Long-Term Nonsteroidal Antiinflammatory Drug Use and Helicobacter py1orj Infection DAVID Y. GRAHAM, MARTIN D. LIDSKY, ANNALEE M. COX, DOYLE J. EVANS, Jr., DOLORES G. EVANS, LESLEY ALPERT, PETER D. KLEIN, SANDRA L. SESSOMS, PATRICE A. MICHALETZ, and ZAHID A. SAEED Department of Medicine, Division of Molecular Virology, Department of Pathology, Department of Pediatrics, and United States Department of Agriculture/Agricultural Research Service Children’s Nutrition Research Center, Baylor College of Medicine; and Veterans Affairs Medical Center, Houston, Texas
This study investigates whether patients who take nonsterodial antiinflammatory drugs are more likely to have Helicobacter pylori gastritis than agematched individuals who do not take nonsteroidal antiinflammatory drugs, and whether patients who take nonsteroidal antiinflammatory drugs who are also infected with H. pylori are more likely to have dyspepsia, mucosal damage, or ulcers than those who are not infected. Two studies were performed, one serological and the other endoscopic, both in arthritis patients receiving nonsteroidal antiinflammatory drugs chronically. The presence of H. pylori was identified with a sensitive enzyme-linked immunosorbent assay test. One hundred eighty-three patients participated in the serological study and 75 patients in the endoscopic study. The frequency of H. pylori infection increased with age, independent of nonsteroidal antiintlammatory drug use; the ageadjusted frequency of H. pylori infection in arthritis patients paralleled that of 351 asymptomatic individuals without arthritis. The frequency of H. pylori infection increased from 30.7% in age group 21-30 years to 73.4% in age group 61-75 years. Nonsteroida1 antiinflammatory drug-induced mucosal injury, either hemorrhages or erosions, was more frequent in those without H. pylori infection than with infection (61% vs.32% for hemorrhages and 57% vs. 34% for erosions for those without and with H. pylori infection; only the difference in the frequency of hemorrhages was significant, P < 0.05). No difference was observed in the presence of dyspeptic symptoms between those with and without H. pylori infection. These data suggest that nonsteroidal antiinflammatory drug-induced damage to the gas-
troduodenal mucosa does not increase the susceptibility to H. pylori infection.
t is now accepted that Helicobacter pylori causes gastritis, but it is not known whether preexisting damage renders the gastroduodenal mucosa more susceptible to H. pylori infection or is required for the infection to become established (1,2). The association between H. pylori infection and peptic ulcer disease is strongest for the presence of duodenal ulcer (3). The use of nonsteroidal antiinflammatory drugs (NSAIDs) is also associated with peptic ulcer disease (1). One might now classify peptic ulcer disease into three main categories according to the proposed pathophysiology: H. pylori-associated, NSAID-associated, and hypersecretory (e.g., associated with the ZollingerEllison syndrome). Recently, interest has focused on the possible relationship between H. pylori infection and NSAID-associated dyspepsia, ulcers, and minor mucosal damage (2,4-7).We sought to determine whether arthritis patients who take NSAIDs are more likely to have H. pylori gastritis than age-matched individuals who do not take NSAIDs and whether patients who take NSAIDs who are also infected with H. pylon’ are more likely to have dyspepsia, mucosal damage, or ulcers than those not infected with H. pylori.
Abbreviation used in this paper: ELISA, enzyme-linked immunosorbent assay. o 1991 by the American Gastroenterological Association 0016-5065/91/$3.00
GRAHAM ET AL.
Materials Assessment Infection
GASTROENTEROLOGY Vol. loo, No. 6
of the Presence of an H. pylori
For all groups of individuals studied, we used a sensitive and specific enzyme-linked immunosorbent assay (ELISA) (positive predictive value, 100%; negative predicthe presence of H. pylori tive value, 98.6%) to identify infection (8). In patients with ulcers, the urea breath test, gastric biopsy with Warthin-Starry stain, or both were used to confirm active H. pylori infection (9,lO). Each individual completed a self-administered questionnaire and the results were reviewed by a trained interviewer. In addition to demographic information, the instrument asked what medication they were taking and for how long. Each individual was questioned about the presence and the frequency of symptoms referable to the upper gastrointestinal tract occurring over the preceding year, including indigestion, heartburn, sour stomach, nauseaJ vomiting, and burning pain in the stomach area and burning pain in the chest. Drug dosage was completed by the physician. Two studies were performed. One study was a compari(by ELISA) son of the prevalence of H. pylori infection between 351 healthy individuals with no symptoms referable to the upper gastrointestinal tract (11) and a group with rheumatoid arthritis, osteoarthritis, ankylosing spondylitis, or systemic lupus erythematosus who were also receiving chronic NSAID therapy. The second study asked whether patients who take NSAIDs who are also infected with H. pylori are more likely to have mucosal damage or ulcers than NSAID users without H. pylori gastritis. Enrollees in the second study underwent upper gastrointestinal endoscopy. Enrollment was restricted to patients with rheumatoid arthritis or osteoarthritis receiving chronic NSAID therapy (there was an overlap of 12 patients between the groups). At endoscopy (Fujinon videoendoscopes; Wayne, NJ), the degree and severity of NSAID-induced gastroduodenal damage was assessed using a modification of the scoring system proposed by Lanza et al. (12) (Table 1). An ulcer was defined as a circumscribed break in the mucosa 2 5 mm in diameter with an exudate and apparent depth. Four antral and one lesser curvature gastric corpus biopsy specimens were taken. In addition, blood was taken for determination of ABO blood type. Patients were excluded if they had peptic ulcer diagnosed within the previous 12 months, were taking antiulcer medications, or had recently taken antibiotics or bismuthcontaining medications.
