Gastrin and colorectal cancer: A prospective study

GASTROENTEROLOGY 1998;115:275–280

Gastrin and Colorectal Cancer: A Prospective Study CHRISTINE M. THORBURN,* GARY D. FRIEDMAN,‡ CHRIS J. DICKINSON,§ JOSEPH H. VOGELMAN,\ NORMAN ORENTREICH,\ and JULIE PARSONNET*,# Departments of *Medicine and #Health Research and Policy, Stanford University School of Medicine, Stanford, California; ‡Division of Research, Kaiser Permanente Medical Care Program, Oakland, California; §Department of Pediatrics, University of Michigan Medical Center, Ann Arbor, Michigan; and \Orentreich Foundation for the Advancement of Science, Cold Spring on Hudson, New York

Background & Aims: Gastrin is a putative promoter of colorectal carcinomas. The aim of this study was to evaluate the temporal relationship between gastrinemia and development of colorectal malignancy. Methods: We conducted a nested case-control study among 128,992 subscribers to a health maintenance program who had participated in a multiphasic health checkup between 1964 and 1969. Serum had been frozen since the checkup and the cohort followed up for cancer. Of 1881 incident colorectal carcinoma cases, 250 were randomly selected; 1 control without cancer was matched to each case by age, sex, education, and date of serum collection. Stored sera were tested for Helicobacter pylori immunoglobulin G and for gastrin and glycine-extended gastrin. Results: Verified cases included 166 colon cancers, 58 rectal cancers, and 9 with cancer in both locations. A mean of 15.3 years had elapsed between serum collection and diagnosis of cancer. Median gastrin levels were similar in cases and controls (41.7 vs. 40.7 pg/mL). However, a gastrin level above normal was associated with increased risk for colorectal malignancy (odds ratio, 3.9; 95% confidence interval, 1.5–9.8). If this association is causal, 8.6% of colorectal cancers could be attributed to high serum gastrin level. Conclusions: Hypergastrinemia is associated with an increased risk of colorectal carcinoma.

n the United States, colorectal carcinoma is one of the most common cancers and is the second leading cause of cancer-related mortality.1 Multiple environmental and hereditary risk factors contribute to its etiology. Together, these cause initiation (irreversible genetic alteration), promotion (reversible induction of clonal expansion), and progression from benign to malignant growth.2 For colorectal cancer, one posited promoter that has been studied extensively is gastrin. Although gastrin’s primary action is to stimulate acid secretion, some data indicate that it also stimulates growth of normal colonic epithelium.3–5 Yet, this remains controversial.6,7 Early studies showing colonic epithelial proliferation used pentagas-

I

trin, which also has affinity for cholecystokinin receptors4,8; thus, the proliferative effects may not be specifically caused by gastrin. Moreover, while synthetic gastrin stimulates protein, RNA, and DNA synthesis in colonic cell cultures,9 in vivo studies in rodents are more inconsistent. Chu et al.5 measured increased colonic crypt size and higher percentage of goblet cells in hamsters with hypergastrinemia induced by gastric fundectomy, but Oscarson et al.10 measured no increased mucosal weight or thickness. Data are even more discrepant for human beings.11 Patients with hypergastrinemia caused by Zollinger– Ellison syndrome have comparatively higher than normal indices of colonic proliferation.12 However, patients with Zollinger–Ellison syndrome have not been found to have an increased risk for colorectal cancer.13 Some casecontrol studies have found higher fasting plasma gastrin levels in patients with colorectal carcinoma than in controls, but other studies have not.14–19 Declines in gastrin levels after tumor resection raise the additional concern that the association between gastrin and colon cancer reflects autocrine production of gastrin by the tumor.15,16,19 Some investigators hypothesize that the conflicting results observed in human studies can be explained by differences in the forms of gastrin being detected.20 The fully processed form, amidated G-17 (which for sake of simplicity will be called ‘‘gastrin’’), is the form typically detected by commercial assays. Yet, unprocessed forms of gastrin can also cause cell proliferation. In one study, glycine-extended gastrin (G-Gly) caused equal proliferation of colon tumor cells as gastrin, although by different receptor mechanisms.21 Colon tumors and normal colon cells produce G-Gly in excess of the fully processed Abbreviations used in this paper: CI, confidence interval; G-Gly, glycine-extended gastrin; KPMCP, Kaiser Permanente Medical Care Program; MHC, multiphasic health checkup; NCCC, Northern California Cancer Center; OR, odds ratio; PAR, population attributable risk percent. r 1998 by the American Gastroenterological Association 0016-5085/98/$3.00

