Calcium, Vitamin D, Dairy Products, and Mortality Among Colorectal Cancer Survivors: The Cancer Prevention Study-II Nutrition Cohort

The overall 5-year relative survival for colorectal cancer is 64% in the United States but decreases to 12% for distant metastatic disease.¹ The associations of dietary factors with colorectal cancer incidence have been extensively reported,² but their roles for colorectal cancer survival are largely unknown.³ Current dietary guidelines for cancer survivors are primarily based on incidence studies.⁴ Empirical knowledge of modifiable prognostic factors, including diet, for patients with colorectal cancer is needed for the more than 3.5 million colorectal cancer survivors worldwide.⁵

Higher intakes of calcium, vitamin D, and dairy products are generally associated with lower risk of colorectal cancer incidence in observational studies.^6,7 In addition, a major randomized, clinical trial of 1,200 mg of supplemental calcium versus placebo among 930 colorectal adenoma patients reported a 19% reduced risk of adenoma recurrence.⁸ In contrast, the Women's Health Initiative clinical trial reported no effect of calcium plus vitamin D supplementation on colorectal cancer incidence,^9–11 but suggestive benefits were observed among those not taking personal calcium or vitamin D supplements¹² and those not concurrently randomly assigned to estrogen therapies.¹³ Two studies reported null associations of prediagnostic calcium intake with colorectal cancer survival.^14,15 The main circulating biomarker of vitamin D, 25(OH)D, was associated with lower risk of mortality among patients with colorectal cancer.^16–19 To our knowledge, no study has examined whether total dairy or milk are associated with survival among patients with colorectal cancer. We investigated associations of pre- and postdiagnosis calcium (total, dietary, and supplemental), vitamin D (total and dietary), and dairy product (total and milk only) intakes with all-cause and colorectal cancer–specific mortality in a prospective study of men and women diagnosed with invasive, nonmetastatic colorectal cancer.

Participants were, on average, age 64 years at baseline and 73 years at diagnosis. Fifty-six percent of participants were men, and most reported their race as white. There were no differences across quartiles of prediagnostic total calcium intake in the distributions of year of diagnosis, sex, tumor stage, grade, sub-site, treatment, or history of hypertension, myocardial infarction, diabetes, and stroke (Table 1). High calcium consumers were slightly older, better educated, more physically active, leaner, more likely to use nonsteroidal anti-inflammatory drugs and postmenopausal hormones (women only), less likely to smoke, and more likely to have a healthier overall diet.

Among the 2,284 patients included in the prediagnosis analyses, 949 deaths occurred (408 from colon or rectal cancer) during a mean follow-up of 7.5 years (standard deviation, 4.6 years; range, 2 days to 18.1 years). No statistically significant associations were observed for any of the prediagnosis diet variables with any of the mortality outcomes (Table 2 and Appendix Table A2). The results were not meaningfully different after additional adjusting for other covariates or after additional sensitivity analyses (data not shown). The results were also null after we included patients with metastatic or unknown tumor stage (Appendix Table A3). In analyses restricted to the 1,111 participants who were included in the postdiagnosis analyses, prediagnosis use of supplemental calcium ≥ 250 mg/d was statistically significantly associated with higher risk of all-cause mortality (RR, 1.65; 95% CI, 1.16 to 2.35; Appendix Table A4); this risk was primarily because of an increased RR for cardiovascular disease mortality (RR, 1.83; 95% CI, 0.82 to 4.09).

Among the 1,111 patients included in the postdiagnosis analyses, 429 deaths occurred (143 from colon or rectal cancer) during a mean follow-up of 7.6 years (standard deviation, 3.4 years; range, 20 days to 11.3 years). The mean time between diagnosis and completion of the postdiagnosis questionnaire was 2.6 years. As shown in Table 3, comparing the highest to the lowest quartiles, total calcium (RR, 0.72; 95% CI, 0.53 to 0.98; P_trend = .02) and milk (RR, 0.72; 95% CI, 0.55 to 0.94; P_trend = .02) intakes were associated with lower all-cause mortality. Additional adjustment for prediagnostic total calcium and milk had no discernible effect on the results. A marginally statistically significant inverse association with all-cause mortality was observed for total dairy (RR, 0.75; 95% CI, 0.56 to 1.01; P_trend = .05). Total calcium was also inversely associated with colorectal cancer–specific mortality (highest v lowest quartile RR, 0.59; 95% CI, 0.33 to 1.05; P_trend = .01; Appendix Table A5).

Because postdiagnosis diet and supplement use may be influenced by serious illness preceding death (reverse causation), we conducted a sensitivity analysis excluding deaths within the first 2 years of follow-up after completion of the postdiagnosis questionnaire. The results after this exclusion seemed similar to the original results. The RRs for the highest compared with the lowest quartile of total calcium and milk, respectively, were 0.69 (95% CI, 0.48 to 0.98; P_trend = .03) and 0.68 (95% CI, 0.50 to 0.93; P_trend = .02) for all-cause mortality; and 0.53 (95% CI, 0.24 to 1.19; P_trend = .02) for total calcium and colorectal cancer–specific mortality.