Endoscopic Scoring System for Nonsteroidal Antiinflammatorv Drug-induced Mucosal Iniurv
Mucosal hemorrhages 0 No evidence of mucosal hemorrhages 1 One area of mucosal hemorrhage 2 More than one area of mucosal hemorrhage but neither numerous or widespread 3 Numerous (three or more) areas of mucosal hemorrhage 4 Widespread involvement of the gastric mucosa with mucosal hemorrhage Global score 0 Normal stomach (includes grade 1,mucosal hemorrhages) A Mucosal hemorrhage (grade 2 or greater mucosal hemorrhage) B Erosions (one or two) C Three or more areas of gastric erosion D Large areas of erosion with widespread involvement of the stomach or duodenum or ulcer NOTE. The same system was used separately for the stomach and duodenal bulb mucosa.
Results Serology Study
We obtained symptom assessment and serum samples from 183 patients: 96 with rheumatoid arthritis, 45 with osteoarthritis, 21 with ankylosing spondylitis, and 21 with systemic lupus erythematosus. There were 127 men and 56 women between the ages of 21 and 75. H. pylori infection was present in 55.6%. The frequency of H. pylori infection increased with age from 30.7% in those aged 21-30 years to 73.4% in those ages 61-75 years and was not statistically different from the age-adjusted rate of infection in the control group of 351 asymptomatic individuals (11) without arthritis (Figure 1). No relationship was found between any of the symptoms assessed and the presence of H. pylori infection. 100
got 00 -70.A’/
Data Analyses Comparison between groups for the frequency of H. pylori infection was done by logistic regression using the SAS program (SAS Institute, Cary, NC). Categorical data was analyzed by Fisher’s Exact Test. Significance was assumed for Pvalues I 0.05. The protocol was approved by the local institutional review committees in July 1988.
Figure 1. The percentage of chronic NSAID users with H. pylori infection is shown (n = 183). For comparison, the frequency of If. pylori infection in 351 asymptomatic adults who were not chronic NSAID users is also shown.
H. PYLORI AND NSAID-INDUCED GASTRODUODENAL
group None Duodenal Gastric
OABCDOABCDO12.34 58 4
0 9 0
Table 4. Relationship Between the Presence of Ulcer and Nonsteroidal Antiinflammatory Drug Used
Scores for Patients With Arthritis
Table 2. Endoscopic
Seventy-five patients (72 men, 3 women) between the ages 28 and 80 years (median age, 62 years) underwent endoscopic examination. A duodenal ulcer was found in 9 (12%), 8 of whom had H. pylori infection. Gastric ulcer was found in 4 (5%), 3 of whom had an H. pylori infection, and no ulcer was present in 62 (83%), 42 of whom had an H. pylori infection (in 1 the H. pylori status was unknown). The presence/absence of H. pylori infection in ulcer patients was confirmed in each instance by the presence stain on of H. pylori as determined by Warthin-Starry at least two antral biopsies. In 1 patient with duodenal ulcer, the data from ELISA were discordant from the histology; that patient had a negative ELISA but a positive urea breath test and histological evidence of H. pylori infection. The infection became recent as the ELISA was positive when retested after 2 months. The data were analyzed according to groups (no ulcer, duodenal ulcer, and gastric ulcer). There were no statistical differences between groups in relation to the endoscopic scores of NSAID-induced mucosal damage for either global scores or scores for gastric mucosal hemorrhages (Table 2). There were no statistical differences between groups in relation to presence of symptoms, presence of H. pylori infection, or ABO blood type (Table 3). The data were also analyzed to look for possible differences between patients with rheumatoid arthritis (n = 24) and osteoarthritis (n = 51); none were found. Finally, no NSAID appeared more or less ulcerogenic than the others, nor did combinations of an NSAID and aspirin or an NSAID and methotrexate (Table 4). In the group of 60 patients without ulcer in whom the H. pylori status was known, we compared whether a relationship was observed between the severity of visible gastroduodenal mucosal damage and H. pylori Table 3. Relationship Between the Presence of Ulcer and ABO Blood Type in Patients With Arthritis Blood type Ulcer
None Duodenal Gastric
