Aspirin and Nonsteroidal Anti-inflammatory Agents and Risk for Colorectal Adenomas

Transcription

1 GASTROENTEROLOGY 1998;114: Aspirin and Nonsteroidal Anti-inflammatory Agents and Risk for Colorectal Adenomas ROBERT S. SANDLER, JOSEPH C. GALANKO, SHARON C. MURRAY, JAMES F. HELM, and JOHN T. WOOSLEY Division of Digestive Diseases and Nutrition, Center for Gastrointestinal Biology and Disease, and Departments of Epidemiology, Pathology, and Biostatistics, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina Background & Aims: Aspirin and nonsteroidal antiinflammatory drugs have been reported to protect against the development of colorectal cancer. Because adenomas are precursors to most colorectal cancers, the aim of this study was to examine the relationship of these medications to the risk for colorectal adenomas in a colonoscopy-based case-control study. Methods: Study participants were drawn from patients who underwent colonoscopy at the University of North Carolina Hospitals. Medication use was assessed by telephone using a comprehensive list of prescription and nonprescription drugs as well as questions about dietary and lifestyle factors that might be relevant for adenoma development. Results: There were 210 patients with adenomas and 169 adenomafree controls. After adjusting for potential confounders, regular users were half as likely to currently have adenomas compared with nonusers (adjusted odds ratio, 0.56; 95% confidence interval, ). Regular users who stopped medication at least 1 year before colonoscopy were still protected (adjusted odds ratio, 0.59; 95% confidence interval, ), although small numbers make this conclusion tentative. The protective effects of aspirin and the nonaspirin nonsteroidal anti-inflammatory drugs were similar. Conclusions: The results suggest that aspirin and nonsteroidal anti-inflammatory drugs cause early disruption of the adenoma-carcinoma sequence. The challenge for the future will be to learn more about dose, duration, and mechanism of action. There is increasing evidence to suggest that aspirin and other nonsteroidal anti-inflammatory drugs (NSAIDs) reduce the risk for colorectal cancer. 1 The mechanism for cancer protection is unknown and speculative but may involve the induction of apoptosis by these medications. 2 The observation is supported by animal experiments that show fewer tumors per animal and fewer animals with tumors after administration of several different NSAIDs. 3 Human studies also show lower risk for colonic neoplasia among aspirin users. Supportive evidence comes from both case-control 4 12 and cohort studies The reduced risk has been found for men and women, for cancers of the colon and the rectum, and for both aspirin and the other NSAIDs. Earlier detection of lesions as a result of aspirin-induced bleeding does not seem to explain the findings. 13 Colorectal adenomas are currently believed to be the precursor lesions to most colorectal cancers, and there is reason to believe that preventing or removing adenomas can prevent colorectal cancer. The lower than expected cancer incidence after adenoma removal in the National Polyp Study seems to support this hypothesis. Not all adenomas progress to cancer; therefore, the risk factors for adenomas are not necessarily the same as the risk factors for cancer. It would be of interest to determine whether aspirin and NSAIDs protect against adenomas in the same way as they protect against cancer. Previous studies have examined the association between NSAIDs and sporadic adenomas. 5,7,8,13 The present study was designed to explore this association while controlling for a number of potentially confounding dietary and lifestyle factors. We also sought to examine possible interaction with coffee drinking in light of a recent study showing marked synergy between caffeine and the aspirin-like compound sulindac sulfide in inducing apoptosis in a colon adenocarcinoma cell line. 18 Materials and Methods The study was designed to examine a number of dietary and lifestyle factors associated with rectal mucosal proliferation and adenomas. Study participants were drawn from patients who underwent colonoscopy at the University of North Carolina Hospitals between August 15, 1992, and July 25, Participants entered the study if they agreed to have biopsy specimens taken from their large bowel during the course of a clinically indicated colonoscopy, to have a blood sample drawn, and to participate in a telephone interview. Six biopsy specimens were taken from the rectum in consenting Abbreviations used in this paper: BMI, body mass index; CI, confidence interval; OR, odds ratio by the American Gastroenterological Association /98/$3.00

