Tobacco Documents Online

C~pl~. - - - STORAGE N°..~~.5........ Input N° .. . ......... GAS'fROENTEROLOGY 1993;105:1098-1103 Nelicobacter pylorl Infection: Independent Risk Indicator of Gastric Adenocarcinoma LARS-ERIK HANSSON,*•II LARS ENGSTRAND,# OLOF NYREN,II DOYLE J. EVANS, Jr.,' ANDERS UNDGREN,s REINHOLD BERGSTROM * BENGT ANDERSSON,** LEIF ATHLIN,$ OVE BENDTSEN,19 and PIOTR TRACZII° Departments of 'Surgery,'Clinical Microbiology, and 'Pathology, and the aCancer Epidemiology Unit, University Hospital, Uppsala, Sweden; 1Veterans Administration Medical Center, Houston, Texas; "Department of Statistics. University of Uppsala, Uppsala, Sweden; * *Department of Surgery, Sala Hospital, Sata, Sweden; ODepartment of Surgery, University Hospital, Ume9, Sweden; "Department of Surgery, Skellefte9 Hospital, Skellefte9, Swedenc and "iDepartment of Surgery, VasterAs Hospital, Vgsterfis, Sweden eackoround: He!lcobacter pylorl has been implicated as a possible etiologic factor in gastric cancer. This case control study was performed to determine the association between H. pylorl and gastric cancer, tak- ing Into account the possibility of confounding by other background factoid Methods: Sera were collected from 112 incident case patients with gastric cancer and 103 control patients with nongastroenterological diseases, who were frequency-matched with respect to age and sex. Immunoglobulin G antibodies to H. py- lorl were Identified using the HM-CAP Immunoassay (Enteric Products Inc., Wesbury, NY) Results: The prev- alence of H. pylorl seropositivity was significantly higher (P a 0.002) among case patients than control patients. The odds ratio (OR) was 2.60 (95% confi- dence interval, 1.35-5.02). The increased OR asso- ciated with H. pylort Infection was confined to tumors with a noncardla location (OR, 3.06) and men (OR, 4.27). OR increased with decreasing age at cancer diagnosis to reach 9.33 in patients <60 years of age. Multivarlate logistic regression analysis was used as control for potential confounding, but the elevated OR assoclated with H. pylor! infection remained signifl- cantly increased. Conclusions: The results support the hypothesis of H. pylor! infection as an Independent risk indicator of gastric cancer. T he microorganism Hrlicobackrpylori is considered a causative agent of chronic antral gastritis.' Be- cause it is assumed that gastritis is a risk factor for cancer of the stomach (at least when it becomes atro- phic), it has been proposed that H. pylori infection may be involved in gastric carcinogenesis.t High preva- ierlce rates of H. pylori infection have been notcd in populations with a high risk of gastric carcinoma.2-4 In two case control studiess•8 and three nested case con- trol studies within cohorts,T-9 H. pylori infection was found to be associated with an increased risk of gastric cancer. However, as the risk of gastric cancer appears to be partially determined by dietary30 and socioeco- nomic" factors, and the latter also seems to be of im- portance for the risk of contracting H. pylori infec- tion,'Z there is a possibility that the previously described association between H. pylori infection and gastric cancer is a result of cohfounding factors. To further examine the question of a possible associ- ation between H. pylori infection and gastric cancer (taking location, and histological type into consider- ation), we determined the seroprevalence of H. pylori immunogiobulin (IgG) antibodies among gastric cancer case patients and hospital control patients. In- formation on dietary and socioeconomic conditions was obtained from all subjects to find out whether confounding by these variables could be responsible for an association. Subjects and Methods Preoperative venous blood sampleswere collected from 114 consecutive patients below the age of 80 years with gastric cancer. The patients received this diagnosis be- tween 1989 and 1991 at eight hospitals in central and north- ern Sweden. The upper age limit was chosen because we expected difficulties in getting reliable answers from those older than 79 years. These cases constituted 60°1fi of all gas- tric cancer occurring in the study base. According to the results in a larger case control study on gastric cancer,13 of which this study was a part, the remaining patients were excluded because of advanced malignant disease or early death (20%), patient's refusal (3%), mental or physical ill- ness (other than gastric cancer) (3%), or because we were unable to get a blood sample (14%). The control group con- sisted of 103 control patients admitted to the same hospitals for nongastrointestinal diseases, namely, inguinal hernia, Abbreviations used In this paper: Cl. confidence Intervai; OR, odds ratio. ~fl 1993 by the American Gastroenterological Association 0016-5085/93/$3.00

