点击显示 收起
1 From the Division of Nutritional Epidemiology, National Institute of Environmental Medicine, Karolinska Institutet, Stockholm, Sweden (SCL and AW); the Department of Surgery and Centre for Clinical Research, Central Hospital, Västerås, Sweden (LB); the Section of Colorectal Surgery, the Department of Surgery, University Hospital, Örebro, Sweden (JR); the Departments of Nutrition and Epidemiology, Harvard School of Public Health, Boston, Massachusetts (EG); and the Channing Laboratory, Department of Medicine, Brigham and Women's Hospital and Harvard Medical School, Boston, Massachusetts (EG)
See corresponding editorial on page 527.
See corresponding CME exam on page 728.
ABSTRACT
Background: Recent epidemiologic studies have generally reported a modest inverse association between calcium intake and the risk of colorectal cancer. However, findings pertaining to specific subsites in the colorectum have been conflicting.
Objective: Our objective was to prospectively examine the relations between intakes of calcium and dairy foods and the risk of colorectal cancer, overall and by anatomic subsite, in men from the Cohort of Swedish Men.
Design: In 1997, 45 306 men aged 45–79 y and without a history of cancer completed a food-frequency questionnaire. The men were followed through 31 December 2004.
Results: During a mean follow-up of 6.7 y, we ascertained 449 incident cases of colorectal cancer. After adjustment for age and other known or potential risk factors, the multivariate rate ratio (RR) of colorectal cancer for men in the highest quartile of total calcium intake compared with those in the lowest quartile was 0.68 (95% CI: 0.51, 0.91; P for trend = 0.01). A high consumption of dairy foods was also associated with a lower risk of colorectal cancer. The multivariate RR of colorectal cancer for 7 servings/d of total dairy foods compared with <2 servings/d was 0.46 (0.30, 0.71; P for trend = 0.01). For cancer subsites, the corresponding RRs were 0.37 (0.16, 0.88) for proximal colon, 0.43 (0.20, 0.93) for distal colon, and 0.48 (0.23, 0.99) for rectum.
Conclusion: Our findings provide support for inverse associations between intakes of calcium and dairy foods and the risk of colorectal cancer.
Key Words: Colorectal neoplasms • cohort studies • calcium • dairy products • prospective studies
INTRODUCTION
Calcium may confer protection against colorectal cancer through various plausible biological mechanisms. Newmark et al (1) postulated that calcium binds secondary bile acids and fatty acids in the colonic lumen, thereby diminishing the potential proliferative stimulus of these compounds on the colonic mucosa. Calcium may also reduce the risk of colorectal cancer by direct effects on cellular proliferation, differentiation, and apoptosis (2). Animal studies and some, although not all, clinical trials have shown that increased calcium and dairy food intakes can decrease colonic epithelial cell proliferation (3, 4). Findings from clinical trials also suggested that calcium supplementation lowers the risk of colorectal adenoma recurrence (5).
Despite experimental evidence and biological plausibility, case-control and cohort studies of calcium intake and colorectal cancer risk have yielded inconsistent results. Reviews (6, 7) and a meta-analysis (8) of epidemiologic studies published through the mid-1990s suggested little or no protective association between calcium intake and colorectal adenoma and cancer. Nevertheless, a recent pooled analysis of cohort studies showed an inverse association between calcium intake and the risk of distal colon and rectal cancers, although not of proximal colon cancer (9). Two other cohort studies that examined the association with calcium according to subsites in the colon showed no differences between the subsites (10) or found that the association was limited to proximal colon cancer (11).
As for calcium, epidemiologic studies on the association between dairy food consumption and the risk of colorectal cancer have produced mixed results. Although milk consumption was associated with a reduction in colorectal cancer risk in the Pooling Project of Prospective Studies of Diet and Cancer (9), other studies have reported only a weak, statistically nonsignificant, inverse association between total dairy food or milk consumption and colorectal cancer risk, and several studies found no association (11–13). Few studies assessed whether the association varied between subsites in the colon (9, 11). We analyzed prospective data from a cohort of Swedish men, a population with a high intake of calcium from dairy foods, to examine in detail total calcium (from foods and supplements) and dairy food intakes in relation to incident colorectal cancer, overall and by anatomic subsite.
