American Journal of Gastroenterology ISSN 0002-9270 doi: 10.1111/j.1572-0241.2004.30390.x  C 2004 by Am. Coll. of Gastroenterology Published by Blackwell Publishing Risk Factors for Erosive Esophagitis: A Multivariate Analysis Based on the ProGERD Study Initiative Joachim Labenz, M.D., Daniel Jaspersen, Prof., Michael Kulig, M.D., Andreas Leodolter, M.D., Tore Lind, Prof., Wolfgang Meyer-Sabellek, Prof., Manfred Stolte, Prof., Micheal Vieth, M.D., Stefan Willich, Prof., and Peter Malfertheiner, Prof. Department of Medicine, Jung-Stilling Hospital, Siegen, Germany; Department of Gastroenterology, Klinikum Fulda, Germany; Institute of Social Medicine, Epidemiology and Health Economics, Charité Hospital, Umboldt University of Berlin, Germany; Department of Gastroenterology, Otto-von-Guericke University Magdeburg, Germany; AstraZeneca, Mölndal, Sweden; AstraZeneca, Wedel, Germany; Institute of Pathology, Klinikum Bayreuth, Germany; and Institute of Pathology, Otto-von-Guericke University Magdeburg, Germany OBJECTIVES: Gastroesophageal reflux disease can be divided into three categories: nonerosive GERD (NERD), erosive GERD (ERD), and Barrett´s esophagus. A shift among these categories rarely occurs. The aim of the present study was to elucidate potential patient-associated risk factors associated with ERD. METHODS: A total of 6,215 patients with troublesome heartburn were recruited to a large, prospective, multicenter open cohort study comprising an initial treatment phase and a 5-yr follow-up phase. Each center planned to recruit an equal number of patients with NERD and ERD. All patients underwent an interview based on standardized questionnaires, a physical examination, and endoscopy with biopsies. Data were analyzed by multiple logistic regression analysis. RESULTS: Risk factor analysis was performed on 5,289 patients (NERD: n = 2,834; ERD: n = 2,455), which was the intent-to-treat population excluding patients with suspected/proven complicated reflux disease. Stepwise regression analysis identified the following independent predictors of ERD: male gender, overweight, regular use of alcohol, a history of GERD >1 yr, and smoker or ex-smoker. A higher level of education and a positive Helicobacter pylori (H. pylori) status were associated with a lower risk of ERD. CONCLUSIONS: Some patient-associated factors increase the risk of erosive esophagitis as opposed to nonerosive reflux disease. However, no single factor or combination of factors is capable of predicting mucosal damage with clinically sufficient certainty. Thus, endoscopy is still required in all GERD patients if valid information on the state of the esophageal mucosa is needed. (Am J Gastroenterol 2004;99:1652–1656) INTRODUCTION Gastroesophageal reflux disease (GERD) is a common condition that affects around 20–50% of adults in Western countries (1–4). The preeminent factor in the pathophysiology of the condition is a disordering of the antireflux barrier comprising the lower esophageal sphincter and the ural diaphragm (5). In contrast, other mechanisms such as esophageal clearance, epithelial resistance, gastric acid production, and gastric emptying are of secondary importance. The disease can be divided into three clinical categories: nonerosive reflux disease (NERD), erosive reflux disease (ERD), and Barrett’s esophagus (6). Only rarely is a shift among these categories observed. That is, a worsening of the illness over time, in particular to premalignant changes, is the exception and not the rule (6–8). The question thus arises as to why some patients present only with symptoms, but no lesions, that is, have NERD; while others develop mucosal breaks. In terms of the nature and severity of symptoms and impairment of quality of life, no differences are to be seen between the categories of NERD and ERD (1, 9, 10). Although the development of GERD is explained by a dysfunction of the antireflux barrier, no close correlation can be found between the severity of such a disturbance and the severity of the clinical or endoscopic presentation. The treatment and course of GERD are determined by the clinical category, and a noninvasive identification of patients with erosive esophagitis would therefore be desirable. The aim of the analysis presented here was to evaluate the patient-associated risk factors for the development of erosive esophagitis. For this purpose, the data of the ProGERD study, a large international, multicenter cohort study on more than 6,000 patients were subjected to uni- and multivariate analyses. 