wordt een minder betrouwbare trendschatting over de hele peri ode verkregen.
Bij het interpreteren van de toekomstprojecties dient men zich crYan bewust te zijn dat
zij niet het toekomstig aantal paW~nten voorspellen maar slechts cen indica tie geven.
De projecties exploreren mogelijke toekomstige ontwikkelingen gegeven de aannames.
In deze context is het belang van monitoring van diabetes en van de gezondheidstoestand in het algemeen bediscussieerd om de beschikbaarheid en het gebruik van
epidemiologische gegevens voor het gezondheidsbeleid te verbeteren.
Concluderend kan gesteld worden dat de in dit proefschrift beschreven resuItaten
aangegeven dat diabetes mellitus een belangrijk en in omvang toenemend volksgezondheidsprobleem is. De ziekte is dan ook van belang voor het gezondheidsbeleid. Om
uiteenlopende redenen (zoals de validiteit en beschikbaarheid van gegevens, mogelijke
ontwikkelingen in het preventie- en gezondheidszorgbeleid) is het niet eenvoudig om te
voorzien wat het exacte aantal patienten en het beslag op de gezondheidszorg en de
kosten in de nabije toekomst zullen zijn. Om meer onderbouwde toekomstprojecties te
maken en de voorbereiding en evaluatie van gezondheidsbeleid te ondersteunen, dienen
de ziekte en de determinanten ervan 'gemonitored' te worden. Het is echter
onvoldoende om monitoring te beperken tot diabetes aileen, omdat diabetes
gemeenschappelijke detenninanten heeft met andere ziekten en zelf ook een
determinant is voor andere ziekten (zoals hart- en vaatziekten), Veranderingen in het
voorkomen en het beloop van diabetes kunnen dus van invloed zijn op andere
gerelateerde en zelfs niet gerelateerde ziekten (vervangende en concurrerende
morbiditeit en mortaliteit). Het verdient dan ook sterke aanbeveling am een coherent
monitoring- en informatiesysteem op te zetten. Een belangrijke stap hiertoe kan gezet
worden door koppeling te bewerkstelligen tussen bestaande gezondheidsenquetes,
epidemiologische monitoring programma's, huisartsen-registratiesystemen, landelijke
zorg- en verzekeringssystemen en doodsoorzaken-statistieken. Gegevens van zo'n
monitoring- en informatiesysteem kunnen gebruikt worden vaar gezondheidsmodellen
die ontworpen zijn om toekomstige ontwikkelingen te verkennen, en daarmee
belangrijk gereedschap leveren am beslissingen in het gezondheidsbeleid te
ondersteunen.
167
DANKWOORD
Tijdens het Scenarioproject Chronische Ziekten, waar diabetes mellitus deel van
uitmaakte, werd mijn speciale belangstelling gewekt voor deze ziekte. De gelegenheid
die mij geboden werd om op het terrein van de epidemiologie van diabetes onderzoek
uit te voeren heeft uiteindelijk geresuIteerd in dit proefschrift. Aan de totstandkoming
crYan hebben vele menscn bijgedragel1, waarvan ik hief cen aantal in het bijzonder wil
noemen.
Allereerst wil ik mijn (co)promotoren, prof.dr.ir. D. Kromhout, prof.dr. A.F. Casparie
en dr. H. Verkleij bedanken. Beste Daan, je niet aflatende stimulans was noodzakelijk
om naast mijn andere drukke werkzaamheden tot cen afronding van mijn proefschrift te
kamen. Dat met twee woorden volstaan kon worden am duidelijk te maken wat we van
elkaar wilden en de snelheid en accuraatheid waarmee je concept-artikelen en
hoofdstukken van dit proefschrift van commentaar hebt voorzien, bespoedigden die
afronding. BijlOnder bedankt. Beste Ton, voor het eerst leerde ik je kennen als
voorzitter van de Scenariocommissie Chronische Ziekten. Jouw speciale belangstelling
voor diabetes, voorheen ais clinicus en daarna op meer beleidsmatig terrein, heb je op
mij overgebracht. Veel dank voor je positieve inbreng bij het totstandkomen van dit
proefschrift. Beste Harry, vanaf het eerste moment dat ik in Leiden solliciteerde tot nu
toe hebben we elkaar niet uit het oog verloren. De \Vijze \Vaarop je gebruik maakte van
een white-board om op schematische wijze Qllze gedachten te ardellen zal mij steeds
bijblijven. Ik ben je zeer erkentelijk voor je inbreng vanaf de opzet tot en met de
rapportage van dit proefschrift. Gedrieen vormen jullie qua achtergrond een 'mixed
bag'. Jullie stelden mij in de gelegenheid om van ieders expertise maximaal gebruik te
maken.