2 442 SANDLER ET AL. GASTROENTEROLOGY Vol. 114, No. 3 subjects and used for a study of rectal mucosal proliferation, the results of which are being reported separately. The study was approved by the Committee for the Protection of the Rights of Human subjects at the University of North Carolina, and all participants provided written informed consent. Patients were excluded for any of the following: age less than 30 years; presence of polyposis, defined as more than 100 polyps; inability to provide informed consent; colitis of any type (e.g., radiation, infectious, or idiopathic); previous colon resection; current or previous colon cancer; unsatisfactory colon preparation as judged by the colonoscopist; or incomplete examination (i.e., cecum not reached). Indications for the procedure and eligibility and the size, shape, and location of any polyps seen were abstracted from colonoscopy reports. Patients with previous adenomas were included in the study. To be certain that information on previous adenomas was correctly recorded, colonoscopy reports from previous examinations were reviewed. The slides from each polyp were reviewed by the study pathologist ( J.T.W.). Polyps were classified as adenomatous (tubular, tubulovillous, or villous) or nonadenomatous based on standard criteria. If polyps had mixed histology (e.g., hyperplastic and adenomatous elements), they were classified as adenomas. Cases were defined as patients with one or more adenomatous polyps. Controls were defined as individuals with no adenomatous polyps. Patients with previous adenomas, but no current adenomas, were assessed in a separate exploratory analysis. Patients with nonadenomatous polyps (e.g., hyperplastic polyps) were eligible to be controls. A 10% systematic sample of specimens with a polyp (either adenomatous or nonadenomatous) was resubmitted to the pathologist in a blind fashion. There were no instances where adenoma cases were classified as nonadenoma cases when the slides were interpreted a second time. Eligible patients were interviewed by telephone by a trained interviewer who was blind to the status of the patient as a case or control. Subjects were asked about consumption of aspirin, aspirin-containing combination drugs (including combinations commonly used in our area such as Goody s [Goody s Pharmaceuticals, Memphis, TN] and BC powders [Block Drug Co., Jersey City, NJ]), and other NSAIDs during the 5 years before the interview. The list of medications was comprehensive and included all prescription and nonprescription nonsteroidals listed in the Physician s Desk Reference in The list also included acetaminophen and acetaminophen-containing compounds. Patients were asked about medicines that may have been obtained anywhere including a doctor s prescription, hospital or neighborhood clinic, pharmacy, supermarket, store, friends, neighbors, and relatives. They were also asked whether they took these medications for any of the following indications: headache, backache, arthritis, bursitis, rheumatism, joint pain, injury, accident, operation, migraine, sinus trouble, or (for women) menstrual cramps. They were asked how frequently they took the drug and the interval since last use. We defined regular users as those who took any of these medications for a total of more than 15 times per month at some time during the 5 years before the interview. Current users were defined as those who took medication more than 15 times per month during the year before the interview. Past users took the drug more than 15 times per month at some time during the 5 years before the interview, but not during the previous 12 months. Nonusers were defined as those who never took medication more than 15 times in a month. We also obtained detailed information about diet. The dietary instrument consisted of a validated quantitative food frequency questionnaire that was developed at the National Cancer Institute. 19,20 Subjects were asked about their intake of more than 100 foods that have been shown to contribute significantly and importantly to the intake of calories and 17 macronutrients and micronutrients in the national diet. 21,22 Participants were asked to estimate how often, on average, they consumed each of the food items during the previous year. A 1-year period was chosen to provide a full cycle of seasons so that responses, at least in theory, would be independent of the time of year. 23 Subjects were also asked to estimate whether their usual portion size was small, medium, or large for each item. Information on the size of a medium portion was provided by interviewers on request. Nutrient intakes were calculated by an analysis program provided by the National Cancer Institute that incorporates the nutrient content of each item, the frequency, and a portion size based on age. Initially, case and control subjects were compared with respect to NSAID use as well as various demographic and dietary factors in an uncontrolled analysis. We used analysis of variance for comparison of continuous variables and the Pearson s 2 test for general association for categorical variables. Crude odds ratios (ORs) comparing regular NSAID users to controls were calculated. We used logistic regression analysis to assess the effects of regular NSAID use on adenoma status. 24 The three categories for adenoma status were case, control, and controls with a history of adenomas. Two separate analyses were conducted. In the first analysis, cases were compared with controls without a history of having adenoma. This analysis was of primary interest. In a second logistic regression analysis, patients with a history of adenomas were compared with controls with no such history. For each of these analyses, the logistic regression model was combined with a backward stepwise process to determine risk due to regular NSAID use, adjusting for potentially relevant confounding factors. In a separate exploratory analysis, NSAID use was categorized by time of use (current, past, or never) to determine whether the effect of taking NSAIDs diminished with time. NSAID use, age (continuous), sex, and race (white vs. nonwhite) were forced into the model. Other factors considered for inclusion in the model were body mass index (BMI) (BMI Weight [kg]/height [m 2 ]), coffee drinker (yes/no), daily caloric intake, family history of colon cancer (yes/no), fat intake, fiber intake, alcohol drinker (yes/no), smoking (current, past, or never), daily servings of fruit and fruit juices, and daily servings of vegetables. Fat and fiber intake were adjusted for daily caloric intake using the method of Willett and Stampfer. 25 All

