点击显示 收起
1 From the Health Research Center, University of Utah, Salt Lake City (MLS, KPC, and SLE), and the Kaiser Permanente Medical Research Program, Oakland, CA (DMS).
2 The contents of this article are solely the responsibility of the authors and do not necessarily represent the official view of the National Cancer Institute. 3 Supported by grant CA48998 from the National Cancer Institute (to MLS); the Utah Cancer Registry, which is funded by contract no. N01-PC-67000 from the National Cancer Institute; the State of Utah Department of Health; the Northern California Cancer Registry; and the Sacramento Tumor Registry. 4 Reprints not available. Address correspondence to ML Slattery, Health Research Center, University of Utah, 375 Chipeta Way, Suite A, Salt Lake City, UT 84108. E-mail: mslatter{at}hrc.utah.edu.
ABSTRACT
Background: Associations between colon and rectal cancer and intakes of vegetables, other plant foods, and fiber have stimulated much debate.
Objective: We examined the association between rectal cancer and plant food and fiber intakes.
Design: Data from 952 incident cases of rectal cancer were compared with data from 1205 population-based controls living in Utah or enrolled in the Kaiser Permanente Medical Care Program in northern California
Results: Rectal cancer was inversely associated with intakes of vegetables (odds ratio: 0.72; 95% CI: 0.54, 0.98), fruit (0.73; 0.53, 0.99), and whole-grain products (0.69; 0.51, 0.94), whereas a high intake of refined-grain products was directly associated with an increased risk of rectal cancer (1.42; 1.04, 1.92). Similarly, relative to low fiber intakes, high intakes of dietary fiber reduced the risk of rectal cancer (0.54; 0.37, 0.78). The reduced risk of rectal cancer associated with vegetable (0.48; 0.29, 0.80), fruit (0.63; 0.38, 1.06), and fiber (0.40; 0.22, 0.71) intakes was strongest for persons who received the diagnosis after age 65 y. A threshold effect at 5 servings of vegetables/d was needed to see a reduced risk of rectal cancer.
Conclusions: The results suggest that plant foods may be important in the etiology of rectal cancer in both men and women. Age at diagnosis appears to play an important role in the association.
Key Words: Age • colorectal cancer • fiber • fruit • grain products • rectal cancer • vegetables
INTRODUCTION
The debate over the associations between cancer and plant food and fiber intakes is ongoing. Central to the debate is the association between colorectal cancer and plant foods. Most studies of colorectal cancer have evaluated associations for colon cancer or colorectal cancer, whereas few studies have examined associations with rectal cancer (1–7). Many studies of colon cancer have shown inverse associations with intakes of plant foods and fiber, although other studies have not (1–7). Of the plant foods for which associations with colon cancer have been examined, vegetable consumption has been most consistently associated with a reduced risk, whereas the association with fruit intake has been less consistent (1, 7). Although the association between colon cancer and the intake of grain products has been reported less frequently than that for either vegetables or fruit, whole-grain products have been shown to reduce the risk of colon cancer, whereas refined-grain products have been shown to increase the risk of colon cancer; the association differed for proximal and distal tumors within the colon (2, 7).
The inconsistencies regarding the association between plant food and fiber intakes and colorectal cancer could stem from several factors, including age at diagnosis, tumor site, and dose needed to observe an association—factors that could be influenced by study design. The findings of our previously reported population-based case-control study support the association between plant food intake and colon cancer but suggest that age and tumor site may be important modulators of risk (2). Whereas few studies have had adequate power to examine proximal and distal tumors; even fewer have had the ability to examine associations with rectal cancer. Studies that have focused on dietary associations with rectal cancer have generally been limited to a few cases (1, 3–7), which makes estimates of associations imprecise. In this study we evaluated the associations between plant food intakes and rectal cancer and addressed the issues of age and dose on the associations.
SUBJECTS AND METHODS
Study population
Participants in the study were from the Kaiser Permanente Medical Care Program of Northern California (KPMCP) and the state of Utah. The study protocol was approved by the University of Utah Institutional Review Board. All eligible cases within these defined populations were identified and recruited for the study. Cases with a first primary tumor in the recto-sigmoid junction or rectum were identified between May 1997 and May 2001 with the use of a rapid-reporting system. Case eligibility was determined by the Surveillance Epidemiology and End Results Cancer Registries in northern California and Utah. Cases were not eligible for the study if they had had a previous colorectal tumor or known (as indicated on the pathology report) familial adenomatous polyposis, ulcerative colitis, or Crohn disease. In addition to these eligibility criteria, participants were between 30 and 79 y of age at the time of diagnosis, English speaking, and mentally competent to complete the interview.