276

THORBURN ET AL.

gastrin, providing support for the autocrine hypothesis of the gastrin-colon cancer relationship.22,23 Some of the controversy surrounding gastrin and colorectal cancer could be resolved by evaluating gastrin levels in subjects before development of cancer. This minimizes potentially biasing alterations in gastrin levels caused by the presence of tumor or the initiation of cancer treatment in the case. To this end, we conducted a nested case-control study of colorectal cancer and gastrin within the Kaiser Permanente Multiphasic Health Checkup cohort. This is, to our knowledge, the only populationbased, longitudinal study to evaluate this putative relationship.

Materials and Methods Study Population The cohort population consisted of 128,992 adult subscribers to the Northern California Kaiser Permanente Medical Care Program (KPMCP) who had participated in a multiphasic health checkup (MHC) program between 1964 and 1969.24 Since their first multiphasic health examination, members of the cohort have been followed up for cancer outcomes. From enrollment through 1991, 1881 incident colorectal adenocarcinoma cases had occurred among the population. Of these, 250 were randomly selected as case patients. For each case, 1 control subject was matched by 2-year birth cohort, gender, education, site (Oakland or San Francisco), and date of serum collection (same month and year). Sample size calculations indicated that 250 pairs would provide sufficient power to detect a 23-pg/mL difference in serum gastrin between cases and controls. As part of the Surveillance, Epidemiology and End Results Program, pathology reports for most cases identified at KPMCP are reviewed by the Northern California Cancer Center (NCCC) and the diagnoses confirmed. Charts and pathology records from cases that had not been confirmed by NCCC were reviewed in detail, and those not confirmed were excluded.

Serological Assays MHC enrollees between 1964 and 1969 underwent glucose tolerance testing as part of the examination. Subjects were asked about their last food intake (95% of subjects had last eaten 3 hours or more before glucose ingestion; median, 5 hours) and then received 75 g of glucose. Serum samples were drawn 1 hour after the glucose bolus. Aliquots of these sera have been catalogued and maintained, for the initial 10 years at 223°C by KCMCP and subsequently at 240°C by the Orentreich Foundation for the Advancement of Science, Inc.25 Sera were tested for gastrin levels using a commercial double antibody 125I-radioimmunoassay (Diagnostic Products Corp., Los Angeles, CA). This assay is highly specific for low– molecular-weight gastrin with 100% detection of both little gastrin (G17) and for mini-gastrin; it also has 39% crossreactivity for big gastrin (G34). The assay has little affinity for

GASTROENTEROLOGY Vol. 115, No. 2

the other gastrointestinal hormones in this family, including cholecystokinin (11%–18% cross-reactivity). The assay has excellent precision with intra- and interassay coefficients of variation of 5%–7%. The upper reference limit determined by this assay is 90 pg/mL for fasting samples from healthy adults.26 A subset of 50 matched serum pairs were tested for G-gly. Included in these 50 serum pairs were all 34 pairs in which a high gastrin level (.90 pg/mL) had been observed and a random selection of 16 other serum pairs. G-gly was detected by radioimmunoassay using antibody 8237, an antibody specific to carboxy-terminal glycine extensions of gastrin, as previously described.22,27 This assay cross-reacts with less than 1% of the amidated, final forms of the peptide. Serum samples were tested by enzyme-linked immunosorbent assay for Helicobacter pylori immunoglobulin (IgG) as previously described.28 All samples were run in triplicate. For all serological analyses, paired samples were tested simultaneously without knowledge of the case-control status of the subjects within the pair.