SUBJECTS AND METHODS
The Cohort of Swedish Men
The Cohort of Swedish Men was initiated in the autumn of 1997, when all men who were aged 45–79 y and who resided in the Västmanland and Örebro counties of central Sweden received a mailed questionnaire concerning demographics, diet and other lifestyle factors, medical history, and medications. Of the 100 303 eligible men, 48 850 (49%) completed the questionnaire. The study was approved by the Regional Ethical Committee at the Karolinska Institutet (Stockholm, Sweden).
For the analysis, we excluded men with a missing or erroneous national registration number, men with implausibly high or low total energy intakes (ie, 3 SDs from the loge-transformed mean energy intake in the cohort), and men with a previous cancer diagnosis (other than nonmelanoma skin cancer) at baseline. After these exclusions, 45 306 men remained for the analysis.
Dietary assessment
Diet was assessed with a food-frequency questionnaire in which the participants reported their average frequency of consumption of 96 foods and beverages over the past year. The participants also indicated their use of calcium supplements and multivitamins according to duration and number of tablets taken per week. The food-frequency questionnaire included the following dairy foods: low-fat (0.1% and 0.5% fat) milk, medium-fat (1.5%) milk, whole (3%) milk, low-fat cultured milk (sour milk and yogurt), full-fat cultured milk (sour milk and yogurt), low-fat hard cheese, hard cheese, cottage or cream cheese, reduced-fat sour cream, full-fat sour cream, ice cream, and butter. We computed total dairy food consumption by adding the daily servings of all foods in the dairy categories except butter, because it is composed almost exclusively of fat and contains no calcium. Total milk consumption was calculated by adding the daily servings of low-fat milk, medium-fat milk, and whole milk.
Nutrient intakes were computed by multiplying the frequency of consumption of each food item by the nutrient content of age-specific portion sizes. Values for the amounts of nutrients in the foods were obtained from the Swedish Food Administration Database (14). Nutrient intakes were energy-adjusted by using residuals calculated from the linear regression of the log of nutrient intake versus the log of energy intake (15). Residuals were adjusted to 2200 kcal/d, which is the approximate mean energy intake in men in a validation study of the food-frequency questionnaire. Intakes of total calcium and total vitamin D were calculated by adding energy-adjusted nutrients from diet to unadjusted nutrients from supplements.
The food-frequency questionnaire was validated in 248 men by using fourteen 24-h recall interviews collected over a 1-y period as the comparison measure (16). The Spearman correlation coefficient (r) between the 24-h recalls and the questionnaire was 0.77 for both total and dietary calcium intake (16).
Case ascertainment and follow-up
Incident cases of colorectal cancer were identified by computerized linkage of the study cohort to the National Swedish Cancer Register (through 31 December 2003) and the Regional Cancer Register covering the study area (from 1 January 2004 through 31 December 2004). These registries have been estimated to be nearly 100% complete (17). Complementary data concerning the localization of colonic carcinomas were obtained from the regional colon cancer registry of the Uppsala-Örebro region. Only adenocarcinomas were included in the present analysis. Proximal colon cancers were defined as tumors occurring from the cecum to the splenic flexure, whereas distal colon cancers were defined as tumors occurring in the splenic flexure, descending colon, and sigmoid colon. Rectal cancers included tumors in the rectosigmoid junction and rectum. Ascertainment of deaths in the cohort and dates of moving outside Sweden or from the study area (in the last year of the study) was accomplished by linkage to the Swedish Death and Population registers at Statistics Sweden.