1652 Risk Factors for Erosive Esophagitis MATERIALS AND METHODS The ProGERD study is a large, prospective, multicenter open cohort study comprising an initial treatment phase and a 5-yr follow-up. Patients (n = 6,509) were recruited from centers (n = 1,253) in Germany, Austria, and Switzerland, the majority of which were primary care clinics (approximately 90%). Patients with no source data verification, no treatment, no informed consent, incomplete data, or age below 18 yr were excluded, leaving 6,215 patients in the intentto-treat population. Each center planned to recruit an equal number of patients with erosive and nonerosive GERD. For this reason each study site received study material for an equal number of patients with erosive and nonerosive GERD. All patients with predominant heartburn aged 18 yr or older who attended the study site were enrolled into the study. Patients with a history of gastrointestinal surgery, gastrointestinal malignancies, continuous treatment with an antacid for more than 7 days during the preceding 4 wk, any alarming symptoms such as undesired weight loss, hematemesis, melena, or other signs indicating serious or malignant disease were excluded, as were pregnant or nursing women, patients abusing alcohol or drugs, and those with limited language skills. At study entry patients underwent a physical examination and Helicobacter pylori (H. pylori) assessment, and were asked to complete standardized questionnaires assessing demographic, medical, and social characteristics. Upper gastrointestinal endoscopies with biopsy were performed in all patients presenting with heartburn (frequency and severity assessed by the Upper Abdominal Symptom Questionnaire) who were subsequently divided into NERD and ERD groups. ERD patients were classified in accordance with the Los Angeles (LA) classification system (11). A risk factor analysis was performed on 5,289 patients (NERD: n = 2,834; ERD: n = 2,455), which was the intentto-treat population (n = 6,215) excluding 926 patients with endoscopically suspected and/or histologically proven complicated reflux disease (ulcer, stricture, Barrett´s esophagus). All available data were analyzed by univariate statistics using the χ 2 test, t-test, analysis of variance (ANOVA), Mann-Whitney U-test, or Kruskal–Wallis test as appropriate. Significant prognostic factors were then subjected to a multivariate analysis with logistic regression. The model was not tested for interactions. Statistical significance was defined by a two-sided alpha level of 0.05. RESULTS The study population comprised 5,289 patients, of whom 2,455 had erosive esophagitis (ERD). Table 1 shows the univariate comparisons of patients with ERD and NERD. ERD patients were predominantly male while NERD patients were more frequently female (OR 1.774, 95% CI 1.283–1.445). There was only a minor difference of 0.8 yr in the mean age. Patients with ERD were more frequently smokers or ex-smokers (OR: 1.395, 95% CI: 1.248–1.559) and regular 1653 Table 1. Patient Characteristics ERD NERD (n = 2,455) (n = 2,834) Male gender (%) 58.8 44.6 Mean age (SD), yr 53.6 (13.7) 52.8 (14.3) Caucasians (%) 99.0 98.3 70.9 61.9 BMI >25 kg/m2 (%) Smoking∗ (%) 58.7 50.5 Regular alcohol intake† (%) 39.7 28.5 Intake of aspirin (%) 12.7 12.2 Intake of NSAIDs (%) 8.7 7.6 Intake of calcium 10.0 9.3 channel blockers, betamimetics, theophylline, nitrates (%) Peptic ulcer at baseline (%) 1.9 1.3 H. pylori +ve at baseline (%) 25.1 28.8 Duration of disease >1 yr (%) 69.8 63.6 >5 yr (%) 30.4 26.0 p-Value <0.0001 0.0337 0.0319 <0.0001 <0.0001 <0.0001 0.5792 0.1331 0.3627 0.0765 0.0031 <0.0001 0.0003 ∗ Smokers and