Statistical Analyses Data were entered and analyzed with EpiInfo (Centers for Disease Control and Prevention, Atlanta, GA) and Egret (Statistics and Epidemiology Research Corp., Seattle, WA) computer programs. Gastrin levels were not normally distributed and were not normalizable by usual methods of transformation. Therefore, gastrin was analyzed nonparametrically by quartiles and as a dichotomous variable using the upper reference limit of the gastrin assay (90 pg/mL, the 95th percentile for population serum gastrin) as the cutoff. Quartile strata determined using the 233 controls were: ,32.0 pg/mL, 32.0 pg/mL to ,40.1 pg/mL, 40.1 pg/mL to ,51.9 pg/mL, and $51.9 pg/mL. Subsequent post hoc analyses to evaluate the robustness of our primary findings included: a matchedrank test, evaluation by tertiles and evaluation of different dichotomous break points (i.e., the fifth percentile for serum gastrin in controls and at each of the quartile cutoff values). Quartile strata for G-gly were ,185 fmol/mL, 185 fmol/mL to ,282 fmol/mL, 282 fmol/mL to ,450 fmol/mL, and $450 fmol/mL. For univariate analysis of categorical variables, odds ratios (ORs) were determined as the ratio of discordant pairs and tested for statistical significance by the McNemar method. The Wilcoxon’s signed rank test was used to compare gastrin values in the matched cases and controls. For multivariate analysis, conditional logistic regression was used. Variables included in the analyses were as follows: gastrin, H. pylori infection, hours fasted before serum draw (#4 vs. .4 hours), race (white, black, or Asian), cigarette smoking at the time of MHC enrollment (current smoker, former smoker, or never smoked), and body mass index at the time of MHC enrollment. Obesity was defined as a body mass index .27.5 kg/m2. Statistical tests of the regression estimates were based on the x2 approximation for the likelihood ratio statistic with confidence intervals (CIs) determined by Wald’s test. Interaction terms were included to examine the joint association of risk factors on the odds of

August 1998

GASTRIN AND COLORECTAL CANCER

277

developing colorectal cancer. Population attributable risk percent (PAR) was calculated using the equation: PAR 5

p 3 (OR 2 1) [p 3 (OR 2 1) 1 1]

,

where p equals the prevalence of high gastrin concentrations in the control population.

Results Of the original 250 patients with colorectal cancer selected from the Kaiser cohort, cancer was verified in 233. These cases and their matched controls were included in the analyses. Case and control subjects were similar with respect to sex, race, and age at time of serum donation (Table 1). Ninety-five percent had last eaten 3 or more hours before their glucose bolus (median, 5 hours). Because gastrin levels return to the daytime baseline approximately 3 hours after eating a meal, the vast majority of subjects were at their daytime baseline before administration of glucose.24 H. pylori antibodies were similarly common in cases (68.2%) and controls (67.8%) (Table 1). The site of primary tumor in cases was 71% colon, 25% rectum, and 4% with sites in both the colon and the rectum. The mean age at diagnosis of the case was 67.9 years. The average interval between serum donation and diagnosis of cancer was 15.3 years (Table 1). Gastrin levels in all subjects ranged from 12 to 1039 pg/mL and were not normally distributed (Figure 1). However, control subjects had a distribution of gastrin levels expected of a normal population, i.e., more than 95% had gastrin levels ,90 pg/mL, the assay manufacturer’s cutoff for normal. A significantly greater proportion of cases had gastrin levels above the normal threshold, with 25 cases (10.7%) but only 9 controls (3.9%) above Table 1. Characteristics of Case and Control Subjects Characteristic Mean age at enrollment ( yr ) Male, n (% ) Race, n (% ) White Black Asian Other Mean hours last meal to glucose bolus H. pylori infected, n (% ) Mean age at diagnosis ( yr ) Mean interval to cancer ( yr ) Site of primary tumor, n (% ) Colon Rectum Both

Cases (N 5 233)

Controls (N 5 233)

52.7 113 (48.5)

52.6 113 (48.5)