Statistical analysis
For each participant, follow-up time accrued from the date of entry into the cohort and ended at a colorectal cancer diagnosis, death from any cause, date of migration, or 31 December 2004, whichever came first. We categorized calcium intake into quartiles based on the distribution in the study cohort; dairy food consumption was grouped into categories. Rate ratios (RRs) were calculated as the incidence rate of colorectal cancer in men in different categories of calcium or dairy food intakes compared with the incidence rate in men in the lowest category. The data conformed to the proportional hazards assumption, and we used Cox proportional hazards models (18) to estimate RRs and 95% CIs. To control for age as much as possible, we stratified the models on age (in mo) at baseline. All multivariate models were also adjusted for education, family history of colorectal cancer, body mass index, exercise, history of diabetes, smoking, aspirin use, multivitamin supplement use, and intakes of total energy, saturated fat, total vitamin D, alcohol, fruit, vegetables, and red meat.
Tests for linear trends were conducted by assigning the median value for calcium and dairy food intake to each category and modeling this variable as a continuous variable. In addition, we used restricted cubic spline Cox proportional hazards regression with 4 knots to flexibly model total calcium intake as a continuous variable in relation to the risk of colorectal cancer, thereby avoiding the need for prior specification of the risk function or the location of a threshold exposure value (19). We used the Wald statistic (20) to evaluate whether associations differed between subsites in the colorectum (proximal colon, distal colon, and rectum). All analyses were performed with SAS statistical software (version 9.1; SAS Institute Inc, Cary, NC). All reported P values are 2-sided.
RESULTS
The baseline characteristics of the study population are shown by quartiles of total calcium intake in Table 1. Calcium supplement use was reported by 2.1% of the men. The median daily intake of total calcium ranged from 809 mg in the lowest quartile to 1665 mg in the highest quartile. Men with higher total calcium intakes were slightly older, had a higher body mass index, were more likely to have a history of diabetes, and exercised more than did men with a lower intake. Family history of colorectal cancer, cigarette smoking, aspirin use, and multivitamin supplement use did not vary appreciably across quartiles of calcium intake. Total calcium intake correlated positively with intakes of total vitamin D (r = 0.15) and saturated fat (r = 0.41) and inversely with consumption of alcohol (r = –0.18), fruit (r = –0.09), vegetables (r = –0.11), and red meat (r = –0.17).
View this table:
TABLE 1. Baseline characteristics according to quartiles of total calcium intake in 45 306 men in the Cohort of Swedish Men1
A total of 449 incident cases of colorectal cancer were ascertained during 302 931 person-years of follow-up (
View this table:
TABLE 2. Rate ratios (RRs) and 95% CIs of colorectal cancer according to total calcium intake in 45 306 men in the Cohort of Swedish Men (1998–2004)
FIGURE 1.. Multivariate rate ratios of colorectal cancer according to total calcium intake. Data were fitted with the use of a restricted cubic spline Cox proportional hazards model adjusted for age (in mo), education (less than high school, high school graduate, or more than high school), family history of colorectal cancer (yes or no), BMI (in kg/m2; <23.0, 23.0–24.9, 25.0–29.9, or 30.0), exercise (h/wk; 3 categories), history of diabetes (yes or no), cigarette smoking (never, past, or current), aspirin use (no, <10 y, or 10 y), multivitamin supplement use (no, occasional use, or regular use), total energy intake (continuous), and quartiles of saturated fat, total vitamin D, alcohol, fruit, vegetable, and red meat intakes. Participants with total calcium intakes >99th percentile of the distribution were excluded to avoid excessive influence of extreme intakes.
To minimize a potential bias from preclinical disease, we conducted additional analyses after excluding colorectal cancer cases that occurred during the first 2-y follow-up period. The results of the analysis, which included 332 remaining colorectal cancer cases, were somewhat stronger than those observed with all cases (multivariate RR for the highest compared with the lowest quartile of total calcium intake: 0.61; 95% CI: 0.44, 0.85; P for trend = 0.005).
Dietary calcium intake (ie, calcium obtained from foods) was also inversely associated with the risk of colorectal cancer. After excluding users of calcium supplements, the multivariate RRs (95% CIs) across increasing quartiles of dietary calcium intake were 0.83 (0.64, 1.08), 0.73 (0.55, 0.97), and 0.67 (0.50, 0.90) compared with the lowest quartile. We could not perform meaningful analyses of supplemental calcium intake in relation to colorectal cancer risk because of the low prevalence of calcium supplement use (2.1%) in the study population.