3 March 1998 NSAIDs AND COLORECTAL ADENOMAS 443 of these potentially confounding factors were included in a preliminary model. At each step, the least significant of these factors considered for inclusion was eliminated from the model, and parameter estimates were recalculated until contrasts for all of the remaining factors had associated P values of Factors that were forced into the model may not necessarily have had associated P values of To insure that confounding factors with associated P values of 0.05 were not erroneously removed from the model, we added back into the final model, one at a time, each of the potential confounders and examined the change in the OR for NSAID use. If the OR for current NSAID use changed by more than 10%, then the potential confounder was included in the final model. To explore a possible interaction between coffee and NSAID use, an additional model was considered. This model included terms for coffee drinking, NSAID use, their interaction, and other relevant factors as determined previously. Results There were 3488 colonoscopies performed between August 15, 1992, and July 31, 1995; 1572 patients were ineligible and therefore not asked to participate. An additional 115 gave consent but were found to be ineligible at the time of colonoscopy (most commonly because of poor preparation). There were 707 subjects who were potentially eligible but were not asked to participate (generally because the nursing staff were too busy with primary patient care responsibilities) and 602 who were eligible but refused. A total of 492 subjects met eligibility criteria and consented to undergo biopsies and blood sampling. Of this group, 16 were excluded because they had cancer, and 80 were excluded because of incomplete interviews. Another 17 were excluded because of poor-quality dietary data (e.g., improbably low total calories). The analyses that follow are based on 379 patients. The descriptive characteristics of the study population are shown in Table 1. The mean age of the study group was 59.3 years (SE, 0.6). Patients currently with adenomas and subjects with previous adenomas were slightly older than the nonadenoma controls (P ). The proportion of men in the current adenoma group was also greater than in the other two groups (P 0.08), and there were more whites (P 0.03). BMI and family history of colorectal cancer were similar in all groups. Selected dietary characteristics of the groups, including dietary fat, fiber, total calories, coffee consumption, alcohol use, number of daily servings of fruit and fruit juices, and number of daily servings of vegetables were similar in all groups. The proportion of cigarette smokers was greater in the nonadenoma controls. Among patients with adenomas, the most common Table 1. Descriptive Characteristics of the Study Population Variable means and proportions Overall (n 379) Adenoma status Current adenoma (n 142) Previous adenoma (n 68) Controls (n 169) Age (yr) b White (%) a Male (%) BMI (kg/m 2 ) Family history (%) Fat (g/day) Fiber (g/day) Calories per day Fruit (servings/day) Vegetable (servings/ day) Coffee use (%) Alcohol use (%) Cigarette smokers (%) NSAID user (%) a P b P reasons for colonoscopy were follow-up of a polyp (38%) and bleeding, anemia, and positive fecal occult blood test (24.6%). Among nonadenoma controls, the most common reasons for colonoscopy were bleeding, anemia, and positive fecal occult blood test (57.5%). The majority of those with previous adenomas (89.7%) underwent colonoscopy for follow-up of a polyp, as might be expected. A list of the NSAIDs and percent of subjects who reported using them is provided in Table 2. Aspirin use and ibuprofen use at some point during the last 5 years were reported most frequently (71.8% and 53.8%, respectively). The categories are not mutually exclusive, i.e., people often reported taking more than one drug. Also, not everyone who reported taking an NSAID is classified as a regular user, although these people are included in Table 2. Reasons for use of the NSAID drugs are listed in Table 3 for regular users of NSAIDs. Regular users were Table 2. Percent of Subjects Reporting Use of Aspirin and Nonaspirin NSAIDs Drug Percent taking drug in last 5 years a Aspirin 71.8 Ibuprofen 53.8 Naprosyn, Anaprox 8.4 Voltaren 2.4 Feldene 4.3 Ansaid, Orudis, Dolobid, or Tolectin 2.6 Clinoril 1.9 Indocin 2.7 Fiorinal 3.2 a Study subjects (n 379) may have taken more than one type of drug so that the percentages total more than 100%.

