Coffee Consumption and Incidence of Colorectal Cancer in Two Prospective Cohort Studies of Swedish Women and Men

Abstract

Investigators have reported an inverse association between coffee consumption and risk of colorectal cancer in several case-control studies, but prospective studies, most of them involving small numbers of cases, have not supported such a relation. In this analysis, the authors prospectively examined the association of coffee consumption with colorectal cancer risk among participants from two population-based cohort studies: 61,433 women in the Swedish Mammography Cohort and 45,306 men in the Cohort of Swedish Men. Information about coffee consumption was obtained from food frequency questionnaires in 1987–1990 and 1997 for women and in 1997 for men. The authors used Cox proportional hazards modeling for cohort-specific multivariate analyses, and results were pooled using random-effects models. During 1,240,597 person-years of follow-up, 1,279 incident cases of colorectal cancer were diagnosed. Coffee consumption was not associated with risk of colorectal cancer, colon cancer, or rectal cancer in either women or men. For both cohorts combined, the multivariate rate ratio for colorectal cancer for each additional cup of coffee per day was 1.00 (95% confidence interval: 0.97, 1.04). The associations were not modified by colorectal cancer risk factors. The findings from these two large prospective cohort studies do not support the hypothesis that coffee consumption lowers the risk of colorectal cancer.

Coffee consumption could conceivably reduce the risk of colorectal cancer through its content of potentially antimutagenic substances and phenolic compounds with antioxidant properties (1–3). Furthermore, consumption of coffee might decrease colorectal cancer risk by increasing large bowel motility in the rectosigmoid region (4) or by reducing the excretion of bile acids and sterols into the colon (5). However, some constituents of coffee might have genotoxic and mutagenic properties (1).

The association between coffee drinking and colorectal cancer risk has been the subject of several case-control studies and a limited number of prospective cohort studies. Case-control studies have tended to support an inverse association between coffee consumption and risk of colorectal cancer, especially case-control studies conducted in Europe (6, 7). In a meta-analysis of coffee consumption and colorectal cancer risk, the combined results of 12 case-control studies revealed a significant 28 percent reduction in colorectal cancer risk for high coffee consumption versus low consumption (6). Prospective cohort studies have not supported a significant reduction in colorectal cancer risk associated with coffee consumption (6–8), but most have been based on small numbers of cases.

Clarifying the role of coffee drinking in colorectal carcinogenesis is important because of the widespread consumption of coffee and the high incidence of colorectal cancer in Western societies. Therefore, we prospectively examined the relation between coffee consumption and colorectal cancer incidence in Swedish women and men from two large population-based cohort studies: the Swedish Mammography Cohort and the Cohort of Swedish Men, respectively. In a previous analysis of the Swedish Mammography Cohort—based on a single assessment of diet at baseline and 460 colorectal cancer cases, with follow-up from 1987 through 1998—we observed no association between coffee consumption and colorectal cancer incidence (9). In the present study, with almost three times as many cases in the analyses (a total of 1,279 cases among women and men), we had more statistical power to observe moderate associations and to examine risk by subsite within the colon. In addition, for the Swedish Mammography Cohort, we used information on diet collected prospectively from women at two time points to better represent long-term coffee consumption.

MATERIALS AND METHODS

Study populations

Two population-based cohorts provided prospective data for our analyses: the Swedish Mammography Cohort and the Cohort of Swedish Men. The Swedish Mammography Cohort was established between 1987 and 1990, when 66,651 women aged 40–76 years and residing in central Sweden (Västmanland and Uppsala counties) completed a mailed questionnaire regarding their diet, weight, height, and education. In September 1997, an expanded questionnaire was sent to participants in the Swedish Mammography Cohort to update the dietary data and to obtain information on other lifestyle factors and medical history. Likewise, the Cohort of Swedish Men was created in the autumn of 1997, when 48,850 men aged 45–79 years and living in central Sweden (Västmanland and Örebro counties) responded to a mailed questionnaire that was identical to the Swedish Mammography Cohort questionnaire from 1997. Both studies (the Swedish Mammography Cohort and the Cohort of Swedish Men) were approved by the ethics committees at the Karolinska Institutet in Stockholm and Uppsala University Hospital and the regional ethics committee for the Örebro region.

Dietary assessment

Among women in the Swedish Mammography Cohort, diet was assessed in 1987–1990 and in 1997; among men in the Cohort of Swedish Men, diet was assessed in 1997. In the Swedish Mammography Cohort, the food frequency questionnaire used for the baseline dietary assessment included 67 food items. On this questionnaire, women were asked to report their average frequency of consumption of each item, including coffee, over the previous 6 months. The eight predefined responses ranged from “never/seldom” to “four or more times per day.” A more comprehensive food frequency questionnaire (96 items) was used to assess dietary intake in the two cohorts in 1997. On this expanded food frequency questionnaire, participants reported their average frequency of consumption of each item over the previous year. Coffee consumption was inquired about as an open question. In a validation study of the baseline dietary questionnaire for the Swedish Mammography Cohort, the Spearman correlation coefficient for correlation between questionnaire data and data from the average of four 1-week diet records was 0.63 for coffee consumption.