Controls were categorically matched to cases by sex, 5-y age groups, and center. At the KPMCP, controls were randomly selected from membership lists. In Utah, controls aged =" BORDER="0">65 y were randomly selected from Social Security lists, and controls aged <65 y were randomly selected from driver's license lists.
The race or ethnicity of the rectal cancer study population was reported at the time of interview as 82% white, non-Hispanic; 4.1% African American; 7.6% Hispanic; 4.6% Asian; 0.7% American Indian; and 1% multiple. The ethnicity makeup of the cases and controls were the same. A total of 952 rectal cancer cases and 1205 matched controls were included in the analyses. Response rates were 65.2% for cases and 65.3% for controls; cooperation rates (ie, the percentage of the population that we contacted who participated) were 73.2% for cases and 68.8% for controls. More than 95% of the cases and controls identified were contacted (8).
Data collection
Data were collected on laptop computers through in-person interviews conducted by trained and certified interviewers. Data for the rectal cancer portion of the study were collected by using the same study questionnaire and the same quality-control procedures as were used in the colon cancer study. Study participants were asked to recall the year 2 y before the date of selection (the date of diagnosis for cases or date of selection for controls). The interview took 2 h. Quality-control methods used in the study were the same as those used in the colon cancer study and were described in detail previously (9, 10).
Dietary data
Dietary intake was ascertained by using an adaptation of the CARDIA diet history (10, 11). Participants were asked to recall foods eaten, the frequency at which they were eaten, the serving size, and whether fats were added in the preparation. Nutrient information was obtained by converting food intake data into nutrient data by using the Minnesota Nutrition Coding Center nutrient database. The diet-history questionnaire used allowed us to evaluate the intake of many individual plant foods. There were 25 fresh fruits; 16 canned, cooked, or frozen fruits; and 8 dried fruits listed on the questionnaire along with 14 fruit and vegetable juices. Additionally, there were 58 cooked and raw vegetables that could be reported. Standard serving sizes of vegetables were as follows:
Height and weight were measured at the time of the interview, and weight was reported for 2 and 5 y before the interview. The body mass index [wt (kg)/ht2 (m)] was calculated for men and women. Information on smoking history, medical history, family history of cancer, physical activity patterns, and use of aspirin and nonsteroidal antiinflammatory drugs was also collected. Physical activity patterns were determined from a detailed questionnaire that asked about the frequency and intensity of activities performed during the referent period and 10 and 20 y before the referent date (8).
Statistical methods
The SAS statistical package (version 8.2; SAS Institute Inc, Cary, NC) was used to conduct the analysis. Differences in mean intakes of plant foods were assessed by using analysis of covariance models with adjustment for age and sex distribution in the population. Unconditional logistic regression models were used to estimate the risk of rectal cancer from plant food intake. In these models the following variables were included as potentially confounding factors: age at selection, body mass index, recalled family history of colorectal cancer in first-degree relatives, long-term vigorous physical activity, energy intake, dietary calcium, and regular use of aspirin or nonsteroidal antiinflammatory drugs; adjustment for dietary fat intake did not alter these associations. All adjustment variables were treated as continuous variables in the model, except for family history and use of nonsteroidal antiinflammatory drugs, which were categorized as dichotomous variables. The exposure variables plant food and fiber intakes were categorized. Dietary fiber was divided into quintiles on the basis of the control distribution. Intake of foods was based on practical serving sizes as well as population distribution. For comparative purposes and for the evaluation of the shape of the dose-response curve between fruit and vegetable intakes and risk of colon and rectal cancers, associations for tumors in the rectum were compared with those in the colon by using data collected from the same target populations in Utah and at the KPMCP. Age-specific analyses were performed with an age cutoff of 65 y. Linear trend was determined by evaluating the significance of linear associations across categorized variables. To test for a threshold effect when a linear association was not observed, we used chi-square tests to examine differences across categories.
RESULTS
The study population is described in Table 1; 36% of the cases were from Utah, 28% were <55 y of age, and 59% were male. No significant differences in level of education were observed between cases and controls; 65% reported some college education. An examination of mean intakes of plant foods showed that controls in the rectal cancer study consumed more servings of carrots, salad, fruit, fresh fruit, and whole-grain products and fewer servings of refined-grain products than did cases in the rectal cancer study (Table 2).