191 (82.0) 29 (12.4) 8 (3.4) 5 (2.1) 5.5 159 (68.2) 67.9 15.3

199 (85.4) 19 (8.2) 8 (3.4) 7 (3.0) 5.4 158 (67.8)

166 (71.2) 58 (24.9) 9 (3.9)

Figure 1. Distributions of gastrin levels in colorectal cancer cases and controls. Boxes represent the interquartile distance (25th to 75th percentile of values) with the intersecting line indicating the median value. The asterisk denotes the geometric mean gastrin value, and the ascending and descending bars reflect the total range observed. Median and geometric mean gastrin values were similar for cases and for controls.

this level (P , 0.005). This difference was not explained by meal stimulation of gastrin release because subjects with high gastrin levels had similar mean hours since their last meal (6.1 hours) as did the normal gastrin group (6.5 hours). None of the persons with high gastrin levels had eaten within 2 hours of glucose bolus. Median gastrin levels were similar in cases and controls (41.7 and 40.7 pg/mL, respectively; Figure 1), and no trend was observed between increasing gastrin quartile and malignancy. Similarly, when gastrin levels in case-control pairs were compared using the Wilcoxon’s signed rank test, gastrin was not significantly associated with cancer (P 5 0.35). In multivariate analysis, gastrin levels above the normal range were significantly associated with an increased risk for colorectal cancer (OR, 3.9; 95% CI, 1.5–9.8; Table 2 ). This association was more pronounced for rectal tumors (55 evaluable pairs: OR, 10.2; 95% CI, 1.2–88.0) than for colon tumors (139 evaluable pairs: OR, 2.5; 95% CI, 0.9–7.2). No other dichotomous cutoffs for gastrin (i.e., at quartile values or at the lowest 5%) yielded a significant association between gastrin and

278

THORBURN ET AL.

GASTROENTEROLOGY Vol. 115, No. 2

Table 2. Multivariate Analysis of the Association Between High Gastrin Levels and Colorectal Carcinoma Variable Serum gastrin .90 pg/mL Infection with H. pylori History of cigarette smoking b Hours fasted before glucose bolus .4 h #4 h

No. of cases

No. of controls

Adjusted OR a

95% CI

25 159 121

9 158 102

3.9 0.9 1.5

1.5–9.7 0.5–1.5 1.0–2.4

172 61

176 57

1.0 1.2

0.8–2.0

aAdjusted

odds ratios were calculated by conditional logisticregression analysis. Also adjusted for race. Subjects were matched on age and sex. bDefined as current or former smoker; 12 cases and 13 controls had unknown smoking history.

colorectal cancer. Overall, the proportion of colorectal tumors attributable to gastrin levels above the normal range was 8.6%. Current and former cigarette smokers were also at increased risk for colorectal cancer (OR, 1.5; 95% CI, 1.0–2.4). This association appeared to be particularly strong in the subgroup with rectal tumors (55 evaluable pairs: OR, 2.8; 95% CI, 1.2–6.4) and not in the subgroup with colon tumors (139 evaluable pairs: OR, 1.2; 95% CI, 0.7–2.1) H. pylori infection status was not independently associated with colorectal cancer, but infection was linked with high gastrin levels. Among cases and controls with gastrin levels above the normal range, 92% (31 of 34 subjects) were infected with H. pylori, whereas only 66% (286 of 432) of remaining subjects were infected (OR, 5.3; 95% CI, 1.6–27.3). To exclude the possibility that high gastrin level indicated autocrine production by an early tumor, multivariate analysis was repeated excluding cases who developed cancer within 5 years of serum donation. The OR for these 185 pairs was still significant at 4.3 (95% CI, 1.6–11.7). Similar results were obtained using the 158 cases who developed cancer more than 10 years after serum draw (OR, 6.2; 95% CI, 1.7–21.8). G-gly quartiles inversely correlated with gastrin quartiles (P 5 0.01, x2 test for trend). G-gly was not associated with colon cancer in univariate analysis (P 5 0.2), although those subjects in the middle two quartiles appeared to have a somewhat increased risk for cancer compared with the highest and lowest quartiles. When we evaluated the proportion of G-gly to gastrin, we observed a strong negative association with cancer; subjects with the lowest proportion of G-gly to gastrin had a risk of cancer 11-fold higher than those with the highest proportion (P 5 0.006; P value for overall trend in quartiles 5 0.06). Because this post hoc analysis was conducted on sera already known to have high gastrin