We additionally evaluated the consumption of total dairy foods and of specific dairy foods in relation to the risk of colorectal cancer (Table 3). The RR of colorectal cancer was statistically significantly reduced for men in the higher quartiles of total dairy food consumption compared with those who consumed <2 servings/d. When we dichotomized total dairy food consumption into <2 servings/d and 2 servings/d, the multivariate RR in a comparison of the 2 categories was 0.56 (95% CI: 0.40, 0.77). The association between total dairy food consumption and colorectal cancer risk did not differ significantly between subsites (P for heterogeneity = 0.91 for the highest category of dairy food consumption). Most individual dairy foods showed inverse associations with the risk of colorectal cancer, but only the results for milk consumption were statistically significant (Table 3). Men who consumed 1.5 glasses of milk/d had a 33% lower risk of colorectal cancer than did those who consumed <2 glasses/wk. When we simultaneously included low-fat, medium-fat, and whole milk in the multivariate model, the RRs (95% CIs) in a comparison of those who consumed 1.5 glasses of milk/d with those who never consumed milk were 0.78 (0.57, 1.07) for low-fat milk, 0.65 (0.47, 0.91) for medium-fat milk, and 0.65 (0.44, 0.99) for whole milk.
View this table:
TABLE 3. Multivariate rate ratios (RRs) and 95% CIs of colorectal cancer according to categories of dairy food consumption in 45 306 men in the Cohort of Swedish Men (1998–2004)1
To examine whether the inverse associations of dairy foods and milk consumption with the risk of colorectal cancer could be attributed to calcium, we entered total calcium intake and either total dairy foods or milk consumption simultaneously into a multivariate model (r between total calcium and total dairy foods = 0.68; r between total calcium and milk = 0.50). After control for total calcium intake, the multivariate RRs (95% CIs) for the highest compared with the lowest category of milk consumption were attenuated from 0.46 to 0.58 (0.31, 1.06) for total dairy foods and from 0.67 to 0.75 (0.56, 1.00) for total milk consumption. Similarly, the multivariate RR in a comparison of the 2 extreme quartiles of total calcium intake was attenuated from 0.68 to 0.76 (95% CI: 0.55, 1.05) when we added milk consumption to the multivariate model.
DISCUSSION
In this population-based prospective cohort of men, higher intakes of calcium and dairy foods, in particular milk, were associated with a significantly lower risk of colorectal cancer, independent of other known risk factors. Possible threshold effects of total calcium intake and milk consumption were suggested by regression splines. The associations did not vary significantly by subsite in the colorectum.
Although observational epidemiologic studies published through the mid-1990s provided no strong support for a protective association between calcium intake and colorectal cancer risk, an inverse association (risk reductions in the range of 15% to 40% for the highest compared with the lowest intake categories) has been observed in several published cohort studies (9–11, 21) and in a large case-control study (22). A recent cohort study conducted in women (13), however, found no such association. Studies that showed no significant inverse association with calcium were generally limited by a relatively small sample size and were thus underpowered to detect risk reductions in the range of 15% to 40%. In fact, earlier studies did have nonsignificant associations consistent with this range (7, 8). Few cohort studies have examined cancer risk by subsites in the colorectum, and the results have been conflicting. In the Pooling Project of Prospective Studies of Diet and Cancer (9), the inverse association with calcium intake was limited to cancers of the distal colon and rectum (9). In contrast, in the Cancer Prevention Study II Nutrition Cohort (11), a high total calcium intake was related to a reduced risk of proximal colon cancer, but not of distal colon and rectal cancer. Another cohort study (10) reported similar inverse associations for proximal and distal colon cancer but observed no effect in the rectum.