Case ascertainment and follow-up

Follow-up of the cohorts was achieved through computerized record linkage to the Swedish cancer, death, and population registers. Follow-up for cancer is estimated to be nearly 100 percent complete (10). Only cases of invasive adenocarcinoma were included in this analysis. In the Swedish Mammography Cohort, 830 incident cases of colorectal cancer (560 colon, 263 rectal, and seven at both sites) were diagnosed from 1987–1990 through 2004; 274 incident cases of colorectal cancer (191 colon, 81 rectal, and two at both sites) were diagnosed from 1998 through 2004. In the Cohort of Swedish Men, 449 incident cases of colorectal cancer (276 colon and 173 rectal) were diagnosed from 1998 through 2004.

Statistical analysis

For our primary analyses, 61,433 women (Swedish Mammography Cohort) and 45,306 men (Cohort of Swedish Men) were eligible after exclusion of participants with an erroneous or incomplete national registration number, those with an implausibly low or high total energy intake (i.e., three standard deviations from the mean value for log_e-transformed energy intake for women and men separately), and those diagnosed with cancer (except nonmelanoma skin cancer) prior to baseline. We performed additional analyses by starting follow-up in 1998 for the Swedish Mammography Cohort (follow-up started in 1998 for all analyses of the Cohort of Swedish Men); 36,616 women completed the follow-up questionnaire and were free from diagnosed cancer before 1998.

Person-time of follow-up was contributed by each participant from 1987–1990 or 1998 for women and from 1998 for men until the date of colorectal cancer diagnosis, the date of death from any cause, the date of moving out of the study area, or December 31, 2004, whichever occurred first. Coffee consumption was analyzed both as a categorical variable and as a continuous variable. Incidence rates of colorectal cancer were calculated by dividing the number of incident cases by the number of person-years in each category of coffee consumption. We computed rate ratios for each of the upper categories by dividing the rates in these categories by the rate in the reference (lowest) category.

For the main analyses of the Swedish Mammography Cohort, we used dietary information from the baseline and 1997 dietary questionnaires to better represent long-term average coffee consumption and to reduce random within-person measurement error. In these analyses, cancer incidence from baseline through 1997 was related to coffee consumption reported on the baseline dietary questionnaire, and cancer incidence from 1998 through 2004 was related to the average of coffee consumption at baseline and in 1997.

Rate ratios were estimated using Cox proportional hazards models (11) stratified by age in months. The models for which results are presented all satisfied the proportional hazards assumption. In multivariate analyses, in addition to age, we simultaneously adjusted for education (less than high school, high school graduation, or more than high school), family history of colorectal cancer (yes/no), history of diabetes (yes/no), smoking (never, past, or current), body mass index (weight (kg)/height (m)²; <23.0, 23.0–24.9, 25.0–29.9, or ≥30), physical activity (low (≤1 hour/week), medium (2–5 hours/week), or high (≥6 hours/week)), aspirin use (“no”; “yes, <10 years”; or “yes, ≥10 years”), multivitamin supplement use (“no”; “yes, occasionally”; or “yes, regularly”), and daily intakes of total energy (kcal; continuous), fruit (quartiles), vegetables (quartiles), milk (quartiles), and red meat (quartiles). In analyses of women, we also adjusted for postmenopausal hormone use (yes/no). Information about family history of colorectal cancer, history of diabetes, smoking, physical activity, and use of aspirin and multivitamin supplements was first obtained in the 1997 questionnaire. Therefore, for women we could only control for these variables in the analyses based on follow-up from 1998 through 2004.

To assess whether the association of coffee consumption with risk of colorectal cancer was modified by age, smoking, body mass index, physical activity, aspirin use, or postmenopausal hormone use, we performed analyses stratified by these variables. Statistical interaction was tested by likelihood ratio tests. We pooled the data from the two cohorts using a random-effects model for the log of the rate ratios (12). Before pooling, we conducted tests of heterogeneity using the Q statistic (12) to evaluate whether associations differed between women and men. All rate ratios are presented with 95 percent confidence intervals, and all reported p values are two-tailed.

RESULTS

Characteristics of the participants in 1997, by gender and coffee consumption, are presented in table 1. The average number of cups of coffee consumed per day was three among both women and men. In both cohorts, higher coffee consumption was associated with a greater frequency of smoking and with lower education and less multivitamin supplement use. Women and men with high coffee consumption were younger and had a higher total energy intake than persons with lower coffee consumption. The proportion of women who had used postmenopausal hormones decreased with higher coffee consumption.