levels, these findings may reflect selection bias. However, we can conclude that the higher risk of cancer observed with high gastrin levels cannot be explained by higher G-gly levels. After data analysis was complete, we reviewed the medical charts of 30 subjects with high gastrin levels to determine if we could identify the source of gastrin (records of 4 of 34 total subjects [2 cases and 2 controls] were not available for review). None of the 30 subjects had been diagnosed with pernicious anemia or Zollinger– Ellison syndrome. Two of the cases had markedly increased mean red cell volumes, but no B12 measurements were taken to confirm pernicious anemia. At chart review, we found that 1 of 7 controls with high gastrin had metastatic colon cancer diagnosed at postmortem examination but after the diagnosis of colon cancer in the matched case. When this pair was dropped from the analysis, the OR for the association between high gastrin level and colon cancer remained unchanged.

Discussion In this prospective study of healthy adults, serum gastrin levels above the normal range (.90 pg/mL) were associated with a 3.9-fold increased risk of colorectal carcinoma. H. pylori infection, race, hours since last meal, and obesity did not independently alter this risk, although high gastrin levels and H. pylori infection were strongly correlated. Importantly, gastrin was not a risk factor for colorectal cancer when evaluated using nonparametric ranking methods. Within the normal range of serum gastrin, there was no evidence for a dose-response effect. The relationship of tumors to only abnormally high gastrin levels, however, coheres with laboratory studies that show increased cell proliferation only at the highest concentrations of gastrin.29 Because few persons exhibit such high gastrin levels, only a small proportion of colon tumors could be expected to be linked to this factor. Indeed, the proportion of colon tumors attributable to high gastrin, if the association is causal, was only 8.6%. Although this study provides temporal and statistical evidence that high gastrin levels precede development of tumors, one cannot conclude that the association is causal. A high gastrin level could itself be a marker for other conditions that predispose to development of malignancy. One particular analytic concern was that increased gastrin reflected autocrine production of the hormone by preexisting tumors or polyps. Because the relative risk for cancer in persons with high gastrin levels actually increased when early-onset cases were excluded, this explanation is unlikely. Moreover, subjects with

August 1998

hypergastrinemia were statistically more likely than persons with normal gastrin levels to have H. pylori infection, suggesting a nonautocrine, gastric source for the hormone. The rationale for exploring an association between gastrin and colorectal cancer is the putative role of the hormone in cell growth. Yet there remains considerable controversy about this hypothesis. Some well-conducted studies neither confirm the presence of gastrin receptors on normal colonic tissue nor show increased cell proliferation in response to gastrin stimulation.11 Interestingly, we observed gastrin to be somewhat more closely associated with rectal cancer (OR, 10.2) than colon cancers (OR, 2.5). Although this difference was not statistically significant (P 5 0.18), it conforms with animal experiments showing proliferative responses to gastrin only in the distal colon.30 An alternative explanation for our observation is that hypergastrinemia signals the presence of atrophic gastritis and hypochlorhydria. Decreased acid in the stomach would allow colonic colonization with acid-sensitive microbial flora. This, in turn, has been postulated to promote tumor growth.31 If this were the case, high gastrin levels would not be directly in the carcinogenic pathway, but nonetheless would signal a risk from sustained hypochlorhydria. Gastrin levels are commonly elevated by the use of acid-inhibitory medications. None of the subjects in this study could have been taking histamine blockers or proton pump inhibitors because serum was drawn before these drugs were marketed. Yet, because these now commonly used medications can cause a 2–4-fold increase in gastrin levels, any link between hypergastrinemia and tumor growth is a potential cause for concern.32–34 This particularly merits consideration now that several of these medications are available over the counter. Several large, longitudinal studies do not implicate acid-inhibiting drugs in colorectal carcinogenesis.35–37 Moreover, in vitro, at least one proton pump inhibitor is thought to decrease colonic tumor cell growth.29,38 It remains possible that the high gastrin levels in our cancer group reflect an underlying predisposition to cancer independent of acid conditions in the stomach. Because selfadministered acid-inhibitory medications are now common and because the effects of gastrin on tumor growth may be restricted to a small subset of subjects with a marked hypergastrinemic response, it is important to pursue more longitudinal studies that evaluate the connections among gastrin, H. pylori, and acid-inhibitory therapies. Several potential weaknesses of this study must be considered. Because serum gastrin values were evaluable from only one point in time, these values may not be

GASTRIN AND COLORECTAL CANCER

279

representative of their long-term gastrin levels. Although it is clear that serum gastrin levels fluctuate on a daily basis, these variations are unlikely to explain the extremely high gastrin levels observed. Studies suggest that a single reading of a high gastrin is likely to reflect a sustained process.39 An additional issue in this study may be the validity of using serum samples that had been frozen for more than 25 years. Although the serum values conform to the expected distribution of gastrin in the U.S. population, values presented in this article should not be taken as precise. Because cases and controls were matched on date of serum donation, however, the internal validity of this study should not be impaired by duration of freezing. One unanticipated finding was the association between cigarette smoking and later development of colorectal carcinomas. This upholds other studies that indicate a role for cigarettes in colorectal adenomas and cancers.40–43 However, as with gastrin levels, the proportion of cancers attributable to cigarettes was small. This study supports the hypothesis that, for a subset of colorectal cancer patients, hypergastrinemia may play a role in the development of their tumors. Among our cohort, high gastrin levels accounted for only 8.6% of the colorectal cancer cases. Because the proportion of disease attributable to a risk factor is dependent on the factor’s prevalence, however, the percentage of tumors related to high gastrin can be expected to increase if hypergastrinemia becomes more prevalent. Moreover, given the high prevalence of colorectal cancer in the industrialized world, even a small proportion of cases represents a substantial number of individual lives. We believe that given the increasing incidence of colorectal cancers worldwide, our finding warrants still further investigation of the relationships among gastrin, H. pylori infection, acid-inhibitory medications, and colorectal carcinoma.

References 1. Potter JD, Slattery ML, Bostick RM, Gapstur SM. Colon cancer: a review of the epidemiology. Epidemiol Rev 1993;15:499–545. 2. Boone CW, Kelloff GJ, Freedman LS. Intraepithelial and postinvasive neoplasia as a stochastic continuum of clonal evolution, and its relationship to mechanisms of chemopreventive drug action. J Cell Biochem 1993;17G:14–25. 3. Creutzfeldt W, Lamberts R. Is hypergastrinemia dangerous to man? Scand J Gastroenterol 1991;26(suppl 180):179–191. 4. Sirinek KR, Levine BA, Moyer MP. Pentagastrin stimulates in vitro growth of normal and malignant human colon epithelial cells. Am J Surg 1985;149:35–39. 5. Chu M, Rehfeld JF, Borch K. Effects of gastric fundectomy and antrectomy on the colonic mucosa in the hamster. Digestion 1992;53:28–34. 6. Hakanson R, Sundler F. Trophic effects of gastrin (suppl). Scand J Gastroenterol 1991;180:130–136. 7. Ryberg B, Axelson R, Hakanson R, Sundler F, Mattsson H. Trophic

280

8.

9.

10.

11.

12.

13.

14.

15.

16.

17.

18.

19.

20. 21.

22.

23.

24. 25.

26.

27.