I October 1993 fractures, varicose veins, benign thyroid or prostatic disease, or benign or malignant breast disease. The control patients were frequency-matched with respect to 10-year age period, gende ; and hospital of admission. As there was no individ- ual matching, the number of case and control patients was not strictly equal. The case patients were interviewed about their past access (at the time of adolescence and 20 years ago) to a refrigerator and freezer (only 20 years ago) and consumption of vegetables, citrus fruits, and salt. The inter- view also included questions about the consumption of alco- hol and coffee 20 years ago. Further questions concerned the occupation of the respondent's and of their father, geograph- ical place of residence, history of smoking and snuff dip- ping, past medical history, and intake of analgesics and HZ- receptor antagonists. The control patients completed a questionnaire containing the same questions as those posed to the case patients. Socioeconomic status was determined on the basis of the occupational classification as devised by Statistics Sweden" and was based on the occupation of lon- gest duration. The case and control patients were classified into one of three socioeconomic categories. The location of the tumor within the stomach was determined through a combined evaluation of special reports from the clinicians to the investigators, case records, and histopathologic re- ports. The histological specimens of all cancer patients were independently reviewed by an experienced pathologist (A. Lindgren), who was unaware of the results of the other investigation parameters. After confirmation of the diagno- sis of adenocarcinoma, the tumors were classified as pro- posed by Lauren's into the intestinal or diffuse histological type. Two patients were found at this review not to have adcnocarcinoma and were thus excluded. Blood was collected by venepuncture and centrifuged. The serum samples were then stored at -20°C and kept frozen on dry ice during transport to the laboratory. A mi- crobiologist (L. Fngstrand), who did not know the case/ control status of the patient, examined the sera for IgG anti- bodies to H. pylo,i using the 1-1M-CAP immunoassay (I3nteric Products Inc., Westbury, NY). The sensitivity and specificity of this assay has been shown to be 98.7% and 100%, respectively.76 Scra were diluted 1:100 and analyzed according to the instructions of the manufacturer. The study was approved by the Ethical Review Board of the Medical Faculty, Uppsala University, and all case and control patients gave informed consent to interviews. Statistical Methods '!-ests of significance in 2 X 2 tables were performed by the Mantcl-Haens..cl x2 test. For odds ratios (ORs) ob- tained from 2 X 2 tables, test-based confidence limits were constructed using the Cornfield approximation" or an exact computation when appropriate.ts The logistic regression model was used in multivariate analyses of the relationship hctwcen gastric cancer risk and possible risk factors. This tnodcl assumes that the logarithm of the odds of having gastric cancer is a linear function of explanatory variables H. PYLOR! AND GASTRIC CANCER 1099 such as H. py/orl infection, vegetable consumption, etc. By considering transformations of basic variables or catego- rizcd variables, nonlinear relationships are also possible. The model was estimated by the maximum likelihood method. On the basis of the estimated (3 parameters and their standard errors, ORs with confidence intervals (CI) were computed. The OR is an estimation of relative risk. Results There were 112 case and 103 control patients, of whom 71 (63g'o) and 68 (66%), respectively, were men, The mean age of the case and the control patients was 67 years. The prevalence of H. pylori seropositivity- was significantly higher (P = 0.002) in case patients (90 of 112; 80%) than in controls (63 of 103; 61%), giving an OR of 2.60 (95% Cl, 1.35-5.02) (Table 1). A stratified analysis with respect to gender showed a significantly elevated OR in men (OR, 4.27; 95% Cl, 1.75-10.62) but not in women (OR, 1.26; 95% Cl, 0.43-3.71) (Table 1). In a multivariate logistic regres- sion analysis including gender, H. pylori status, and an interaction term for gender and H. pylori status, the OR associated with H. pylori scropositivity and male gender was 4.27 (95% Cl, 1.87-9.74) and the interaction term was 0.30 (95% Cl, 0.08-1.05), which means that the sex difference was not quite statistically significant. The age group <60 years had an OR of 9.33, decreas- ing to 4.27 in the group aged 60-69 years and to 1.15 in the oldest age group (Table 1). A multivariate logis- tic regression analysis including age in continuous form, H. pylorl status, and an interaction term between age and H. pylorr status ((age - 40) X H. pylori status] gave an OR for H. pylori seropositivity (at the age of 40) of 41.5 (95% Cl, 3.78-456), with an interaction term of 0.91 (95t1U CI, 0.84-0.98), showing a significant in- teraction between age and H. pylori status. This means that from the age of 40, OR decreases by 9% annually if a linear effect of age is assumed (OR at 50, 60, and 70 years was 16.2, 6.3, and 2.5, respectively). The de- creasing OR with age was attributable to an increasing prevalence of H. pylori seropositivity with age in the control patients and a decreasing prevalence of this seropositivity with age among the case patients. In a strati6ed analysis with respect to tumor loca- tion within the stomach, only the OR for a noncardia location was found to be significantly increased (3.06) (Table 2). There were eight stump neoplasms (89% with positive H. pylori serology), which were referred to the noncardia group. If these stump neoplasms were excluded from the noncarclia group, there was still a significant association with positive H. pylorr serology (OR, 2.96; 95% Cl, 1.42-6.23).