Ibuprofen Naproxen Indomethacin Sulindac Piroxicam Tolmetin Trisalate Salsalate NSAID + aspirin NSAID + methotrexate
22 20 7
2 2 1 1 8
0 0 0 1 1
0 0 0 0
infection (Table 5). The frequency of any visible damage was greater in patients with H. pylori infection (49% compared with 24% for those without H. pylori infection), but the difference did not achieve statistical significance (P > 0.06). The subgroup, patients with gastric mucosal hemorrhages, showed a difference; hemorrhages were significantly less frequent in patients with H. pylori infection than those without H. pylori infection (P < 0.05) (Table 6). H. pylori Infection and Dyspepsia No obvious or statistical relationship was observed between the presence or absence of symptoms referable to the upper gastrointestinal tract and H. pylori infection. In the large serological study, we specifically looked for a possible relationship between H. pylori infection and dyspepsia, occurring at least once a week, or “severe” dyspepsia (defined as occurring at least once daily and which occurred in 17.8%); no relationship was observed. In the endoscopic study, symptoms referable to the upper gastrointestinal tract were present in 53% of patients without ulceration, 89% (8 of 9) of duodenal ulcer patients, and 75% (3 of 4) of gastric ulcer patients (P = NS for comparisons between those with and without ulcer). In the endoscopic study, a larger proportion of patients with H. pylori infection had symptoms than those without H. pylori infection, but the difference was not significant (P > 0.09). Table 5 shows the
Table 5. Relationship Between Endoscopic Finding, H. uvlori Infection, and DVSDeDSia H. p_vlori Present Absent
(n = 53) (n = 20)
Ulcers ll(91) 2 (50)
“Hemorrhages and erosions were not mutually exclusive categories; those with ulcers were not included and two patients with hemorrhages did not have data about H. pylon’ infection or symptoms. bNumber in parentheses shows percentage with dyspepsia.
GRAHAM ET AL.
Table 6. Relationship Between H. pylori Status and Mucosal Lesions in Patients Taking Nonsteroidal Antiinflammatory Drugs H. pylori
Positive Negative P
26 5 0.067
17 13 0.04
18 12 0.11
11 2 0.33
“Hemorrhages, erosions, and ulcer are not mutually exclusive; one patient with both hemorrhages and erosions was of unknown H. pylori status.
relationship between the presence of H. pylori infection, symptoms, and endoscopic findings. Patients with normal endoscopic findings and H. pylori infection were more likely to have symptoms (10 of 27) compared with those without H. pylori infection (1 of 7) (P < 0.04). The significance of this finding is unknown and because of the large number of comparisons may represent a type I error. Discussion
Because no difference was observed in the age-adjusted frequency of H. pylori infection between chronic NSAID users and healthy volunteers, it is suggested that NSAID-induced damage to the gastroduodenal mucosa does not result in increased susceptibility to H. pylori infection. Our study confirms and extends observations from an older study (13) that showed that resistance to aspirin-induced microbleeding correlated with the presence of polymorphonuclear cells in the gastric cytology (an indirect method of assessing the presence of an H. pylori infection). These results, if confirmed in acute NSAID administration studies in volunteers, suggest that subsequent endoscopic studies in which NSAIDs are administered to normal volunteers to compare drugs or to assess the ability of some therapeutic maneuver to prevent or reduce mucosal damage must either take the presence of H. pylori infection into account or use scoring systems that do not emphasize the presence of mucosal hemorrhages. We found that the positive predictive value of symptoms referable to the upper gastrointestinal tract for the presence of an ulcer was 26%; the negative predictive value was 93%. Our study thus reconfirms previous observations that the presence of symptoms, even those occurring at least once a week, does not predict the presence of an ulcer (1,14). The lack of an association between H. pylori infection and the presence of symptoms referable to the upper gastrointestinal tract suggests that H. pylori gastritis is either not responsible for NSAID-associated dyspepsia or, at best, is responsible for only a very small proportion of cases.