During 1,240,597 person-years of follow-up of 61,433 women (1987–2004) and 45,306 men (1998–2004), 1,279 incident cases of colorectal cancer (830 among women and 449 among men) were diagnosed. Coffee consumption was not significantly associated with risk of colorectal, colon, or rectal cancer in either women or men (table 2). The crude incidence rate of colorectal cancer was 92 cases per 100,000 person-years of follow-up among participants who consumed less than one cup of coffee per day; the corresponding rate among participants who consumed four or more cups of coffee per day was the same. The pooled multivariate rate ratio for colorectal cancer for each additional cup of coffee per day was 1.00 (95 percent confidence interval (CI): 0.97, 1.04). Additional adjustment for consumption of alcohol and tea in the multivariate analyses did not alter the rate ratios shown in table 2. When we used data on coffee consumption for women obtained from the baseline questionnaire only, the findings were similar to those obtained using updated information about coffee consumption (data not shown). Coffee consumption was not associated with risk of either proximal or distal colon cancer; the pooled multivariate rate ratios for each additional cup of coffee per day were 1.01 (95 percent CI: 0.94, 1.08) for proximal colon cancer and 0.97 (95 percent CI: 0.90, 1.05) for distal colon cancer.

Pooled rate ratios for colorectal, colon, and rectal cancer according to coffee consumption in 1997 are shown in table 3. A higher level of coffee consumption (≥6 cups/day) was not significantly associated with colorectal cancer risk. The pooled rate ratio for colorectal cancer for six or more cups of coffee per day as compared with less than one cup per day was 1.04 (95 percent CI: 0.73, 1.49) with adjustment for age only and 1.06 (95 percent CI: 0.74, 1.52) after further adjustment for education, family history of colorectal cancer, history of diabetes, smoking, body mass index, physical activity, aspirin use, multivitamin supplement use, and intakes of total energy, fruit, vegetables, milk, and red meat.

To avoid potential bias due to preclinical conditions, we performed additional analyses after excluding colorectal cancer cases diagnosed during the first 2 years of follow-up. The results were essentially unchanged (data not shown). There were no significant associations of coffee consumption with colorectal cancer risk within strata of colorectal cancer risk factors, including age, smoking, body mass index, physical activity, aspirin use, and (among women) postmenopausal hormone use (table 4). None of the tests for statistical interaction between coffee consumption and these variables produced significant results in either cohort.

DISCUSSION

In this prospective analysis of data from two large population-based cohorts of Swedish women and men comprising 1,279 cases of colorectal cancer, we observed no association between consumption of coffee and risk of colorectal cancer, colon cancer, or rectal cancer. The results were similar for women and men and did not vary across levels of other risk factors for colorectal cancer.

Our null finding for coffee consumption and colorectal cancer risk is consistent with a recent report combining data from two large US cohort studies—the Nurses' Health Study and the Health Professionals Follow-up Study—including 1,433 colorectal cancer cases (8). Moreover, summary results from five other cohort studies (a total of 931 cases) showed no association of coffee consumption with risk of colorectal cancer (for high consumption vs. low consumption, rate ratio = 0.97, 95 percent CI: 0.73, 1.29) (6). In contrast, case-control studies, most of them conducted in Europe, have found that coffee consumption is associated with a lower risk of colorectal cancer (6, 7). However, results from case-control studies should be interpreted with caution because of the retrospective assessment of diet in such studies.

The prospective design of our study eliminated recall bias, which is a potential threat to the validity of case-control studies. Other strengths of this study included a population-based design, a large number of cases of colorectal cancer, information on many potential risk factors for colorectal cancer, the use of data from two completely separate cohorts, and the completeness of follow-up of the study population. In addition, by updating information on diet during follow-up of the Swedish Mammography Cohort, we were able to account for changes in coffee consumption over time and reduce random within-person measurement error. Because coffee consumption was self-reported by questionnaire, some misclassification of exposure is inevitable. Misclassification of coffee consumption may have attenuated the results to some degree. However, our validation study of the dietary questionnaire indicated that coffee consumption was reported reasonably accurately.

In the present study, we did not have data on type of coffee (i.e., whether it was filtered or boiled, decaffeinated or not). However, findings from a previous study (13) indicated that in the Stockholm area, approximately 15 percent of the population consumes boiled coffee only or both boiled and filtered coffee. Furthermore, decaffeinated coffee is very uncommon in Sweden.

In summary, in this analysis of data from two prospective cohort studies of Swedish women and men, coffee consumption was not associated with risk of colorectal cancer. This finding, along with results from a previous large prospective study of US women and men (8), suggests that consumption of filtered coffee is unlikely to have a substantial impact on colorectal cancer risk.

This study was supported by research grants from the Swedish Research Council/Longitudinal Studies, the Swedish Cancer Foundation, the Swedish Foundation for International Cooperation in Research and Higher Education (STINT), the Västmanland County Research Fund Against Cancer, the Örebro County Council Research Committee, and the Örebro Medical Center Research Foundation.

Conflict of interest: none declared.

References