4 444 SANDLER ET AL. GASTROENTEROLOGY Vol. 114, No. 3 Table 3. Reported Conditions for NSAID Use Among 194 Regular Users Condition Number Percent Headaches Back trouble, back pain, or sciatica Arthritis, bursitis, rheumatism, or other joint soreness Pain from injury, accident, or operation Migraine headaches Sinus trouble Menstrual pain or cramps Other Total NOTE. Regular users took aspirin or nonaspirin NSAIDs more than 15 times per month. Subjects reported one condition for each medication taken. individuals who took NSAID medications more than 15 times per month. Information on condition was collected for each drug reported. For example, an individual who took aspirin for headaches and ibuprofen for arthritis would appear twice in the table. Therefore, the total number of conditions reported in Table 3 (572) is much greater than the number of regular users (194). Most commonly reported conditions were headaches (28.0%) and arthritis, bursitis, rheumatism, or other joint soreness (20.5%). NSAID use by adenoma status is shown in Table 4. There was a protective effect of NSAID use on adenoma development, with a crude OR of 0.73 (95% confidence interval [CI], ). After adjusting for age, sex, race, and BMI, the effect was even stronger, with an adjusted OR of 0.56 (95% CI, ). To determine which variable had the most influence on the difference between the crude and the adjusted OR, we built a series of models adjusting for one factor at a time. The ORs for NSAID use adjusting singly for age, sex, race, or BMI were 0.66, 0.70, 0.71, and 0.71, respectively. These models indicate that age was the factor most responsible for the difference between the crude and adjusted ORs. Comparing cases to the nonadenoma control group and categorizing regular use of NSAID by time of use, there were 138 current users, 20 past users, and 143 nonusers Table 4. Risk for Adenoma Development Associated With NSAID Use NSAID use Crude Adjusted a OR 95% CI a OR 95% CI a Never (n 150) Regular user (n 157) b a Adjusted for age, sex, race, and BMI. Significance levels for age, sex, race, and BMI in the adjusted model were , , , and , respectively. b Four subjects were missing covariate information and were excluded from analyses. All 4 were classified as regular NSAID users. of NSAIDs. Current NSAID users, that is, those who were regular users of NSAIDs in the previous year, were protected, and ORs were similar to those found in the analysis of regular users (defined by regular use in the past 5 years). Past NSAID users, that is, individuals who had regularly consumed NSAIDs at some point during the 5 years before the interview, but not in the 12 months before the interview, were protected, but small numbers of past users lead to wide CIs (crude OR, 0.66; 95% CI, ; adjusted OR, 0.59; 95% CI, ). Ten subjects were excluded from this analysis; 9 were excluded because of missing information about whether the medication had been taken in the past year and 1 because of missing covariate information. We also examined the effect of NSAIDs in a group of 68 subjects who had an adenoma in the past but did not have an adenoma at the time of the qualifying colonoscopy. Small numbers in this category resulted in unstable estimates. The proportion of NSAID users in the previous adenoma group was similar to that in the nonadenoma control group in the crude analysis, with an OR ratio of 0.80 (95% CI, ). After adjusting for age, sex, race, and BMI, the adjusted OR was 0.63 (95% CI, ). The adjusted OR is similar to that for individuals with current adenomas. The results would imply that individuals with a previous adenoma were less likely to have an adenoma at present because of NSAID use, but these conclusions must be tempered by the fact that we do not have good information on when the past adenoma occurred in relation to NSAID use. If NSAIDs are protective against adenoma development, specificity of the effect might be expected. That is, analgesics in other categories would not be expected to be protective. Our questionnaire sought information on non-nsaid analgesics such as acetaminophen, propoxyphene, and codeine. Defining non-nsaid users as people who took these medications more than 15 times per month, we were able to compare non-nsaid users with people who used neither NSAIDs nor non-nsaid analgesics. Excluding regular users of NSAIDs, 75 case and 75 control subjects remained for this analysis. Of these, 13 cases and 15 controls were classified as regular users of non-nsaid pain medications (unadjusted OR, 0.84; 95% CI, ). We also compared the risk for aspirin with that of nonaspirin NSAIDs. To make this comparison we excluded from analysis those who took both aspirin and nonaspirin NSAIDs more than 15 times per month (10 cases and 22 controls). There were 88 users of aspirin, 38 users of nonaspirin NSAIDs, and 153 nonusers for this analysis. The OR for use of aspirin was 0.84 (95% CI,