ex-smokers. †5 drinks/wk. alcohol consumers (OR: 1.647, 95% CI: 1.463–1.854). Moreover, ERD patients had a longer disease history (OR: 1.247, 95% CI: 1.106–1.406) and were less frequently infected with H. pylori (OR: 0.831, 95% CI: 0.734–0.940). Conversely, peptic ulcers were found more frequently in NERD patients although this difference was not significant. The proportion of patients taking aspirin, NSAIDs, or drugs believed to lower LES pressure was similar in both groups. No differences were found between the groups with respect to concomitant diseases such as hypertension, coronary heart disease, other cardiovascular disorders, pulmonary disorders, diabetes, joint and spine disorders including inflammatory diseases. Agestratified reanalysis (3 categories: ≤ 40 yr; 41–60 yr; >60 yr) showed similar results in all age groups with one exception: H. pylori infection was less frequently detected in patients with ERD and 40 yr of age or less (ERD: 23.8%; NERD: 28.5%; p = 0.0813), and was a significant protective factor in patients aged 41–60 yr (ERD: 25.2%; NERD: 31.5%; p = 0.0008), while the proportion of patients above 60 yr with H. pylori infection was similar in both groups (ERD: 28.4%; NERD: 28.5%; p = 0.992). Interestingly, there was no clear increase in the prevalence of H. pylori infection with increasing age. Logistic regression analysis identified the following factors as independent predictors of ERD: male gender, overweight (body mass index >25 kg/m2 ), regular intake of alcohol, a history of GERD of >1 yr, and smoker or ex-smoker (Table 2). However, a higher level of education and a positive H. pylori status were associated with a lower risk of ERD. Although the probability of having erosive esophagitis increased with the number of risk factors present, as many as 36.1% of the male smokers or ex-smokers older than 40, who regularly drank alcohol and had a BMI of > 30 kg/m2 had non-erosive reflux disease. 1654 Labenz et al. Table 2. Variables Associated with Any Grade of Erosive Esophagitis: Results of the Multivariate Analysis (complete model) Risk Factors Reference Male Age >40–60 yr Age >60 yr Caucasian BMI >25–30 kg/m2 BMI >30–40 kg/m2 BMI >40 kg/m2 Any alcohol intake Regular alcohol intake Duration of disease >1–5 yr Duration of disease >5 yr Smoker/ex-smoker Single Employed Retired Manual worker Nongraduate Secondary school Highschool education Intake of analgesics except than aspirin and NSAIDs Intake of aspirin Intake of NSAID Intake of other concomitant medication∗ Family history of GERD H. pylori positive Duodenal or gastric ulcer Arterial hypertension Coronary heart disease Circulatory disorder Pulmonary disorder Diabetes Joint or spine disorder Inflammatory joint disease ∗ Female Other age groups ≤60 yr Non-caucasian Other BMI Other BMI ≤40 kg/m2 No intake No/occasional intake Other duration ≤5 yr Nonsmoker Married Unemployed or retired Employed/unemployed Nonmanual worker Graduate Primary school Nonhighschool education No intake No intake No intake No intake No history H. pylori negative No ulcer No hypertension No CHD No circulatory disorder No pulmonary disorder No diabetes No joint or spine disorder No inflammatory joint disease OR (95% CI) p-Value 1.590 (1.381–1.830) 1.103 (0.934–1.304) 1.153 (0.902–1.474) 1.639 (0.938–2.865) 1.337 (1.161–1.538) 1.470 (1.229–1.759) 1.570 (0.871–2.831) 1.100 (0.945–1.281) 1.290 (1.105–1.506) 1.201 (1.041–1.387) 1.281 (1.093–1.500) 1.157 (0.999–1.339) 1.154 (0.983–1.355) 0.956 (0.794–1.152) 0.974 (0.786–1.207) 1.028 (0.889–1.188) 1.079 (0.718–1.624) 0.993 (0.864–1.142) 0.798 (0.651–0.977) 0.903 (0.693–1.179) 1.046 (0.866–1.265) 1.212 (0.961–1.528) 1.138 (0.894–1.447) 0.994 (0.866–1.142) 0.854 (0.743–0.980) 1.518 (0.917–2.514) 0.982 (0.836–1.155) 0.816 (0.643–1.036) 1.139 (0.782–1.659) 0.956 (0.749–1.220) 1.080 (0.805–1.449) 0.917 (0.699–1.203) 0.910 (0.381–2.175) <0.0001 0.2479 0.2562 0.0828 <0.0001 <0.0001 0.1337 0.2174 0.0012 0.0123 0.0022 0.0517 0.0804 0.6378 0.8101 0.7131 0.7140 0.9244 0.0290 0.4542 0.6392 0.1036 0.2935 0.9350 0.0247 0.1050 0.8302 0.0947 0.4980 0.7171 0.6071 