. 1100 HANSSON ET AL. GASTROENTEROLOGY Vol. 105, No. 4 .a Table 1. Risk of Gastric Adenocarcinoma Associated with H. pylori infection: Stratified Analysis With Respect to Age and Gender Gastric cancer patients Control patients Stratificption group Mean No. age (yr) seropositive (%) Mean age (yr) No. seropositive (%) OR 95% Cl :.• AII sub)ects 67 90/112 (80.4) 67 63/103 (61.2) 2.60 1.35-5.02 Men ` 66 61/71 (85.9) 64 40/68 (58.8) 4.27 1.75-10.62 Women 69 29/41 (70.7) 68 23/35 (65.7) 1.26 0.43-3.71 Age <60 yr - 21/23 (91.3) - 9/17 (52.9) 9.33 1.38-100.73 60-69 yr - 27/33 (81.8) - 20/39 (51.3) 4.27 1.31-15.26 00 yr - 42/56 (75.0) - 34/47 (72.3) ' 1.15 0.44-3.02 Seventy-Bve of the cases (67%) had the intestinal type of adenocarcinoma, 28 (259'0) had the diffuse type, and 9 (8%) had the mixed type (Table 2). An increased OR was noted for the intestinal and diffuse types, but the increase only reached statistical significance for the intestinal type. When the analysis was restricted to noncardia cases, 81% of those with the intestinal type (OR, 2.71; 95% CI, 1.19-6.29) and 85% of those with t6e diffuse type (OR, 3.49; 95% Cl, 1.07-14.84) had positive serology for N. py/ori. Background data on dietary and socioeconomic characteristics showed that the two groups were simi- lar in terms of dietary habits, with the exception of a higher consumption of citrus fruits (OR, 0.31; 95% Cl, 0.15-0.64) among the control patients and a higher consumption of coffee (OR, 4.93; 95% C1, 2.66-9.17) and hard liquor (OR, 2.53; 95% Cl, 0.99-6.61) among the case patients. There was a higher percentage of blue collar workers among the case patients (OR, 1.83; 9546 Cl, 1.00-3.34), whereas farmers and white collar workers dominated slightly among the control pa- tients. A separate analysis aimed at revealing factors asso- ciated with; the risk of acquiring H. py/ori infection was performed within the control group. The mean age of the seropositive and seronegative subjects was 69 and 65 years, respectively. H. pylorr infection tended to be more prevalent among blue collar workers (OR, 1.92; 95% Cl, 0.80-4.66) than nonblue collar workers. Al- though statistical significance was not reached, sero- positive subjects were more likely to have a lower con- sumption of vegetables (OR, 0.48; 95% CI, 0.19-1.17), citrus fruits (OR, 0.49; 95% CI, 0.20-1.21), beer (OR, 0.51; 95% Cl, 0.14-1.86), wine (OR, 0.41; 95% Cl, 0.03-3.80), and liquor (OR, 0.19; 95% Cl, 0.02-1.15). Seropositive subjects tended to have a higher con- sumption of coffee (OR, 1.72; 95% CI, 0.66-4.58) and more had a history of cigarette smoking (OR, 1.53; 95% Cl, 0.64-3.68), peptic ulcers (OR, 3.13; 95% CT, 0.89-13.85), use of H2-receptor antagonists (because none of the control patients who had used H2-receptor antagonists was seronegative, the OR could not be cal- culated), and previous gastric resection (OR, 2.74; 95% CI, 0.26-138.18). The numbers of subjects in the de- scribed groups, however, were small. Multivariate logistic regression analysis was used to examine the risk of gastric cancer associated with H. pylori seropositivity while taking into account possible confounding factors. OR for seropositivity remained virtually unchanged and was signi6cantly,inereased in Table 2. Risk of Gastric Adenocarcinoma Associated With H. pylor! Infection: Stratified Analysis With Respect to Tumor Location and Histological Type Patient group Mean age (yr) No. seropositive (%) OR' 95% Cl All with gastric adenocarcinoma 67 90/112 (80.4) 2.60 1.35-5.02 Site of cancer° Noncardia 68 77/93 (82.8) 3.06 1.49-6.31 Cardia 61 13/19 (68.4) 1.38 0.44-4.77 Histological type° Intestinal 69 60/75 (80.0) 2.54 1.21-5.38 Diffuse 63 22/28 (78.6) 2.33 0.82-7.60 Mixed 68 8/9 (88.9) 5.08 0.63-230.92 'Reiative to control subjects. °The whole control group was used as reference category in these stratified analyses.