5 March 1998 NSAIDs AND COLORECTAL ADENOMAS ), whereas the OR for use of nonaspirin NSAIDs was 0.74 (95% CI, ). One of our hypotheses was that coffee consumption might potentiate the effect of NSAIDs based on in vitro studies. We added a main effect for coffee drinking as well as a term representing an interaction effect between regular NSAID use and coffee drinking to the model described in Table 4. When a main effect for coffee drinking, but no interaction term, was included, there was a modest but nonsignificant effect for coffee (OR, 0.73; 95% CI, ). When both coffee and the coffee plus NSAID interaction were added to the model, there was no evidence to support interaction (OR for coffee plus NSAID interaction, 1.01; 95% CI, ). Discussion We found a strong protective effect of NSAID medications against the development of colorectal adenoma. Regular users of NSAIDs (more than 15 times per month during the 5 years before the interview) were about half as likely to develop adenomas. Subjects who stopped taking NSAIDs during the year before colonoscopy did not lose their protection, although small numbers make this conclusion tentative. Because we asked a comprehensive set of questions about diet and lifestyle, we were able to control for a number of potentially confounding factors. These other factors did not diminish the effect. The findings are similar to estimates reported in studies of NSAIDs and colorectal cancer, suggesting that NSAIDs operate at a step in the adenoma-carcinoma sequence that is proximal to adenoma development. This result supports the use of aspirin and NSAIDs in chemoprevention trials in which adenomas are the end point. The mechanism whereby NSAIDs might prevent the development of colorectal adenomas is not known, although several potential mechanisms have been proposed. The effects of this class of pharmacological agents are exceedingly diverse, and any one of several mechanisms might be operative. It would be logical to hypothesize that the effects were caused by inhibition of prostaglandins. Experimental animal and human tumors have been shown to contain or produce large quantities of prostaglandins, 26,27 and prostaglandin synthesis has been shown to be associated with tumor promotion and metastatic potential. 28 Prostaglandins, especially prostaglandin E 2, are biological response modifiers that can affect cell proliferation and tumor growth. 29 The precise role of prostaglandins in the development of human colorectal adenomas and cancers remains uncertain. A number of other mechanisms have been postulated for the chemopreventive effects of these drugs. NSAIDs inhibit cell growth and [ 3 H]thymidine incorporation into cellular DNA in cell culture models. 28 Indomethacin and other NSAIDs inhibit DNA synthesis, cyclic adenosine accumulation, synthesis of growth factors in response to tumor promoters, and the progression of cultured cells through the cell cycle. 30 In cell culture, indomethacin inhibits G 1 to S phase cell cycle transition and reduces overall DNA synthesis. 29 NSAIDs interfere with membrane-associated processes such as G protein signal transduction, transmembrane calcium flux, and cell-tocell binding; they inhibit other enzymes such as phosphodiesterase, cyclic adenosine 3,5 -monophosphate dependent protein kinase, and some folate-dependent enzymes. 28 Aspirin and NSAIDs may have an effect on immune function. There are certain aspects of our study that distinguish it from previous studies. First, our study was specifically designed to examine the NSAID-adenoma hypothesis rather than to use existing data to examine the hypothesis retrospectively. We obtained very detailed information about all available NSAID medications, the frequency of their use, and indications. Second, we obtained comprehensive information about a range of dietary and lifestyle factors that could potentially confound the association. We used a well-validated dietary instrument administered by carefully trained interviewers. Third, all pathological slides were reviewed by a single pathologist using