0.5321 0.88324 Other concomitant medication: calcium channel blockers, betamimetics, theophylline, nitrates. A separate analysis of variables associated with severe erosive esophagitis (grades C and D according to the Los Angeles classification) is shown in Table 3. Logistic regression analysis confirmed male gender, overweight, regular intake of alcohol, smoking habits, and a longer disease history as independent predictors of severe erosive disease, with odds ratios higher than those identified in the overall group Table 3. Significant Variables Associated with Severe Erosive Reflux Disease (Los Angeles C/D): Results of the Multivariate Analysis (Model Includes All Variables Shown in Table 2) Risk Factors Male gender Age >60 yr BMI >25–30 kg/m2 BMI >30–40 kg/m2 Regular alcohol intake Duration of disease >1–5 yr Duration of disease >5 yr Smoker/ex-smoker Single Retired H. pylori positive OR (95% CI) 2.177 (1.576–3.008) 1.869 (1.115–3.314) 1.703 (1.237–2.344) 1.971 (1.327–2.926) 1.706 (1.232–2.362) 1.475 (1.067–2.040) 1.607 (1.137–2.272) 1.333 (1.004–1.771) 1.638 (1.184–2.265) 0.615 (0.390–0.967) 0.610 (0.440–0.845) of patients with erosive esophagitis. In contrast to the overall group, aging was associated with a higher risk of severe esophagitis. H. pylori infection was associated with a lower risk for severe erosive disease, with the odds ratio being lower as compared to the overall group of patients with erosive disease. Living alone increased the risk of severe esophagitis, while retirement lowered the risk. Patients’ factors not related to ERD were: age; family history of GERD; ethnicity; marital status; use of acetylsalicylic acid, nonsteroidal antiinflammatory drugs (NSAIDs), other analgesics, calcium channel blockers, betamimetics, theophylline, or nitrates; concomitant diseases. p-Value <0.0001 0.0177 0.0011 0.0008 0.0013 0.0187 0.0072 0.0469 0.0029 0.0354 0.0029 DISCUSSION This study investigating large groups of patients with nonerosive and erosive reflux disease clearly showed that no single patient-associated factor is capable of predicting erosive esophagitis, although overweight males regularly using alcohol carry a higher risk, especially if they have a longer disease history. A number of previous studies have shown that symptom severity is not correlated with the severity Risk Factors for Erosive Esophagitis of esophageal damage, and this has been confirmed in the present study (1, 9, 10). However, a large study involving 10,294 patients with erosive esophagitis identified duration and severity of heartburn along with male gender and obesity as independent risk factors of severe erosive esophagitis (Los Angeles grades C and D) (12). Our prospective study is one of the largest ever done on GERD patients. However, the fact that it is not populationbased is a shortcoming, and the possibility of selection bias has thus to be considered. In our opinion, a selection bias is an unlikely explanation of the reported findings, since patients with NERD and ERD were identified by the same physicians on the basis of symptom severity following a unique selection process. However, there is a possibility of misclassification bias, since prior antisecretory therapy may have led to healing of erosions or a lower grade of erosive esophagitis. Therefore, we reanalyzed the data comparing patients with nonerosive disease and those with severe erosive esophagitis (grades C and D). This revealed an even stronger association of the variables identified in the overall group, clearly suggesting that they are important. Male gender has been frequently cited as a risk factor for GERD. However, a population-based study in the United States failed to show any differences between males and females with respect to the prevalence of reflux symptoms (4). This does not exclude the possibility that males are more likely to develop erosive esophagitis and females are more likely to present with nonerosive disease, as has been the case in our study and a recent international trial on patients with GERD in primary care (13). Moreover, male gender has been identified as a risk factor for more severe erosive reflux disease (grades C and D) (12), and this has now been confirmed by our analysis. Overweight and alcohol intake are controversial risk factors for the development of GERD (14–17), and the effect of weight reduction and cessation of alcohol consumption on the clinical course of GERD is marginal at best (18–20). However, in this study these factors that might increase acid reflux into the esophagus by various mechanisms were associated with a higher risk of erosive esophagitis, suggesting a higher acid load of the esophagus in patients with these risk factors. It has been hypothesized that GERD can be divided into three categories, two of which are nonerosive and erosive reflux disease (6). On the basis of epidemiological data it has also been hypothesized that a shift from one category to another is a very rare event (6, 8). If this is true, then we would expect a similar duration of the disease history in patients with NERD and ERD. In our study, a longer duration of the disease was, however, associated with a higher risk of erosive esophagitis. This difference could not be explained by the small but significant higher age of patients with erosive esophagitis. Our findings are in contradiction to previous reports (21). However, the number of patients in the previous study might be too small to reveal differences of the magnitude found in our trial. On the other hand, the study by El-Serag and Johanson on patients with ero- 1655 sive reflux esophagitis identified the severity and duration of heartburn as independent predictors of more severe erosive esophagitis (12). On the basis of our findings we hypothesize that some patients with NERD will progress to more severe disease. The long-term follow-up of our study population will show whether this actually is true, and whether this occurs so frequently that clinical management strategies have to be adopted. In the present study H. pylori infection was associated with a somewhat lower risk for the development of mucosal breaks in the esophagus in patients with GERD, thus corroborating previous studies in smaller patient groups (22). Potential mechanisms by which H. pylori could protect the esophagus include ammonia generation, decreased gastric acidity as a consequence of corpus gastritis, and gastrin liberation increasing the pressure of the lower esophageal sphincter (23). It is, and remains, highly controversial whether or not H. pylori really does protect the esophagus, and whether eradicating the infection may provoke or unmask reflux disease (22, 23). Available data concerning the role of NSAIDs in the development of esophagitis are controversial. We failed to find any difference between the NERD and ERD groups, which is in line with recent findings in large double-blind, placebo controlled studies comparing conventional nonselective NSAIDs, COX-2-selective NSAIDs, and placebo (24). In another case-control study NSAIDs increased the risk of esophageal ulcers only, possibly via the mechanism of topical esophageal injury (25). In conclusion, some patient-associated factors increase the risk of erosive esophagitis as opposed to nonerosive reflux disease. However, no single factor is capable of predicting mucosal damage with clinically sufficient certainty, and endoscopy is therefore still necessary in all GERD patients if valid information on the state of the esophageal mucosa is needed. ACKNOWLEDGMENT This study was supported by a grant from AstraZeneca. Reprint requests and correspondence: Dr. Joachim Labenz, Department of Medicine, Jung Stilling Hospital, Wichernstr. 40, 57074 Siegen, Germany. Received November 29, 2003; accepted March 23, 2004. REFERENCES 1. Aro P, Ronkainen J, Storskrubb T, et al. Quality of life in a general population with gastroesophageal reflux symptoms and/or esophagitis. A report from the Kalixanda Study. Gastroenterology 2003;124(Suppl 1):A-168. 2. Eisen G. The epidemiology of gastroesophageal reflux disease: What we know and what we need to know. Am J Gastroenterol 2001;96(Suppl):S16–S18. 