In our study, milk was the dairy food that was most strongly inversely associated with the risk of colorectal cancer. This finding is consistent with the results from the Pooling Project (9), in which persons in the highest category of milk consumption (250 g/d; ie, 1 glass/d) had a 15% lower risk of colorectal cancer than did those in the lowest category of milk consumption (<70 g/d). A meta-analysis of case-control and cohort studies (including a few of the studies in the Pooling Project) reported an inverse association with milk consumption for the cohort studies (20% reduction in risk for the high compared with the low consumption group) but not for the case-control studies (12). In this same meta-analysis, the RR for high compared with low total dairy food consumption was 0.62 (95% CI: 0.52, 0.74) for cohort studies (12), which is consistent with our findings. As in our study, the Pooling Project (9) and the meta-analysis (12) showed no significant association of yogurt or cheese consumption with colorectal cancer risk. A potential explanation for the lack of significant association with cheese and yogurt may be that these foods are not important sources of dietary calcium intake. National surveys in Sweden (23) and in the United States (24) have reported that milk is the most important contributor to dietary calcium intake.
Our finding of a possible threshold effect of calcium intake on the risk of colorectal cancer agrees with the results from the Pooling Project (9), although the threshold value was located at slightly different calcium intakes (1400 mg/d in our study and 1000 mg/d in the Pooling Project). A possible reason why the optimal intake varied between the 2 studies may be because calcium intake was adjusted to different energy intakes; in our study, calcium intake was energy-adjusted to 2200 kcal/d, whereas it was adjusted to 1600 kcal/d for women and 2100 kcal/d for men in the Pooling Project. Another potential explanation is that the dietary questionnaires had different degrees of error in the measurement of calcium intake.
Dairy foods contributed most to total calcium intake in our study population. In addition to calcium, dairy foods contain conjugated linoleic acid, sphingolipids, and milk proteins (including lactoferrin), which have been shown to inhibit colorectal carcinogenesis in animal models (25–28). When we additionally adjusted for total calcium intake, our results suggested that the association between the consumption of total dairy foods and milk and the risk of colorectal cancer was only partly explained by calcium intake, although adjustment for calcium intake was somewhat limited by its close correlation with dairy food consumption. Results from the Pooling Project (9) also showed that the inverse relation with milk consumption persisted after adjustment for total calcium intake. In our study and in the Pooling Project, the inverse association of total calcium intake with colorectal cancer risk also remained (although it was not statistically significant) after control for milk consumption, which suggests an independent effect of calcium on colorectal carcinogenesis. Additional support for a role of calcium in reducing colorectal cancer risk comes from randomized clinical trials (5) and from observational epidemiologic studies conducted in the United States (10, 11, 22, 29), which indicate that supplemental calcium may reduce the risk of colorectal adenoma recurrence and colorectal cancer.
The strengths of our study include a population-based and prospective design and detailed information on diet. The prospective design and practically complete cohort follow-up minimized the possibility of recall bias and bias caused by loss to follow-up. The extensive data on known and possible risk factors for colorectal cancer allowed comprehensive adjustment of confounders. Although the assessment of calcium intake obtained through the dietary questionnaire in the present study had high accuracy (16), some misclassification of calcium intake is inevitable. However, any measurement error in assessing dietary intake in a prospective setting would not be related to the outcome, and nondifferential misclassification would have attenuated rather than exaggerated true associations. We cannot rule out the possibility that some unmeasured confounders accounted for the observed associations, but residual confounding by the covariates that we adjusted for is unlikely given the similar results in age-adjusted and multivariate analyses.
In conclusion, the findings from this prospective cohort study of men provide support for inverse associations between intakes of calcium and dairy foods and colorectal cancer risk. Future studies should examine the relation of other components of dairy foods, such as conjugated linoleic acid, sphingolipids, and milk proteins, with the risk of colorectal cancer.
ACKNOWLEDGMENTS
SCL and AW were responsible for the study concept and design; AW was responsible for data collection; SCL was responsible for the statistical analyses and the writing of the manuscript writing. SCL, LB, JR, EG, and AW were responsible for interpreting the results and for critical revision of the manuscript. All authors reviewed the final manuscript. None of the authors had any financial or personal conflicts of interest.
REFERENCES
Related articles in AJCN: