GASTROENTEROLOGY 1999;116:1305 1309 Helicobacter pylori and Risk of Ulcer Bleeding Among Users of Nonsteroidal Anti-inflammatory Drugs: A Case-Control Study CLAUS AALYKKE,* JENS MARTIN LAURITSEN,*, JESPER HALLAS,*, SUSANNE REINHOLDT, KAREN KROGFELT, and KARSTEN LAURITSEN* Departments of *Medical Gastroenterology and Geriatrics, Odense University Hospital, Odense; Institute of Community Health and Department of Clinical Pharmacology, Odense University, Odense; and Department of Gastrointestinal Infections, Statens Serum Institut, Copenhagen, Denmark Background & Aims: Peptic ulcer complications related to use of nonsteroidal anti-inflammatory drugs (NSAIDs) are among the most common serious adverse drug reactions. Whether Helicobacter pylori infection potentiates this gastrointestinal toxicity of NSAIDs is still unresolved. In this study, we investigated the role of H. pylori as a cause of bleeding peptic ulcer among NSAID users. Methods: A case-control study of current users (n 132) of NSAIDs (including acetylsalicylic acid), admitted because of bleeding peptic ulcer, was performed. Controls were 136 NSAID users without gastrointestinal complications. H. pylori was diagnosed by either increased levels of serum immunoglobulin G or by 13 C-urea breath test. Results: Fifty-eight (44%) case subjects had a bleeding gastric ulcer, 54 (41%) had a bleeding duodenal ulcer, 12 (9%) had both gastric and duodenal ulcers, and 8 (6%) had hemorrhagic gastritis. H. pylori was present in 75 (57%) cases compared with 59 (43%) controls. The adjusted odds ratio of bleeding peptic ulcer among NSAID users associated with H. pylori infection was 1.81 (95% confidence interval, 1.02 3.21). H. pylori accounted for approximately 24% of bleeding peptic ulcers among elderly NSAID users. Conclusions: NSAID users infected with H. pylori have an almost twofold increased risk of bleeding peptic ulcer compared with NSAID users without H. pylori. Alow prevalence of Helicobacter pylori has been reported among patients with complicated ulcer disease, raising the question whether H. pylori infection is important for this outcome. 1 Intervention studies have shown that H. pylori eradication after ulcer bleeding reduces the rebleeding rate. 2,3 Whether this applies to the risk of bleeding peptic ulcer among users of nonsteroidal anti-inflammatory drugs (NSAIDs) is still unresolved, and only few studies have examined this relationship. 4 8 We describe the results of a prospective case-control study designed to establish whether H. pylori infection is a risk factor for bleeding peptic ulcer among NSAID users. Patients and Methods Selection of Case and Control Subjects Case subjects in this prospective case-control study were consecutive patients presenting with signs of upper gastrointestinal bleeding who fulfilled three criteria: (1) admission to Department of Medical Gastroenterology, Odense University Hospital, between October 31, 1994, and March 8, 1996; (2) current NSAID use including acetylsalicylic acid (aspirin) in analgesic or antithrombotic doses, with current use defined as at least one dose in the week before the gastrointestinal bleeding episode; and (3) endoscopy-verified peptic ulcer or hemorrhagic gastritis found by the endoscopist to be a probable bleeding source. Eligible controls were current users of NSAIDs including aspirin without signs of ulcer bleeding at the interview, recruited from the Outpatient Clinic of Rheumatology and Department of Geriatrics, Odense University Hospital. Current use was defined as at least one dose taken in the week before interview. Because the risk of bleeding peptic ulcer depends on age and sex, 9 controls were frequency-matched by age in 10-year bands and by sex to increase study efficiency. 10 Neither case nor control subjects were subjected to exclusion criteria with respect to past ulcer history or other comorbidity. A structured interview covered potential risk factors among cases and controls: a history of peptic ulcer, a history of bleeding peptic ulcer, dyspepsia within the last 3 months, duration and dose of NSAID use, smoking habits, alcohol consumption, and use of oral corticosteroids and anticoagulants. Case subjects were typically interviewed the day after admission. Dose of NSAID was expressed as the number of defined daily doses (DDD) 11 consumed per week. One DDD of aspirin was defined as 150 mg, and low-dose aspirin was defined as a weekly consumption of 7 DDD or 1050 mg. Cases and controls were excluded if the interview or 13 C-urea breath test were not possible or unethical to perform. Nine cases were excluded a posteriori: 6 cases with doubtful bleeding source, 1 case with no NSAID use in the week before ulcer bleeding, and 2 cases who were not enrolled consecutively. One Abbreviations used in this paper: CI, confidence interval; DDD, defined daily dose. 1999 by the American Gastroenterological Association 0016-5085/99/$10.00
1306 AALYKKE ET AL. GASTROENTEROLOGY Vol. 116, No. 6 case was excluded because he died before the interview, and 1 control was excluded because both serological test and breath test results were missing. Informed consent to study participation was obtained before study enrollment. Of the eligible subjects, 43 control and 1 case subjects refused participation. The study was approved by the Regional Ethics Committee of Funen and Vejle Counties. H. pylori Infection Serum IgG antibodies to H. pylori were measured in duplicate by a second-generation enzyme-linked immunosorbent assay (Pylori-stat; Bio-Whittaker, Walkersville, MD), previously validated in a Danish population. 12 According to the manufacturer, an immunoglobulin (Ig) G level of 1 U was considered positive. 13 C-Urea breath test was performed after fasting, and a standard meal was given, followed by intake of 100 mg 13 C-labeled urea dissolved in 100 ml water. Duplicate breath samples were taken before ingestion of the urea and after 30 minutes and considered positive if the increase in 13 C/ 12 C ratio was 5.8 per mil. Among case subjects, breath test was done as soon as oral feeding was allowed after the bleeding episode. A patient was considered H. pylori positive if either the serological test or breath test was positive. 13,14 The determination of the H. pylori status was based on the results of the remaining test if one of the test results was missing (serological test, n 4; breath test, n 26). Statistics A sample size of 137 cases and controls was calculated as sufficient to detect an odds ratio of 2.0 (2 0.05, 0.2), given a prevalence of H. pylori infection of 50% among the controls. 15 Data were analyzed with the SPSS statistical package for Windows (Norusis MJ. SPSS for Windows, release 7.0.; Chicago, IL: SPSS, Inc., 1996). Crude odds ratios and 95% confidence intervals (95% CIs) were calculated for H. pylori infection and potential risk factors. The unconditional multivariate logistic regression analysis done by SPSS advanced statistics included the potential risk factors of sex, age, history of peptic ulcer, history of bleeding peptic ulcer, dyspepsia within 3 months, duration and dose of NSAID use, smoking habits, and alcohol consumption. Results Eventually, 132 case (63 women and 69 men; median age, 72 years; range, 30 95) and 136 control (77 women and 59 men; median age, 66 years; range, 31 93) subjects were included in the study. Case subjects were older than controls, and there were more male cases. However, these differences were adjusted in the logistic regression analysis. 10 Fifty-six (42%) cases were exclusively aspirin users, 43 (33%) were NSAID users, and 33 (25%) used both NSAIDs and aspirin. For comparison, 42 (31%) of the controls were aspirin users only, 75 (55%) NSAID users only, and 19 (14%) controls had used both. Among users of aspirin only, 30 (23%) cases vs. 33 (24%) controls had used low-dose aspirin. Fifty-eight (44%) cases had upper gastrointestinal bleeding caused by gastric ulcer, 54 (41%) had bleeding duodenal ulcer, 12 (9%) had both gastric and duodenal ulcer, and 8 (6%) had hemorrhagic gastritis. The mean hemoglobin by admission was 5.6 mmol/l (range, 2.2 9.6). The normal range is 8 11 mmol/l for men and 7 10 mmol/l for women. Fifty-six (41%) controls had rheumatoid arthritis, 33 (24%) had other rheumatic diseases, and 47 (35%) had nonrheumatic diseases. The frequency of various risk factors and the corresponding crude odds ratios for bleeding peptic ulcer among NSAID users are listed in Table 1. H. pylori infection was present in 75 (57%) cases compared with 59 (43%) controls. The crude odds ratio for bleeding peptic ulcer associated with H. pylori infection among users of NSAIDs was 1.72 (95% CI, 1.06 2.78). A history of bleeding peptic ulcer, current dyspepsia, and alcohol consumption of more than 5 drinks per week were also risk factors in the crude analysis. A reliable estimate for use of oral corticosteroids Table 1. Crude Odds Ratios of Bleeding Peptic Ulcer Among NSAID Users Associated With Potential Risk Factors Risk factor Cases (n 132) Controls (n 136) Crude odds ratio (95% CI) H. pylori infection No 57 (43) 77 (57) Reference Yes 75 (57) 59 (43) 1.72 (1.06 2.78) History of peptic ulcer No 103 (78) 117 (86) Reference Yes 29 (22) 19 (14) 1.73 (0.92 3.28) History of bleeding peptic ulcer No 111 (84) 129 (95) Reference Yes 21 (16) 7 (5) 3.49 (1.43 8.51) Dyspepsia within 3 mo No 50 (38) 87 (64) Reference Yes 82 (62) 49 (36) 2.91 (1.77 4.78) Smoking No 86 (65) 84 (62) Reference Yes 46 (35) 52 (38) 0.86 (0.53 1.42) Alcohol 5 drinks/wk 89 (67) 110 (81) Reference 5 drinks/wk 43 (33) 26 (19) 2.04 (1.17 3.58) Duration of NSAID/aspirin use 3 mo 67 (51) 50 (37) Reference 3 12 mo 21 (16) 34 (25) 0.46 (0.24 0.89) 12 mo 44 (33) 52 (38) 0.63 (0.37 1.09) Dose of NSAID/aspirin use 3.5 DDD/wk 13 (10) 14 (10) Reference 3.6 7.0 DDD/wk 42 (32) 65 (48) 0.70 (0.30 1.63) 7.0 DDD/wk 77 (58) 57 (42) 1.45 (0.64 3.33) NOTE. Data are numbers of subjects with percentages in parentheses.
June 1999 H. PYLORI AND ULCER BLEEDING IN NSAID USERS 1307 could not be assessed because the controls were mainly enrolled from the Outpatient Clinic of Rheumatology. Likewise, because anticoagulant therapy only was used by 2 case and 3 control subjects, analysis of this variable was precluded. Exclusion of subjects without breath test or without serological test (12 cases and 18 controls) did not affect the main result (crude odds ratio, 1.84; 95% CI, 1.10 3.08). The association between H. pylori and bleeding peptic ulcer was similar among aspirin users and NSAID users (crude odds ratio, 1.78; 95% CI, 0.79 3.99; vs. 1.84; 95% CI, 0.86 3.95). In the logistic regression analysis (Table 2), H. pylori infection persisted as an independent risk factor for bleeding peptic ulcer among NSAID users (adjusted odds ratio, 1.81; 95% CI, 1.02 3.21). As in the crude analysis, history of peptic ulcer bleeding, current dyspepsia, and alcohol consumption were associated with an increased risk for bleeding peptic ulcer among NSAID users. The odds ratio of bleeding peptic ulcer associated with H. pylori in an NSAID user was 2.21 (95% CI, 0.35 19.64) in individuals with a history of bleeding peptic ulcer compared with 1.78 (95% CI, 1.07 2.99) in those without such a history. To investigate the joint effect of H. pylori infection and history of bleeding peptic ulcer, we used a common reference group, individuals with no history of bleeding peptic ulcer and uninfected with H. pylori. The adjusted odds ratio was 1.73 (95% CI, 0.95 3.15) for those infected with H. pylori but without a history of ulcer bleeding; 3.29 (95% CI, 0.85 12.64) for those with history of bleeding peptic ulcer but without H. pylori infection; and 9.74 (95% CI, 1.73 54.64) for Table 2. Adjusted Odds Ratios of Bleeding Peptic Ulcer Among NSAID Users Associated With Potential Risk Factors Risk factor Adjusted odds ratio (95% CI) a H. pylori infection 1.81 (1.02 3.21) History of peptic ulcer 0.93 (0.41 2.12) History of bleeding peptic ulcer 4.08 (1.37 12.13) Dyspepsia within last 3 mo 2.85 (1.63 5.00) Smoking 0.91 (0.48 1.71) Alcohol ( 5 drinks/wk) 2.39 (1.16 4.89) Duration of NSAID/aspirin use 3 mo Reference 3 12 mo 0.49 (0.23 1.04) 12 mo 0.69 (0.37 1.29) Dose of NSAID/aspirin use 3.5 DDD/wk Reference 3.6 7.0 DDD/wk 0.95 (0.34 2.68) 7.0 DDD/wk 2.25 (0.84 6.05) a Logistic regression analysis adjusting for sex, age ( 60 years, 61 75 years, and 75 years), history of peptic ulcer, history of bleeding peptic ulcer, dyspepsia within 3 months, duration of NSAID use, dose of NSAID use, smoking, and alcohol. those with both risk factors. The relative excess risk due to this interaction 10,16 was 5.72 (95% CI, 0.92 to 12.36). A hypothesis-generating subanalysis of cases with gastric and duodenal ulcers was performed to reveal if the effect of H. pylori infection was confined to any specific site of bleeding. H. pylori infection was present in 55% of patients with bleeding gastric ulcer and 56% of patients with bleeding duodenal ulcer. H. pylori infection increased the risk of bleeding gastric (adjusted odds ratio, 2.39; 95% CI, 1.09 5.26) as well as of bleeding duodenal ulcer (adjusted odds ratio, 1.66; 95% CI, 0.78 3.55), although this was not statistically significant for the latter group. According to a population-based prevalence of H. pylori infection of 25.9%, 17 the population attributable risk percent 18 of H. pylori infection for bleeding peptic ulcer among NSAID users can be estimated as 16% (95% CI, 7 to 34) and among the elderly NSAID users as 24% (95% CI, 3 41). Discussion This case-control study suggests that H. pylori infection is an independent risk factor roughly doubling the risk of bleeding peptic ulcer among NSAID users. This increased risk persisted even after extensive adjustment for confounding factors. Consistent with previous reports, 1,7 only 57% of our cases were H. pylori positive in contrast to the usually much higher prevalences found in uncomplicated ulcers. We restricted the study group to NSAID users because our primary goal was to investigate the role of H. pylori specifically among NSAID users. We could not confirm previous data reporting either no increased risk 5 or a decreased 7 risk associating H. pylori with ulcer bleeding. Different methodology may account for some of these discrepancies. In one study 7 the controls consisted exclusively of patients with uncomplicated ulcer. In another study 5 a clinically relevant effect of H. pylori among NSAID users cannot be ruled out because confidence intervals were wide. In contrast to others, 6 we found no difference between aspirin and NSAID users with respect to the strength of association between H. pylori and bleeding peptic ulcer. Because only NSAID users were recruited for the study, straightforward estimates of the interaction between NSAID use and H. pylori infection as risk factors for upper gastrointestinal bleeding were precluded. However, because the odds ratio of H. pylori in our study parallels risk estimates in studies 4,5 without restriction to NSAID users, major interaction between the two risk factors is unlikely.
1308 AALYKKE ET AL. GASTROENTEROLOGY Vol. 116, No. 6 Age and sex were precluded from risk assessment for bleeding peptic ulcer because they were used as matching criteria. Like others, 9 we found a history of bleeding peptic ulcer was an independent risk factor for bleeding peptic ulcer among NSAID users. The presence of H. pylori infection probably further enhanced that risk, suggesting these patients as the most gratifying candidates for eradication of H. pylori infection. Recently, the efficacy of H. pylori eradication was compared with omeprazole maintenance therapy for preventing recurrent ulcer hemorrhage. 8 Eradication of H. pylori appeared as effective as maintenance therapy among aspirin users, whereas in NSAID users, maintenance therapy was superior to the eradication therapy. However, it was uncertain to what extent H. pylori eradication in itself reduced the risk of recurrent bleeding because no placebo group was provided. A possible limitation of this study might be the heterogeneity of NSAIDs included. However, no NSAID, including prophylactic aspirin, has yet proven to be without gastrointestinal toxicity. 19 21 Furthermore, a substantial part of the differences in gastrointestinal toxicity of NSAIDs seems to be dose related. 19 21 Although we could not adjust for the different type of NSAIDs, we did adjust for the NSAID dose. Also, we found no substantial difference in H. pylori prevalence between NSAID and aspirin users. A further limitation of the present study could have been selection bias, because the primary source for control selection was rheumatic patients. However, the prevalence of H. pylori infection in arthritic patients has been reported to be identical to asymptomatic individuals without arthritis. 22 Furthermore, the 43% prevalence of H. pylori infection among our controls is similar to the H. pylori prevalence in the population of comparable age. 17 Selection bias among the case subjects was unlikely because they were enrolled consecutively. Diagnostic biases regarding the H. pylori status were unlikely because both diagnostic procedures were performed blinded to case or control status. Information of other potential risk factors may have been liable to recall bias. However, by using the same structured interview for cases and controls we endeavored to minimize this type of bias. Furthermore, selective recall would only have an indirect effect on the results (i.e., on the confounder adjustment), because our primary exposure, H. pylori status, was unknown to all subjects. There was hardly any major confounding in our study, judged by the small difference between the unadjusted and adjusted odds ratios (1.72 vs. 1.81). The population attributable risk percent of H. pylori infection takes into account both the strength of the association and the prevalence in the population. It may also be interpreted as the percent of patients with bleeding peptic ulcer among NSAID users that could be prevented if H. pylori infection was eradicated. A population attributable risk percent of 24% among elderly NSAID users suggests H. pylori infection is important, although of less significance, than the approximately 50% observed in uncomplicated peptic ulcer. 23 In summary, NSAID users infected with H. pylori have an almost twofold increased risk of bleeding peptic ulcer compared with NSAID users without H. pylori infection. The result from this prospective case-control study suggests a potential for halving the risk of ulcer complications among NSAID users by eradicating H. pylori. In high-risk patients, this seems insufficient as a preventive measure and should be accompanied by prophylactic use of an appropriate antiulcer drug. References 1. Laine L, Peterson WL. Bleeding peptic ulcer. N Engl J Med 1994;331:717 727. 2. Jaspersen D, Koerner T, Schorr W, Brennenstuhl M, Raschka C, Hammar CH. Helicobacter pylori eradication reduces the rate of rebleeding in ulcer hemorrhage. Gastrointest Endosc 1995;41: 5 7. 3. Rokkas T, Karameris A, Mavrogeorgis A, Rallis E, Giannikos N. Eradication of Helicobacter pylori reduces the possibility of rebleeding in peptic ulcer disease. Gastrointest Endosc 1995;41: 1 4. 4. Köhl H, Leverkus F, Labenz J. H. pylor infection increases the risk of peptic ulcer bleeding a case control study (abstr). Gut 1996;39:(suppl 3):A231. 5. Cullen DJ, Hawkey GM, Greenwood DC, Humphreys H, Shepherd V, Logan RFA, Hawkey CJ. Peptic ulcer bleeding in the elderly: relative roles of Helicobacter pylori and non-steroidal antiinflammatory drugs. Gut 1997;41:459 462. 6. Hawkey GM, Stack WA, Pearson G, Everitt S, Logan RFA, Hawkey CJ. Non steroidal anti inflammatory drugs, aspirin and Helicobacter pylori as risk factors for bleeding peptic ulcers (abstr). Gut 1997;41:(suppl 3):A5. 7. Pilotto A, Leandro G, Di Mario F, Franceschi M, Bozzola L, Valerio G. Role of Helicobacter pylori infection on upper gastrointestinal bleeding in the elderly a case-control study. Dig Dis Sci 1997;42: 586 591. 8. Chan FKL, Sung JY, Suen R, Lee YT, Leung WK, Leung VKS, Wu JCY, Chung SCS. Eradication of H. pylori versus maintenance acid suppression to prevent recurrent ulcer hemorrhage in high risk NSAID users: a prospective randomized study (abstr). Gastroenterology 1998;114:(suppl):A87. 9. Garcia Rodriguez LA, Jick H. Risk of upper gastrointestinal bleeding and perforation associated with individual non-steroidal anti-inflammatory drugs. Lancet 1994;343:769 772. 10. Rothman KJ. Modern epidemiology. Boston: Little Brown, 1986: 237 283; 311 326. 11. WHO collaborating centre for drug statistics and methodology. Anatomical therapeutic chemical (ATC) classification index including defined daily doses (DDDs) for plain substances. Oslo, Norway: 1995. 12. Jensen AKV, Andersen LP, Wachmann CH. Evaluation of eight commercial kits for Helicobacter pylori IgG antibody detection. APMIS 1993;101:795 801. 13. Feldman RA, Evans SJ. Accuracy of diagnostic methods used for