THORBURN ET AL.

effects of continuous infusion of [leu15]-gastrin-17 in the rat. Gastroenterology 1990;98:33–38. Johnson LR, Aures D, Yuen L. Pentagastrin induced stimulation of the in vitro incorporation of [14C]-leucine into protein of the gastrointestinal tract. Am J Physiol 1969;217:251–254. Yassin RR, Clearfield HR, Little KM. Gastrin’s trophic effect in the colon: identification of a signaling pathway mediated by protein kinase C. Peptides 1993;14:1119–1124. Oscarson J, Hakanson R, Liedberg G, Lundqvist G, Sundler F, Thorell J. Variated serum gastrin concentration: trophic effects on the gastrointestinal tract of the rat. Acta Physiol Scand 1979; S475:1–27. Freston JW, Borch K, Brand SJ, Carlsson E, Creutzfeldt W, Hakanson R. Effects of hypochlorhydria and hypergastrinemia on structure and function of gastrointestinal cells: a review and analysis. Dig Dis Sci 1995;40 (suppl):50S–62S. Sobhani I, Lehy T, Laurent-Puig P, Cadiot G, Ruszniewski P, Mignon M. Chronic endogenous hypergastrinemia in humans: evidence for a mitogenic effect on the colonic mucosa. Gastroenterology 1993;105:22–30. Orbuch M, Venzon DJ, Lubensky IA, Weber HC, Gibril F, Jensen RT. Prolonged hypergastrinemia does not increase the frequency of colonic neoplasia in patients with Zollinger–Ellison sydrome. Dig Dis Sci 1996;41:604–613. Smith JP, Wood JG, Solomon TE. Elevated gastrin levels in patients with colon cancer or adenomatous polyps. Dig Dis Sci 1989;34:171–174. Seitz JF, Giovannini M, Gouvernet J, Gauthier AP. Elevated serum gastrin levels in patients with colorectal neoplasia. J Clin Gastroenterol 1991;13:541–545. Wong K, Beardshall K, Waters CM, Calam J, Poston GJ. Postprandial hypergastrinaemia in patients with colorectal cancer. Gut 1991;32:1352–1354. Yapp R, Modlin IM, Kumar RR, Binder HJ, Dubrow R. Gastrin and colorectal cancer: evidence against an association. Dig Dis Sci 1992;37:481–484. Kikendall JW, Glass AR, Sobin LH, Bowen PE. Serum gastrin is not higher in subjects with colonic neoplasia. Am J Gastroenterol 1992;87:1394–1397. Penman ID, El-Omar E, Ardill JES, McGregor JR, Galloway DJ, O’Dwyer PJ. Plasma gastrin concentrations are normal in patients with colorectal neoplasia and unaltered following tumor resection. Gastroenterology 1994;100:1264–1270. Dickinson CJ. Relationship of gastrin processing to colon cancer. Gastroenterology 1995;109:1384–1388. Watson SA, Michaeli D, Grimes S, Morris TM, Robinson G, Varro A, Justin TA, Hardcastle JD. Gastroimmune raises antibodies that neutralize amidated and glycine-extended gastrin-17 and inhibit the growth of colon cancer. Cancer Res 1996;56:880–885. Kochman ML, DelValle J, Dickinson CJ, Boland CR. Posttranslational processing of gastrin in neoplastic human colonic tissues. Biochem Biophys Res Commun 1992;189:1165–1169. Singh P, Xu Z, Dai B, Rajaraman S, Rubin N, Dhruva B. Incomplete processing of progastrin expressed by human colon cancer cells: role of noncarboxyamidated gastrins. Am J Physiol 1994;266: G459–G468. Collen JF, Davis LF. The multitest laboratory in health care. J Occup Med 1969;11:355–360. Friedman GD, Blaner WS, Goodman DS, Vogelman GH, Brind JL, Hodve R, Fireman BH, Orentreich N. Serum retinol and retinolbinding protein levels do not predict subsequent lung cancer. Am J Epidemiol 1986;123:781–789. Lindstedt G, Olbe L, Kilander AF, Armbrecht U, Jagenburg R, Runsteen D, Lundberg PA. Analytical and clinical evaluation of a radioimmunoassay for gastrin. Clin Chem 1985;31:76–82. DelValle J, Sugano K, Yamada T. Glycine-extended processing

GASTROENTEROLOGY Vol. 115, No. 2

28.

29.

30.

31.

32.

33.

34.

35.

36.

37.

38.

39.

40.

41.