3. Hollenz M, Stolte M, Labenz J. Prevalence of gastrooesophageal reflux disease in general practice. Dtsch Med Wochenschr 2002;127:1007–12. 1656 Labenz et al. 4. Locke III GR, Talley NJ, Fett SL, et al. Prevalence and clinical spectrum of gastroesophageal reflux: A populationbased study in Olmsted County, Minnesota. Gastroenterology 1997;112:1448–56. 5. Galmiche JP, Zerbib F. Mechanisms of gastro-oesophageal reflux disease and potential targets for anti-reflux therapy. In: Farthing MJG, Bianchi Porro G, eds. New horizons in gastrointestinal and liver disease: Mechanisms and management. Paris: John Libbey Eurotext 1999:3– 15. 6. Fass R, Ofman JJ. Gastroesophageal reflux disease—Should we adopt a new conceptual framework? Am J Gastroenterol 2002;97:1901–9. 7. Cameron AJ, Arona AS. Barrett’s esophagus and reflux esophagitis: Is there a missing link? Am J Gastroenterol 2002;97:273–8. 8. El-Serag HB, Sonnenberg A. Associations between different forms of gastro-oesophageal reflux disease. Gut 1997;41:594–9. 9. Avidan B, Sonnenberg A, Schnell TG, et al. There are no reliable symptoms for erosive oesophagitis and Barrett’s oesophagus: Endoscopic diagnosis is still essential. Aliment Pharmacol Ther 2002;16:735–42. 10. Tew S, Jamieson GG, Pilowsky I, et al. The illness behavior of patients with gastroesophageal reflux disease with and without endoscopic esophagitis. Dis Esophagus 1997;10:9– 15. 11. Lundell L, Dent J, Bennett J, et al. Endoscopic assessment of oesophagitis: Clinical and functional correlates and further validation of the Los Angeles classification. Gut 1999;45:172–80. 12. El-Serag HB, Johanson JF. Risk factors for the severity of erosive esophagitis in Helicobacter pylori-negative patients with gastroesophageal reflux disease. Scand J Gastroenterol 2002;37:899–904. 13. Carlsson R, Dent J, Watts R, et al. Gastro-oesophageal reflux disease in primary care: An international study of different treatment strategies with omeprazole. Eur J Gastroenterol Hepatol 1998;10:119–24. 14. Lagergren J, Bergström R, Nyrén O. No relation between body mass and gastro-oesophageal reflux symptoms 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. in a Swedish population based study. Gut 2000;47:26– 9. Locke GR, Talley NJ, Fett SL, et al. Risk factors associated with symptoms of gastroesophageal reflux. Am J Med 1999;106:642–9. Lundell L, Ruth M, Sandberg N, et al. Does massive obesity promote abnormal gastroesophageal reflux? Dig Dis Sci 1995;40:1632–5. Wilson LJ, Ma W, Hirschowitz BI. Association of obesity with hiatal hernia and abnormal gastroesophageal reflux. Am J Gastroenterol 1999;94:2840–4. Kjellin A, Ramel S, Rössner S, et al. Gastro-oesophageal reflux in obese patients is not reduced by weight reduction. Scand J Gastroenterol 1996;31:1047–51. Mathus-Vliegen EMH, Tytgat GNJ. Gastro-oesophageal reflux in obese subjects: Influence of overweight, weight loss and chronic balloon distension. Scand J Gastroenterol 2002;37:1246–52. Meining A, Classen M. The role of diet and lifestyle measures in the pathogenesis and treatment of gastroesophageal reflux disease. Am J Gastroenterol 2000;95:2692–7. Smout AJPM. Endoscopy-negative acid reflux disease. Aliment Pharmacol Ther 1997;11(Suppl 2):81–5. Labenz J. Helicobacter pylori eradication leads to gastrooesophageal reflux disease. In: Hunt RH, Tytgat GNJ, eds. Helicobacter pylori. Basic mechanisms to clinical cure 2002. Dordrecht: Kluwer Academic Publishers, 2003:243– 52. Dent J. Helicobacter pylori eradication does not lead to gastro-oesophageal reflux disease. In: Hunt RH, Tytgat GNJ, eds. Helicobacter pylori. Basic mechanisms to clinical cure 2002. Dordrecht: Kluwer Academic Publishers, 2003:243– 52. Laine L, Goldstein JL, Fort JG. Do NSAIDs induce esophagitis? The effect of non-specific NSAIDs and COX-2 specific inhibitors ± low-dose Aspirin in prospective, randomized, double-blind, placebo-controlled endoscopic trials. Gastroenterology 2003;124(Suppl 1):A-511. Avidan B, Sonnenberg A, Schnell TG, et al. Risk factors for erosive reflux esophagitis: A case-control study. Am J Gastroenterol 2001;96:41–6.