View this table:
TABLE 1. . Description of the study population
View this table:
TABLE 2. . Average daily servings of plant foods for rectal cancer cases and controls and colon cancer controls1
High vegetable intakes were associated with a 28% reduction in the risk of rectal cancer in both sexes combined (Table 3). Assessment of specific types of vegetables did not indicate that any specific types of vegetables influenced risk differently (data not shown). High fruit intakes were significantly protective (odds ratio: 0.73; 95% CI: 0.53, 0.99), relative to the lowest intake, when data from men and women were combined, although specific types of fruit did not appear to confer unique protection. A threshold effect was observed for fruit intake, with a significant P value (P = 0.02; chi-square test). There were no significant differences in either vegetable or fruit intake between the sexes. Legume intake did not alter the risk of rectal cancer. Refined-grain products were associated with an increased risk of 40% overall, with a P value (chi-square test) for the threshold effect of <0.01. There was also a significant interaction between sex and refined grain intake (P = 0.05). Assessment of high-fat, refined-grain products showed that, among women, there was an increased risk at the highest intake (odds ratio: 1.70; 95% CI: 0.96, 3.02) and a significant threshold effect (P 0.01); the P value for the interaction between sex and high-fat refined grain intake was 0.06. Intake of whole-grain products was associated with a 30% reduced risk of rectal cancer for both men and women combined; the interaction between intake of whole-grain products and sex was not significant (P = 0.27), and the P value for the threshold effect was 0.02. We did not assess intake of high-fat whole-grain products because almost all those who reported consuming whole-grain products reported consuming whole-grain products that were neither high in fat nor high in sugar. Increasing fiber intake showed a strong protective effect for both men and women combined (Table 4).
View this table:
TABLE 3. . Associations between plant food intakes (servings/d) and rectal cancer in men and women
View this table:
TABLE 4. . Associations between fiber intake (g/d) and rectal cancer in men and women
Intakes of vegetables, fruit, and fiber were more strongly associated with older age at diagnosis (Table 5). There was no significant interaction between age and intakes of nuts, refined grains (including those high in fat), and whole-grain products. Intakes of high-fat, refined-grain products appeared to be associated with a greater risk in those with a diagnosis at a younger age (P for interaction = 0.03). There was no indication of risk for either younger or older participants relative to legume intake (data not shown).
View this table:
TABLE 5. . Associations between plant food (servings/d) and fiber intakes (g/d) and rectal cancer by age at diagnosis
DISCUSSION
The findings from this study suggest that intake of plant foods, especially vegetables and whole-grain products, reduces the risk of rectal cancer. Dietary fiber was also shown to have a strong protective effect for rectal tumors. On the basis of a threshold effect, intake of refined-grain products—especially those with a high-fat content—increases the risk of rectal cancer. Most other associations were consistent for men and women. Age appeared to be an important modulator of risk for all plant food intakes. Intakes of vegetables, fruit, and fiber—especially of insoluble fiber—had a strong effect in older persons. Support for our findings between plant food intake and rectal cancer risk is mixed, although few studies have had adequate power to examine dietary associations with rectal cancer specifically.
Our results concerning the association between vegetable and fiber intakes and rectal cancer risk are not different from those that we observed for colon cancer using the same methods in the same target population (2). In both studies we observed a 30% reduction in risk at =" BORDER="0">5 servings of vegetables/d. In men in our previous colon cancer study (2), we observed the strongest inverse associations with vegetable intake in older men; fiber was more strongly associated with colon cancer in both older men and women than in those who were younger at the time of diagnosis. Of interest is our previous observation that vegetable and fiber intakes are more protective against proximal tumors than against distal tumors (2). It is possible that even though the associations were similar for colorectal subsites, the mechanisms differed. The associations with proximal tumors may imply a mechanism involving insulin, because it has been hypothesized that proximal tumors are more associated with endogenous factors than are distal tumors (12), whereas transit time may be a more important mechanism for rectal tumors.
Several studies that have included mostly older persons have obtained results similar to those of the current study. For instance, the association between =" BORDER="0">4 servings of vegetables/d compared with <2 servings/d and colorectal cancer in the women from Iowa—a cohort of older women—was 0.73 (95% CI: 0.47, 1.13) (4); the association between vegetable intake and colorectal cancer in elderly women in the Leisure World Study was 0.72 (95% CI: 0.45, 1.16) (13); in a cohort study conducted in the Netherlands, an inverse association was observed between the upper and lower quintiles of fruit and vegetable intake (odds ratio: 0.66; 95% CI: 0.44, 1.01) (6). In the Health Professionals' Follow-up Study, the risk estimate for =" BORDER="0">6 servings of vegetables/d relative to 2 servings of vegetables/d for rectal cancer (n = 89 cases) was 0.66; a reduction in the risk of rectal cancer was not observed on the basis of data from the Nurses Health Study (n = 155 cases). As noted, many of the associations were not statistically significant; however, estimates of risk were almost identical to our significant point risk of 0.72. Although point estimates may differ in their level of significance because of power issues, data from cohort studies in older persons support the 30% reduction in risk reported here. One study that examined rectal cancer that did not report a similar magnitude of association did not adjust for dietary and activity factors simultaneously (14).