Ook al zeg ik allereerst dank aan de (co)promotoren, minstens net lOveel dank komt
toe aan drs. R. Gijsen en ir. R.T. Hoogenveen. Beste Ronald en Rudolf, jullie
dagelijkse enthousiaste inzet bij de analyses van gegevens en het opstellen en
implementeren van modellen was onontbeerlijk. De vele vragen tussendoor waren nooit
een probleem. Ik stel het dan ook zeer op prijs dat jullie bereid zijn om mij als
paranimfen bij te staan.
De samenwerking met andere instituten vormde een wezenlijk onderdee!. A.!.M.
Bartelds, huisarts en projectmanager van de Continne Morbiditeits Registratie van het
NIYEL te Utrecht, stelde mij in de gelegenheid om onderzoek op te zetten en uit te
voeren onder de Peilstationsartsen. Beste Aad, ik ben jou, het secretariaat en de
169
Dankwoord
PeiIstationsartsen veel dank verschuldigd. Dr. R.A. Hirasing, kinderarts en projectIeider
van het incidentie-onderzoek onder 0-19 jarigen bij TNO Preventie en Gezondheid te
Leiden, stelde mij in de gelegenheid om mee te werken aan dat onderzoek. Beste
Remy, je weet hoe zeer ik de samenwerking met jOll waardeer.
Dr.ir. E.LM. Feskens, beste Edith, je bereidheid om tllssendoor van gedachten te
wisselen over de epidemiologie van diabetes heb ik zeer op prijs gesteld. Bedankt voor
het grondig doornemen van een van de laatste versies van het proefschrift. M. Scholsz,
beste Miranda, veel dank ben ik je verschuldigd voor de logistieke werkzaamheden en
de redactionele ondersteuning. M. Gould, beste Michael, veel dank voor de snelle,
adequate en leerzame wijze waarop je Engelse teksten corrigeerde. M.D. Cornelissen-
Kuyt, beste Marianne, je in familiekring bekende creativiteit heb je opniellw laten
blijken in de omslag. Hartelijk dank hiervoor.
Ook al was er minder directe betrokkenheid van de overige collega's op het Centnlln
voor Volksgezondheid Toekomst Verkenningen van het RIVM, de wanne
belangstelling heb ik zeer gewaardeerd. Dr. P.G.N. Kramers, beste Pieter, bij deze wiI
ik jall en via jOll aIle andere collega's hiervoor bedanken. Jullie inzet om mij in de
laatste fase van een aantal klllssen te ontzien is voor mij van ongekende waarde. De
Directie van het RIVM dank ik voor de mimte die mij is geboden om dit proefschrift
af te ronden.
Last but not least, het afronden van een proefschrift naast een drukke baan steIt hoge
eisen aan het thuisfront. Lieve Helina, zander jouw positieve instelling om mij die
gelegenheid te bieden was het nooit gelllkt. Het ontnemen van huishoudelijke taken en
de morele steun zijn onbetaalbaar. Ollze klnderen maakten mij er voortdurend van
bewust dat er meer is op aarde dan aIleen promoveren. De balans tussen werk en prive
dient dan ook hersteld te worden.
170
ABOUT THE AUTHOR
Dirk Ruwaard was born on August 17th, 1960 in Katwijk aan Zee, the Netherlands.
After completing secondary school (Atheneum at the Pieter Groen College in Katwijk
aan Zee) in 1978, he studied Mathematics and Physics in Delft for one year. He started
to study Medicine in September 1979 at Leiden University Medical School, the Netherlands. He obtained his medical degree cum laude in September 1986. Subsequently, he
worked at the Central Laboratory of the Netherlands Red Cross Blood Transfusion
Service, in Amsterdam, and as an army conscript. In February 1988 he started research
at the Institute of Social Medicine, at the Medical Faculty of the University of Leiden
(Head: Prof D. Kromhout), on the project 'The future burden of chronic diseases for
Dutch society: scenarios for diabetes mel1itus~ chronic non-specific lung diseases and
rheumatoid arthritis 1990-2005'. Since 1989 this project has been carried out at the
National Institute of Public Health and the Environment CRIVM) in Bilthoven, the
Netherlands. This project was the starting-point for the further research activities
described in this thesis. In 1988 and 1989 he attended the Summer Course in Epidemiology at the Leiden/Erasmus Rotterdam University Medical Schools and the Summer
Program in Epidemiology (focusing on epidemiology, health risk assessment and health
services reseach) at the Department of Epidemiology of the Johns Hopkins University
School of Hygiene and Public Health in Baltimore, USA, respectively. From 19911993 he specialized in Social Medicine at the TNO Institute for Prevention and Health,
Leiden. In 1992 he became head of the Public Health Analysis Branch within the
Department for Public Health Forecasting, and project leader of the project 'Public
Health Status and Forecasts' at RIVM. In January 1996 he was appointed deputy head
of the Department for Public Health Forecasting. Furthermore, he is a member of a
number of committees and societies in the [wid of public health research.
171
View publication stats