October 1993 the model shown in Table 3. The model includes fac- tors suggested to be associated with gastric cancer in this or previous studies. Discussion The main finding in this study was the signifi- cantly increased prevalence of H. pylori scropositivity among patients with noncardia gastric cancer com- pared with control patients. This difl'erence could not be explained by dietary or socioeconomic factors. Our finding of an overall OR of 2.60 is quite consistent with two earlier case control studies with ORs of 2.675 and 2.21 6 An association between H. pylori infection and gastric cancer has also been found in prospective studies, two Americana.9 and one British; with ORs ranging from 2.77 to 6.0. In one of these studies,° the risk increase was confined largely to women (OR, 18.0), in contradiction to the moderately and statisti- cally nonsignificant risk increase in women (OR, 1.26) found in our study. However, the high OR in women in the American studya could at least partly be ex- plained by the fact that only 39.4% of the women con- trol patients were infected, compared with 69.7% of the men. In our study, the percentage of control pa- tients infected was similar in men and women. This is in line with the finding in a number of studies in both developed and developing countries that the age-spe- ciFic prevalence of infection was the same in both genders.19-20 In the analysis with stratification by age, OR was significantly increased in only the two younger age groups. The reason for the absence of an excess risk arnong the oldest cases was twofold. Firstly, the preva- lence of H. pylori seropositivity increased with age in the control patients. As the great majority of the popu- lation will ultimately become infected, any differences in infection rates'will tend to disappear in the oldest Table 3. OR of Gastric Cancer and 95% Cl in a Multiple Logistic Regression Model OR Cl H. pylorl seropositlvity 2.73 1.29-5.79 White collar worker' 0.87 0.38-2.00 Farmer• 0.24 0.07-0.76 Access to refrigerator 20 yrs ago 0.26 0.05-1.25 High vegetable consumption 20 yrs ago 1.65 0.80-3.4 t High citrus fruit consumption 20 yrs ago 0.22 0.09-0.54 Coffee consumption 1-3 cups/day° 2.12 0.59-7.58 Coffee consumption >3 cups/day' 10.50 2.92-37.70 High liquor, consumption 4.45 1.42-13.96 History of cigarette smoking 1.86 0.90-3.38 'Relerence category, blue collar worker. °Reference category, coffee consumption < t cup/day. H. PYLORI AND GASTRIC CANCER 1101 age groups when the rates among the control patients catch up with those in the case patients. Secondly, the percentage of infected cancer case patients decreased in the oldest age groups. This was not unexpected, as the unfavorable conditions for H. pylod survival asso- ciated with the.precancerous atrophic gastritist•21 may cause the organism to disappear in some cases. Our finding is in accordance with previous results of case control and cohort studies.6•9 Thus, it seems as if early acquisition of the infection increases the risk; this is quite reasonable considering the probably long induc- tion time for gastric cancer via gastritis, atrophy, and metaplasia. In patients with carcinoma located at the gastric cardia, we found no significant association with H. pylori infection in accordance with previous epidemio- logicals," and clinical22 studies. There are several epi- demiological features that distinguish cancer of the gastric cardia from more distal gastric cancer,2s and the difference in the association of these two cancer sites with H. pylorr infection may be another indication of different etiologies. It has been hypothesized (mainly on the basis of the pattern of occurrence) that the pathogenesis of the in- testinal type of gastric carcinoma may be linked to environmental factors, whereas the diffuse type may have other causes, possibly genctic. As chronic atro- phic gastritis (type B) has mainly been linked to the intestinal type of gastric carcinoma,2'.2s H. pylorr infec- tion would be expected to show the same relationship. However, we found that H. pylod infection was asso- ciated with an increased risk of both the intestinal and the diffuse types of gastric carcinoma. The same obser- vation was made in other seroepidemiological stud- ies s.a.s,9 In one22 of two22~16 studies based on histologi- cal diagnosis of H. pylori infection, a significantly higher prevalence of this infection was noted among patients with the intestinal type compared with the diffuse type of gastric carcinoma. Nevertheless, most data seem to indicate that there is no important differ- ence between the two histological types ofgastric carci- noma as far as their association with H. pylod is con- cerned. Could the increased prevalence of H. pylod among the case patients be caused by increased susceptibility of the cancerous stomach to become infected? Because the blood samples were taken after the cancer diagno- sis, this question would seem justified. Our finding that OR was highest in the youngest age group and the fact that precancerous conditions, such as atrophy and intestinal mctaplasia, are unfavorable for H. pylod col- onization contradict this possibility. Another potential

1102 HANSSON ET AL. source of error in this study may be that the control group of hospitalized patients was not representative of the study base. Our finding of a 61% prevalence of H. pylori'seropositivity, howevcr, corresponds well with rcports from other Western populations with the same age distribution.12,2'als Another source of concern is the fact that the case patients underwent face-to-face interviews about their diet and other hacfcground charactcristics as part of a large epidemio- logical project on gastric cancer, whereas the control patients, for cost-saving reasons, were given self-ad- ministered questionnaires. This may potentially result in differential misclassification. Hence, the varying distributions of the variables between the case and control patients, indicating some of them as possible risk factors of gastric carcinoma, should be interpreted with caution. Both H. pylori infection'•27•2° and gastric cancertt are more prevalent in populations with poor socioeco- nomic conditions; this may be a potential confounding factor. However, in our study, the OR associated with H. pylori infection remained significantly increased in multivariate logistic regression models, making con- founding by these variables less likely. Hence, H. pylori seems to be an independent risk indicator of gastric carcinoma. f"l'here arc several possible mechanisms by which H. pylorr infection may be involved in gastric carcinogene- sis.1-1. pylori adversely affects the chemical and physical properties of the mucus laycr,"' which may make the mucosa susceptible to carcinogenic factors. Acute in- fection by 1-1. pylori has been reported to result in pro- longed inhibition of the naturally occurring gastric se- cretion of ascorbic acicl.'t This secretion also seems to be decreased in chronic gastritis, thereby permitting a greater degree of intragastric formation of N-nitroso compounds. Moreover, H. pylori seems to act as a pro- moter in the progression from normal to metaplastic epithelium, possibly by inclucing a hyperprolifcrative state in the inflamed gastric mucosa.° Finally, the chronic inAammation per se may expose the mucosal cells to the oxidativc stress of free radicals generated from the inf-ammatory cells12 There are several issues to be resolved, however, before the association between H. pylorr and gastric cancer can be accepted as causal. There are several populations, e.g., in ChinaZ and Africa,'9 where H. py- lori infection is prevalent at young ages but gastric cancer is uncommon. There is a lack of specificity, as H. pylori infection seems to be a risk indicator of both intestinal and diffuse types of adenocarcinoma,3'9 as well as gastric lymphomas.s•" If, as several studies sug- GASTROENTEROLOGY Vol. 105, No. 4 gest,'7," H. pylori is an important causal factor in duo- denal ulccr, the negative association between duodenal ulcer and later development of gastric cancer's is per- plexing. Thus, the role of H. pylori in the complex mul- tifactorial carcinogenesis of gastric cancer needs to be explored further. References 1. Dixon MF. Campylobacter pylorl and chronic gastritis. In: Rath- bone BJ, Heatley RV, eds. Campylobacter pylori and gastroduo- denal disease. Oxford, England: Blackwell Scientific, 1989; ! 06- 116. 2. Forman D, Sitas F. Newell DG, Stacey AR, Boreham J, Peto R, Campbell TC, U J, Chen J. Geographic association of Hetico- bacter pylort antibody prevalence and gastric cancer mortality in rural China. Int J Cancer 1990;46:608-611. 3. Burstein M, Mouge E, Le6n-Barfa R. Lozano R, Berendson R, Gilman RH, Legua H, Rodriguez C. Low peptic ulcer and high gastric cancer prevalence In a developing country with a high prevalence of infection by Nellcobacter pylori. J Clin Gastroen- terol 1991;13:154-156. 4. Correa P, Fox J, Fontham E, Ruiz B, Lin Y, Zavala D. Taylor N, Mackinley D, de Lima E, Portilla H, Zarama G. Nellcobacter pylori and gastric carcinoma. Serum antibody prevalence in poputa- tions with contrasting cancer risks. Cancer 1990;66:2569- 2574. 5. Talley NJ, Zinsmeister AR, Weaver A, DiMagno EP, Carpenter HA, Perez-Perez G1, Blaser MJ. Gastric carcinoma and Nelicobacter pylorf infection.l Nati Cancer Inst 1991;83:1734-1739. 6. Sipponen P, Kosunen TU, Valle T, Riihela M, Seppala K. Netico- bacter pyforf infection and chronic gastritis in gastric cancer. J Clin Pathol 1992;45:319-323. 7. Forman D, Newell DG, Fullerton F, Yarnell JWG, Stacey AR, Wald N, Sitas F. Association between Infection with Helicobacter py- forl and risk of gastric cancer: evidence from a prospective inves- tigation. BMJ 1991:302:1302-1305. 8. Parsonnet J. Friedman GD, Vandersteen DP, Chang Y, Vogelman IN, Orentreich N, Sibley RK. Helfcobacter pylori infection and the risk of gastric carcinoma. N Eng- J Med 1991:325:1127-1131. 9. Nomura A, Stemmermann GN, Chyou PH. Kato I, Perez-Perez GI, Blaser M1. Helicobacter pytorl Infection and gastric carcinoma among Japanese Americans In Hawaii. N Engl J Med 1991;325:1132-1136. 10. Boeing H. Epidemiological research In stomach cancer: progress over the last ten years. J Cancer Res Clin Oncol 1991;117:133- 143. 11. Howson CP, Hlyama T, Wynder EL. The decline in gastric cancer: epidemiology of an unplanned triumph. Epidemtol Rev 1986; 8:1-27. 12. Taylor ON, Blaser MJ. Epidemiology of Hellcobacter pylorl infec- tions. Epidemlol Rev 199 1; 13:42-59. 13. Hansson LE, Nyran 0, Bergstrtlm R, Wolk A, Lindgren A, Baron J. Adaml HO. Diet and risk of gastric cancer. A population-based case-control study In Sweden. Int J Cancer (in press). 14. Statistics Sweden. Swedish socioeconomic classification. Re- ports on statistical coordination. Stockholm: Statistics Sweden. 1982:4. 15. LaurAn P. The two histological main types of gastric carcinoma. diffuse and so-called intestinal-type carcinoma: an attempt at a histociinicai ctas ircati APMtS 1965 64•31 49 s i n . 16. Evans DJ Jr, Evans DG, Graham DY, Klein PD. A sensitive and specific serologic test for detection of Campylobacfer pylor( in- fection. Gastroenterology 1989;96:1004-1008. 17. Fleiss iL. Statistical methods for rates and proportions. New York: WIley, 1981. QD -- GO Cr Pz