It is likely, given the age effects that we observed, that some of the differences in associations reported in the literature stem from differences in age at diagnosis. Many cohorts include younger persons, whose mean ages were considerably less than those reported here (5). It is possible, if the age effect observed in our data exist in other populations, that studies primarily of younger persons may miss associations that may be important for older populations.
One potential problem in case-control studies is the retrospective recall of diet. In this study we used an interviewer-administered questionnaire that was very detailed, which provided a complete dietary assessment. Arguing against bias from retrospective recall are the similarities in associations for many dietary factors, including sugar, calcium, and folate intakes, observed in our colon cancer study with those reported in large cohort studies (13, 15–19). If recall bias exists, it implies that selective parts of the diet are recalled differently, either because of publicity or because of selective changes in diet resulting from actual dietary changes.
Several dietary factors could fall into the highly publicized risk factors, including fruit, vegetable, and meat intakes. However, one of the major areas of difference in reported associations with colon cancer by study design is for meat intake, with some, but not all cohort studies reporting stronger associations than case-control studies (20–23), whereas the opposite is true for vegetable intake. These differences in associations detected by study design may have important biological meaning that may have to do with the timing of exposure, which is important in the carcinogenic process; with some component of meat or vegetable intake that is important; or with diet and other concomitant lifestyle changes that may influence the risk of these factors.
Arguing against the association between vegetable, fruit, wheat bran, and fiber intakes and colorectal cancer risk are the results from several controlled trials (24, 25). For instance, the Polyp Prevention Trial did not observe a reduced risk in polyp recurrence with a diet high in vegetables, fruit, and fiber (25). This may indicate more about mechanisms and study design than about risk itself, given the short period of follow-up and the number of persons with small adenomas that were included in the study. It is possible that vegetables and fiber do not work at the initiation stage but at a promoter stage of carcinogenesis. It also is possible that most participants were younger, that it was not an effective invention for polyps recurring at a younger age, or that, although polyps are precursor lesions to tumors, not all polyps go on to become tumors. One could summarize that there are limitations to our knowledge that make designing a clinical trial to study polyps as a surrogate for cancer difficult and the results from these trials less meaningful if significant associations are not detected.
Given the strong association with fiber, the primary biological mechanism of vegetables and whole-grain products in reducing the risk of rectal cancer may be through their contribution to fiber. Whole-grain products contribute primarily insoluble fiber, whereas vegetables and fruit contribute both soluble and insoluble fiber. Dietary fiber can decrease transit time and increase fecal weight, which results in less carcinogen contact with the colonic mucosa (26). However, given the lack of association with polyp recurrence, action in the tumor-promoter stage of the disease process may be a more reasonable mechanism. Dietary fiber, especially soluble fiber, could enhance removal from the body of secondary bile acids, which are thought to be tumor promoters. Dietary fiber, especially soluble dietary fiber, can be degraded to form short-chain fatty acids. Short-chain fatty acids inhibit growth and induce differentiation and apoptosis of colon tumor cells (26). Additionally, increased short-chain fatty acid concentrations result in a concomitant decrease in fecal pH that influences the conversion of bile acids, which are effective tumor promoters. Soluble dietary fiber has been shown to modulate glycemia and insulinemia (26); it is possible that fiber could influence colorectal cancer risk through this mechanism also. Given the strongest associations with vegetables in older persons, it is possible that long-term exposure to these factors may be important. Because all types of fiber appear to be protective, it is likely that the association could stem from multiple mechanisms at work.
In addition to the potential difficulties in recall, testing for the threshold effect was done post hoc, given the observed associations. Because testing for the threshold effect was post hoc, a Bonferroni correction for 4 tests (ie, each of the top quartiles compared with the bottom quartile) could be considered. Exact P values are reported; however, because P values are generally <0.01, a significant association remains after the correction is considered.
In summary, these data support the association between high intakes of vegetables and whole-grain products as dietary factors that reduce the risk of rectal cancer, whereas high intakes of refined-grain products increase the risk of rectal cancer. These associations were observed in both men and women and appear to differ by age at the time of diagnosis. The protective effect of plant foods, especially vegetables, appears to be a threshold effect that occurs at =" BORDER="0">5 servings/d. Given the associations detected, it is of interest to identify biological mechanisms that explain these observations and to determine whether specific tumor mutations are linked to these dietary factors.
ACKNOWLEDGMENTS
We acknowledge Joan Benson for help with data collection and Richard Holubkov, KN Ma, Roger Edwards, Leslie Palmer, and Judy Morse for their support with the collection and analysis of the data. We also thank the many study interviewers.
SLE and DMS contributed to the collection and quality control of the data. KPC contributed to data management. MLS contributed to all aspects of the study, including the data analysis and preparation of the manuscript. None of the authors had any financial or personal interest in any company or organization associated with the research.
REFERENCES