Vol. 100, No. 6
This prospective evaluation did find a relationship between ABO blood type and NSAID-associated ulcer. We also did not confirm the previous suggestion that there might be an association between blood type 0 and propensity to develop peptic ulcers in rheumatoid arthritis patients who were NSAID users (15). One caveat is that both the number of patients with rheumatoid arthritis and the number with ulceration was small thus raising the possibility of a type II error; 2 of the 24 patients with rheumatoid arthritis had an ulcer (1 duodenal ulcer and 1 gastric ulcer: blood type 0 and A, respectively). The frequency of blood type 0 and A among the 24 patients with rheumatoid arthritis patients was 11 and 8, respectively. The majority of ulcers in chronic NSAID users was associated with H. pylori infection. It remains unclear whether most NSAID-associated duodenal ulcers are H. pylori-associated ulcers, NSAID-exacerbated H. pylori-associated ulcers, or NSAID-induced ulcers. The known age-related increase in H. pylori infection (11,16) coupled with the relatively older population characteristic of patients with chronic arthritis ensured that, by chance, a large number will both receive NSAIDs and have H. pylori infection. Studies are currently under way to ask whether duodenal or NSAID users do gastric ulcers in H. pylori-infected not recur after eradication of the infection despite continued NSAID therapy. Failure to recur would strongly suggest that the ulcers were at least, in part, related to the H. pylori infection. References 1. Graham DY. Prevention of gastroduodenal injury induced by chronic nonsteroidal antiinflammatory drug therapy. Gastroenterology 1989;96(Suppl):675-681. 2. Peterson WL, Lee EL, Feldman M. Relationship between Compylobacterpylori and gastritis in healthy humans after administration of placebo or indomethacin. Gastroenterology 1988;95: 1185-1197. 3. Graham DY. Compylobacler py1ori and peptic ulcer disease. Gastroenterology 1989;96(Suppl):615-625. 4. Graham DY, Smith JL. Gastroduodenal complications of chronic NSAID therapy. Am J Gastroenterol 1988;83:1081-1084. 5. Iglehart IW III, Edlow DW, Mills L Jr., Morrison SA, Hochberg MC. The presence of Carnpylobacter py!ori in nonsteroidal antiinflammatory drug associated gastritis. J Rheumatol 1989; 16:599-603. 6. Gustavsson S, Phillips SF, Malagelada J-R, Rosenblatt JE. Assessment of Campylobocter-like organisms in the postoperative stomach, iatrogenic gastritis, and chronic gastroduodenal diseases: preliminary observations. Mayo Clin Proc 1987;62: 265-268. Double A, Morris A. Nonsteroidal anti-inflammatory druginduced dyspepsia-is Campylobocterpyloridis implicated? Br JRheumatol 1988;27:110-112. Evans DJ Jr., Evans DG, Graham DY, Klein PD. A sensitive and specific serologic test for detection of Compylobacter pylori infection. Gastroenterology 1989;96:1004-1008, Graham DY, Klein PD, Evans DJ Jr., Evans DG, Alpert LC, Opekun AR, Boutton TW. Campylobacterpylori detected non-
invasively by the 13C-urea breath test. Lancet 1987;1:11741177. Alpert LC, Graham DY, Evans DJ Jr., Yoshimura HH, Hazel1 SL, Evans DG, Klein PD. Diagnostic possibilities for Campylobacter pylori infection. Eur J Gastroenterol Hepatol 1989;1:17-26. Graham DY, Klein PD, Opekun AR, Alpert LC, Klish WJ, Evans DJ, Evans DG, Michaletz PA, Yoshimura HH, Adam E, Boutton TW. Epidemiology of Campylobacterpyloridis infection (abstr). Gastroenterology 1987;92:1411. Lanza FL, Royer GL, Nelson RS. Endoscopic evaluation of the effects of aspirin, buffered aspirin, and enteric-coated aspirin on gastric and duodenal mucosa. N Engl J Med 1980;303:136138. Wood PHN, Harvy-Smith EA, Dixon AJ. Salicylates and gastrointestinal bleeding: acetylsalicylic acid and aspirin derivatives. Br Med J 1962;1:669-675. Larkai EN, Smith JL, Lidsky MD, Graham DY. Gastroduodenal mucosa and dyspeptic symptoms in arthritic patients during chronic nonsteroidal anti-inflammatory drug use. Am J Gastroenter01 1987;82:1153-1158.
15. Semble EL, Turner RA, Wu WC. Clinical and genetic characteristics of upper gastrointestinal disease in rheumatic arthritis. J Rheumatol1987;14:692-699. 16. Graham DY, Adam E, Klein PD, Evans DJ Jr., Evans DG, Hazel1 SL, Alpert LC, Michaletz PA, Yoshimura HH. Epidemiology of Campylobacter pylori. Gastroenterol Clin Biol 1989;13:84B88B.
Received March 13,199O. Accepted November 23,199O. Address requests for reprints to: David Y. Graham, M.D.. Digestive Disease Section (lllD), VA Medical Center, 2002 Holcombe Boulevard, Houston, Texas 77030. This work was supported by the Veterans Affairs Medical Center: grant DK39919 from the National Institute of Diabetes and Digestive and Kidney Diseases; the USDA/ARS Department of Agriculture; McNeil Laboratories; and by the generous support of Hilda Schwartz. The authors thank Dr. Hoda Malaty for assistance with the statistical analyses.