predefined criteria. Finally, all subjects in this study underwent full colonoscopy. If the examination was incomplete, or if the preparation was judged as marginal, the subjects were not included. Given the high sensitivity of colonoscopy, this feature of the study significantly reduces misclassification errors. The study is not without its limitations. The most obvious is the low participation rate. To participate in the study, a subject must have agreed to have six biopsy specimens taken from the rectum during the course of a routine colonoscopy. We believe that the invasiveness of this intervention had a detrimental effect on participation. When epidemiological studies begin to include invasive techniques to obtain biological markers, they may be faced with similar problems of nonresponse. Whether nonresponse may have biased the study is uncertain. The fact that our findings are similar to those of previous studies is somewhat reassuring. Because of the retrospective nature of the study, there are insufficient data to examine dose response. We asked people how often they typically took each of the NSAID medications during the past 5 years. We did not ask about dose or duration because we did not believe that

6 446 SANDLER ET AL. GASTROENTEROLOGY Vol. 114, No. 3 people could provide accurate information. In addition, our study subjects often reported taking more than one of the drugs. We therefore aggregated the data to calculate the average monthly number of episodes of NSAID use. Despite this limitation, we would point out that information from observational epidemiological studies such as ours has been crucial to the preliminary understanding of a number of diseases and has provided the background data to justify clinical trials. Ultimately, detailed information on agent, dose, and duration can only come from clinical trials that are costly and time consuming. Increasing evidence supports the hypothesis that aspirin and NSAIDs inhibit the development of benign and malignant neoplasms of the large intestine. To translate this finding into public health recommendations, we need better information about the dose and duration of medication. Although further observational studies may provide helpful data, the results of ongoing randomized trials may be necessary. It is also important to recognize that this class of medications can have widespread effects. Primary prevention studies have shown that low-dose aspirin can prevent myocardial infarction These studies also show that the risk of certain adverse outcomes such as stroke and heart disease other than myocardial infarction is increased among those taking aspirin. Aspirin may also increase the risk for gastrointestinal hemorrhage. It is necessary to consider the full range of risks and benefits from these medications and not simply focus on protective effects against colorectal neoplasia. In summary, we found that premalignant colorectal adenomas were less likely among individuals who regularly consumed aspirin and NSAIDs. The order of magnitude of the protective effect was similar to that found for colorectal cancer. The results suggest that these medications cause some early disruption of the adenomacarcinoma sequence. The challenge for the future will be to learn more about dose, duration, and mechanism of action. References 1. Greenberg ER, Baron JA. Prospects for preventing colorectal cancer death. J Natl Cancer Inst 1993;85: Shiff SJ, Koutsos MI, Qiao L, Rigas B. Nonsteroidal antiinflammatory drugs inhibit the proliferation of colon adenocarcinoma cells: effects on cell cycle and apoptosis. Exp Cell Res 1996;222: Reddy BS, Rao CV, Rivenson A, Kelloff G. Inhibitory effect of aspirin on azoxymethane-induced colon carcinogenesis in F344 rats. Carcinogenesis 1993;14: Kune GA, Kune S, Watson LF. Colorectal cancer risk, chronic illnesses, operations, and medications: case control results from the Melbourne Colorectal Cancer Study. Cancer Res 1988;48: Martinez, McPherson RS, Levin B, Annegers JF. Aspirin and other nonsteroidal anti-inflammatory drugs and risk of colorectal adenomatous polyps among endoscoped individuals. Cancer Epidemiol Biomarkers Prev 1995;4: Reeves MJ, Newcomb PA, Trentham-Dietz A, Storer BE, Remington PL. Nonsteroidal anti-inflammatory drug use and protection against colorectal cancer in women. Cancer Epidemiol Biomarkers Prev 1996;5: Peleg II, Lubin MF, Cotsonis GA, Clark WS, Wilcox CM. Long-term use of nonsteroidal anti-inflammatory drugs and other chemopreventors