42. 43.

intermediates of gastrin and cholecystokinin in human plasma. Gastroenterology 1989;97:1159–1163. Parsonnet J, Friedman GD, Vandersteen DP, Chang Y, Vogelman JH, Orentreich N, Sibley RK. Helicobacter pylori infection and the risk of gastric carcinoma. N Engl J Med 1991;325:1127–1131. Tobi M, Chintalapani S, Goo R, Maliakkai B, Reddy J, Lundqvist M, Oberg K, Luk G. Omeprazole inhibits growth of cancer cell line of colonic origin. Dig Dis Sci 1995;40:1526–1530. Malecka-Panas E, Fligiel SE, Jaszewski R, Majumdar AP. Differential responsiveness of proximal and distal colonic mucosa to gastrin. Peptides 1997;18:559–565. Moore WE, Moore LH. Intestinal floras of populations that have a high risk of colon cancer. Appl Environ Microbiol 1995;61:3202– 3207. Lamberts R, Creutzfeldt W, Struber HG, Brunner G, Solcia E. Long-term omeprazole therapy in peptic ulcer disease: gastrin, endocrine cell growth, and gastritis. Gastroenterology 1993;104: 1356–1370. Lanzon-Miller S, Pounder RE, Hamilton MR, Ball S, Chionos NAF, Raymond F, Olausson M, Cederberg C. Twenty-four hour intragastric acidity and plasma gastrin concentration before and during treatment with either ranitidine or omeprazole. Aliment Pharmacol Ther 1987;1:239–251. Fraser AG, Sawyer AM, Hudson M, Smith M, Pounder RE. Effects of ranitidine 150 mg four times a day on 24-hour intragastric acidity and 24-hour plasma gastrin concentration. Dig Dis Sci 1994;39:91–96. Colin-Jones DG, Langman MJ, Lawson DH, Logan RF, Paterson KR, Vessey MP. Postmarketing surveillance of the safety of cimetidine: 10 year mortality report. Gut 1992;33:1280–1284. Moller H, Lindvig K, Klefter R, Mosbech J, Moller Jensen O. Cancer occurrence in a cohort of patients treated with cimetidine. Gut 1989;30:1558–1562. McCloy RF. Implications of a review of long-term safety of omeprazole and management strategies for peptic disease. Hepatogastroenterology 1992;39:90–91. Penman ID, el-Omar E, McGregor JR, Hillan KJ, O’Dwyer PJ, McColl KE. Omeprazole inhibits colorectal carcinogenesis induced by azoxymethane in rats. Gut 1993;34:1559–1565. Lindstedt G, Runsteen D, Lundberg P, Bengtsson C, Lapidus L, Nystrom E, Kilande AF. On the natural history of hypergastrinemia. Clin Chem 1985;31:1135–1140. Giovannucci E, Colditz GA, Stamfer MJ, Hunter D, Rosner BA, Willett WC, Speize PE. A prospective study of cigarette smoking and risk of colorectal adenoma and colorectal cancer in U.S. women. J Natl Cancer Inst 1994;86:192–199. Martinez ME, McPherson RS, Annegers JF, Levin B. Cigarette smoking and alcohol consumption as risk factors for colorectal adenomatous polyps. J Natl Cancer Inst 1995;87:274–279. Slattery ML, Potter JD, Friedman GD, Ma KN, Edwards S. Tobacco use and colon cancer. Int J Cancer 1997;70:259–264. Todoroki I, Kono S, Shinchi K, Honjo S, Sakurai Y, Wakasayashi K, Imanishi K, Nishikawa H, Ogawa S, Katsurada M. Relationship of cigarette smoking, alcohol use, and dietary habits with sigmoid colon adenomas. Ann Epidemiol 1995;5:478–483.

Received August 8, 1997. Accepted April 28, 1998. Address requests for reprints to: Julie Parsonnet, M.D., HRP Redwood Building Room T225, Stanford University, Stanford, California 94305-5405. Fax: (650) 725-6951. Supported by National Institutes of Health grants R03CA65801 (to J.P.), RO1 DK47398 (to C.J.D.), and R35CA49761 (to G.D.F.), by the Michigan Peptide Research Center (P30 DK34933; to C.J.D.), and by the Stanford Medical Scholars Program (to C.M.T.).