... I Octotier 1993 18. Mehta CR, Patel,NR, Gray R. Computing an exact confidence Intervat for the common odds ratio In several 2 X 2 contingency tables.l Am Statistical Assoc 1985;80:969-973. 19. MEgraud F, Brassens-Rabb6 M-P, Denis F. Belbouri A, Hoa DQ. Seroepidemlology of Campylobacter pylori infection in various populations.l Clin Microblol 1989;27:1870-1873. 20. Graham DY, Maiaty HM, Evans DG, Evans DG Jr, Klein PD, Adam E. Epidemiology of Welloobacter pylorl in an asymptomatic popu- lation In the United States. Effect of age. race, and socioeco- nomic status. Gastroenterology 1991;100:1495-1501. 21. Siuraia M, Sipponen P. Kekki M. Campylobacter pylorl in a sam- ple of Finnish population; relations to morphology and functions of the gastric mucosa. Gut 1988;29:909-915. 22. Parsonnet J, Vandersteen D, Goates J. Sibley RK, Pritikin J. Chang Y. Helicobacter pylori infection In intestinal and diffuse- type gastric adenocarcinomas. J Nati Cancer inst 1991;83:640- 643. 23. Hansson L-E. SparE:n P. Nyr6n 0. Increasing Incidence of carci- noma of the gastric cardia in Sweden 1970-1985. Br J Surgery 1993;80:374-377. 24. Correa P. Cuello C, Duque E. Burbano I.C. Garcia FT. Bolaftos 0. Gastric cancer In Colombia. Ill. Natural history of precursor le- sions. J Nati Cancer Inst 1976;47:1021-1026. 25. Correa P, Haenszel W, Cuello C. Zavala D, Fontham E, Zarama G. Gastrtc precancerous process in a high risk population: cohort foliow•up. Cancer Res 1990;50:4737-4740. 26. Loffeld RJLF, Willems i, Flendrig JA, Arends JW. Helicobacter py- torJ and gastric carcinoma. Histopathology 1990;17:537-541. 27. Sitas F, Forman D. Yarnell JWG, Btur ML. Elwood PC, Pedley S. Marks KJ. Hellcobacter pylori Infection rates in reiation to age and social class in a population of Welsh men. Gut 1991;32:25- 28. 28. Kosunen TU, Ht1t1k J. Rauleiin HJ, Myliyla G. Age-dependent in- crease of Campylobacter pylorl antibodies In blood donors. Scand J Gastroenterol 1989;24:110-114. 29. The Gastrointestinal Physiology Working Group. Hellcobacter py- H. PYLORI AND GASTRIC CANCER 1103 lorl and gastritis in Peruvian patients: relationship to socioeco- nomic level, age, and sex. Am J Gastroenterol 1990;85:819- 823. 30. Sarosick J, Peura DA, Guerrant PL, Marshall B1, Laszewicz W, Gabryelewicz A, McCallum RW. Mucolytic effects of Helicobacter pylorf. Scand J Gastroenterol 199 1; 187:87:47-55. 31. Sobala GM, Crabtree JE, Dixon MF, Schorah CJ, TaylorJD, Rath- bone BJ, Heatley RV, Axon ATR. Acute Hellcobacter pylorJ infec- tions clinical features, local and systemic immune response, gas- tric mucosal histplogy, and gastric juice ascorbic acid concentrations. Gut l 991;32:14 ! 5-1418. 32. Ames BN, Motchnik P, Profet M. Shigenaga MK, Hagen TM. An- tioxidant prevention of birth defects and cancer. Presented at the international conference on male-mediated developmental toxic- ity; September 16-19, 1992; Pittsburgh, Pennsylvania. New York: Plenum (in press) 33. Wotherspoon AC, Ortiz-Hidalgo C, Falzon MR, Isaacson PG. Hell- cobacter pylori-associated gastritis and primary B-cell gastric lymphoma. Lancet i991;338:1175-1176. 34. Blaser MJ, Perez-Perez Cl, Lindenbaum J, Schneidman D, Van Deventer' G, Marin-Sorensen M, Weinstein WM. Association of infection due to Helicobacter pylorl with specific upper gastroin- testinal pathology. Rev Infect Dis 1991;8(suppi 13):S704- S708. 35. Lee S, lida M, Yao T. Schindo S, Nose K. Akazawa K. Okabe H, Fujishima M. Risk of gastric cancer in patients with non-surgically treated peptic ulcer. Scand J Gastroenterol 1990;25:1223- i226. Received March 3, 1993. Accepted June 3, 1993. Address requests for reprints to: Lars-Erik Hansson, M.D., Depart- ment of Surgery, University Hospital, S-751 85 llppsala, Sweden. Supported by grants from the Swedish Society of Medicine, Pro-' fessor Nanna Svartz Foundatlon, and Tore Nllsson Foundation. The authors thank Ulla Zettersten for excellent technicai assis- tance.