June 1999 H. PYLORI AND ULCER BLEEDING IN NSAID USERS 1309 epidemiological studies of Helicobacter pylori. Aliment Pharmacol Ther 1995;9 (suppl 2):21 31. 14. McColl KEL. The role of Helicobacter pylori eradication in the management of acute bleeding peptic ulcer. Eur J Gastroenterol Hepatol 1995;7:753 755. 15. Strom BL. Pharmacoepidemiology. 2nd ed. Chichester, England: Wiley, 1994. Appendix A: sample size tables. 16. Hosmer DW, Lemeshow S. Confidence interval estimation of interaction. Epidemiology 1992;3:452 456. 17. Andersen LP, Rosenstock SJ, Bonnevie O, Jorgensen T. Seroprevalence of immunoglobulin G, M, and A antibodies to Helicobacter pylori in an unselected Danish population. Am J Epidemiol 1996;143:1157 1164. 18. Lilienfeld DE, Stolley PD. Foundations of epidemiology. 3 ed. New York: Oxford University Press, 1994:226 251 and 316 320. 19. Langman MJS, Weil J, Wainwright P, Lawson DH, Rawlins MD, Logan RFA, Murphy M, Vessey MP, Colin-Jones DG. Risks of bleeding peptic ulcer associated with individual non-steroidal anti-inflammatory drugs. Lancet 1994;343:1075 1078. 20. Henry D, Lim LL, Garcia Rodríguez LA, Perez-Gutthann S, Carson JL, Griffin M, Savage R, Logan R, Moride Y, Hawkey CJ, Hill S, Fries JT. Variability in risk of gastrointestinal complications with individual non-steroidal anti-inflammatory drugs: results of a collaborative meta-analysis. Br Med J 1996;312:1563 1566. 21. Weil J, Colin-Jones D, Langman M, Lawson D, Logan R, Murphy M, Rawlins M, Vessey M, Wainwright P. Prophylactic aspirin and risk of peptic ulcer bleeding. Br Med J 1995;310:827 830. 22. Graham DY, Lidsky MD, Cox AM, Evans DJ, Alpert L, Klein PD, Sessons SL, Michaletz PA, Saeed ZA. Long-term nonsteroidal antiinflammatory drug use and Helicobacter pylori infection. Gastroenterology 1991;100:1653 1657. 23. Kurata JH, Nogawa AN. Meta-analysis of risk factors for peptic ulcer. Nonsteroidal antiinflammatory drugs, Helicobacter pylori, and smoking. J Clin Gastroenterol 1997;24:2 17. Received February 5, 1998. Accepted March 4, 1999. Address requests for reprints to: Claus Aalykke, M.D., Department of Medical Gastroenterology S, Odense University Hospital, DK-5000 Odense C, Denmark. e-mail: Aalykke@med.ou.dk; fax: (45) 6591-7481. Supported by A. J. Andersen og Hustrus Fond and the Institute of Clinical Research, Odense University. Klatskin of the Klatskin tumor Copyright holder unknown. Photo obtained from the National Library of Medicine website (http://www.nlm. gov). Gerald Klatskin (1910 1986) was born in New York City to a family of Russian descent. During his undergraduate years at Cornell he developed a keen interest in histology and in 1933 was awarded his M.D. degree, graduating first in his class. His resident training was at the Yale New Haven Hospital and at the Strong Memorial Hospital in Rochester, NY. In World War II he served as a medical officer in Calcutta, India, where he acquired a particular interest in hepatitis and amebic abscess of the liver. Returning to Yale in 1946, he established a laboratory that became a renowned Liver Study Unit. Histological specimens prepared there with the aid of Hazel Hubbel, a talented technician, were recognized as works of art. Klatskin was among the first to link the Australia antigen to hepatitis B, to relate acute liver injury to prognosis, and to recognize that alcoholic hyperlipidemia could give rise to pancreatitis. In 1965 he described the unique features of cholangiocarcinoma at the bifurcation of the main intrahepatic bile ducts, which came to be called the Klatskin tumor. It was his dream to create a histological atlas of diseases of the liver, an ambition brought to fruition in 1993 by Harold Conn, his first fellow and longtime collaborator at Yale. Known by some of his many students as Great Bear because of his seemingly burly mien, Klatskin was, in fact, a kindly and unpretentious mentor. Contributed by WILLIAM S. HAUBRICH, M.D. Scripps Clinic and Research Foundation, La Jolla, California