and risk of subsequent colorectal neoplasia. Dig Dis Sci 1996;41: Logan RF, Little J, Hawtin PG, Hardcastle JD. Effect of aspirin and nonsteroidal anti-inflammatory drugs on colorectal adenomas: case-control study of subjects participating in the Nottingham faecal occult blood screening programme. BMJ 1993;307: Muscat JE, Stellman SD, Wynder EL. Nonsteroidal anti-inflammatory drugs and colorectal cancer. Cancer 1994;74: Rosenberg L, Palmer JR, Zauber AG, Warshauer ME, Stolley PD, Shapiro S. A hypothesis: nonsteroidal anti-inflammatory drugs reduce the incidence of large-bowel cancer. J Natl Cancer Inst 1991;83: Suh O, Mettlin C, Petrelli NJ. Aspirin use, cancer, and polyps of the large bowel. Cancer 1993;72: Greenberg ER, Baron JA, Freeman DHJ, Mandel JS, Haile R. Reduced risk of large-bowel adenomas among aspirin users: the Polyp Prevention Study Group. J Natl Cancer Inst 1993;85: Giovannucci E, Rimm EB, Stampfer MJ, Colditz GA, Ascherio A, Willett WC. Aspirin use and the risk for colorectal cancer and adenoma in male health professionals. Ann Intern Med 1994;121: Gridley G, McLaughlin JK, Ekbom A, Klareskog L, Adami HO, Hacker DG, Hoover R, Fraumeni JF. Incidence of cancer among patients with rheumatoid arthritis. J Natl Cancer Inst 1993;85: Paganini Hill A., Hsu G, Ross RK, Henderson BE. Aspirin use and incidence of large-bowel cancer in a California retirement community. J Natl Cancer Inst 1991;83: Schreinemachers DM, Everson RB. Aspirin use and lung, colon, and breast cancer incidence in a prospective study. Epidemiology 1994;5: Thun MJ, Namboodiri MM, Heath CWJ. Aspirin use and reduced risk of fatal colon cancer. N Engl J Med 1991;325: Goldberg Y, Qiao L, Tsai LL, Rigas B. Caffeine potentiates the anti-proliferative effect of sulindac sulfide by increasing the amount of apoptosis in HT-29 colon adenocarcinoma cells (abstr). Gastroenterology 1996;110:A Block G, Hartman AM, Dresser CM, Caroll MD, Gannon J, Gardner L. A data-based approach to diet questionnaire design and testing. Am J Epidemiol 1986;124: Block G. Dietary guidelines and the results of food consumption surveys. Am J Clin Nutr 1991;53:356S 357S. 21. Block G, Dresser CM, Hartman AM, Caroll MD. Nutrient sources in the American diet: quantitative data from the NHANESII survey. I. Vitamins and minerals. Am J Epidemiol 1985;122: Block G, Dresser CM, Hartman AM, Caroll MD. Nutrient sources in the American diet: quantitative data from the NHANESII survey. II. Macronutrients and fats. Am J Epidemiol 1985;122: Willett W. Nutritional epidemiology. New York: Oxford University, 1990: Agresti A. Categorical Data Analysis. New York: Wiley, 1990: Willett W, Stampfer MJ. Total energy intake: implications for epidemiologic analyses. Am J Epidemiol 1986;124: Rigas B, Goldman IS, Levine L. Altered eicosanoid levels in human colon cancer. J Lab Clin Med 1993;122:

7 March 1998 NSAIDs AND COLORECTAL ADENOMAS Suh O, Mettlin C, Petrelli NJ. Aspirin use, cancer, and polyps of the large bowel. Cancer 1993;72: Hixson LJ, Alberts DS, Krutzsch M, Einsphar J, Brendel K, Gross PH, Paranka NS, Baier M, Emerson S, Pamukcu R. Antiproliferative effect of nonsteroidal antiinflammatory drugs against human colon cancer cells. Cancer Epidemiol Biomarkers Prev 1994;3: Earnest DL, Hixson LJ, Alberts DS. Piroxicam and other cyclooxygenase inhibitors: potential for cancer chemoprevention. J Cell Biochem Suppl 1992;16I: Waddell WR, Ganser GF, Cerise EJ, Loughry RW. Sulindac for polyposis of the colon. Am J Surg 1989;157: Peto R, Gray R, Collins R, Wheatley K, Hennekens C, Jamrozik K, Warlow C, Hafner B, Thompson E, Norton S, Gilliland J, Doll R. Randomised trial of prophylactic daily aspirin in British male doctors. Br Med J 1988;296: Final report on the aspirin component of the ongoing Physicians Health Study. Steering Committee of the Physicians Health Study Research Group. N Engl J Med 1989;321: Collaborative overview of randomised trials of antiplatelet therapy I: prevention of death, myocardial infarction, and stroke by prolonged antiplatelet therapy in various categories of patients. Antiplatelet Trialists Collaboration. BMJ 1994;308: Received April 30, Accepted November 11, Address requests for reprints to: Robert S. Sandler, M.D., Division of Digestive Diseases and Nutrition, CB # 7080, 423A Burnett Womack Building, University of North Carolina, Chapel Hill, North Carolina Fax: (919) Supported in part by grants P30 DK34987, R01 CA66635, and R01 CA44684 from the National Institutes of Health.