There was no evidence that the inverse associations of postdiagnosis total calcium and milk intakes with all-cause mortality were modified by age at diagnosis (< 70 years v ≥ 70 years), sex, tumor stage (localized v regional), tumor sub-site (colon v rectum), postdiagnosis body mass index (obese v not obese), physical activity (< median v ≥ median), total energy (< median v ≥ median), or total folate (< median v ≥ median) intakes (results stratified by stage shown in Appendix Table A6; other data not shown).

This study suggests that higher intakes of total calcium and milk after colorectal cancer diagnosis are associated with lower risk of mortality. These associations persisted after adjusting for important covariates, such as sex and tumor stage, and after several sensitivity analyses. We found no evidence that calcium, vitamin D, or dairy product intakes before colorectal cancer diagnosis were associated with mortality. To our knowledge, this is the first study to report associations of dairy and milk (both pre- and postdiagnosis) with colorectal cancer survival, and also the first to assess the role of postdiagnosis calcium and vitamin D intakes.

Calcium, 25(OH)D (the major circulating form of vitamin D), and dairy products are associated with lower risk of incident colorectal cancer based on several meta-analyses.^6,7,31 An earlier CPS-II Nutrition Cohort study reported inverse associations of colorectal cancer incidence with total calcium and total vitamin D intakes.³² The World Cancer Research Fund and American Institute for Cancer Research Continuous Update Project³³ concluded in 2011 that calcium and milk were both probable factors associated with lower colorectal cancer risk.

In contrast to the substantial evidence of a role for calcium, vitamin D, and dairy products in colorectal cancer primary prevention, the role of these factors in colorectal cancer survival is less studied.³ In two cohort studies, prediagnosis dietary calcium intake was not associated with all-cause mortality among patients with colorectal cancer, consistent with our findings.^14,15 25(OH)D, either pre- or postdiagnosis, was associated with longer colorectal cancer survival in four previous studies^16–19; in the current study, we observed no association with dietary vitamin D intake, which may not optimally reflect serum vitamin D status. In a large, pooled analysis, vitamin D intake was positively associated with serum 25(OH)D level, but the associations were relatively weak (Spearman correlation was 0.22 for dietary vitamin D and 0.29 for total vitamin D).³⁴

In the current study, we found a statistically significant lower risk of death among patients with higher postdiagnostic intakes of total calcium and milk. Although not completely understood, several possible biologic mechanisms might underlie these associations. Clinical trials conducted among patients with previous colorectal adenoma suggested that daily treatment with calcium, compared with placebo, was associated with lower risk of colorectal adenoma recurrence.^8,35,36 Potential mechanisms include calcium's ability to bind to bile and fatty acids and prevent or lower toxicity^37–39; direct effects on colonocyte proliferation,^40,41 differentiation,⁴² and apoptosis⁴³; and, alterations in K-ras mutations.⁴⁴ Although these mechanisms were originally proposed in the primary prevention context, it is reasonable to hypothesize that calcium may also act through these mechanisms after diagnosis to reduce the risk of cancer recurrence, thus ultimately improving chance of survival. Direct clinical or epidemiologic evidence of calcium in colorectal cancer progression is limited, but in vitro evidence suggests that calcium may promote E-cadherin expression and suppress β-catenin/T-cell factor activation through the calcium sensing receptor, and restrain their malignant behaviors.⁴⁵ Thus, calcium may be capable of limiting growth and distant metastasis from cancer cells that escaped the colon at the time of treatment. In our data, the strong inverse association of postdiagnosis total calcium intake with colorectal cancer–specific mortality was consistent with these mechanisms.

Whereas postdiagnosis calcium intake was associated with lower risk of all-cause and colorectal cancer–specific mortality, there were no such associations with prediagnosis diet. Reasons for these discrepant findings are unclear. It is possible that calcium may have short-term rather than long-term effects on colorectal cancer progression and survival, and therefore only postdiagnosis diet is relevant in this context. It is also important to note that different FFQs were used to assess pre- and postdiagnosis diet. The correlation coefficient (Pearson) between pre- and postdiagnosis total calcium intake was 0.37: this moderate correlation might suggest that participants changed their calcium intake after cancer diagnosis or, alternatively, this could reflect differences in the dietary assessment instruments. The FFQs used in this study included the major food and beverage sources of calcium and vitamin D, and validation studies have shown good agreements between estimates from diet recall and these FFQs (eg, Pearson correlation coefficients ranged from 0.57 to 0.66 for calcium and from 0.52 to 0.88 for dairy products).^27,29,30 Therefore, we believe that the low correlations between pre- and postdiagnosis diet are more likely resulting from real changes in diet after cancer diagnosis.

We observed a potential higher risk of all-cause mortality from prediagnosis supplemental calcium intake, especially when restricting the analysis to the 1,111 participants who were included in the postdiagnosis analyses. Additional research should address whether this is a real potential harm to colorectal cancer patients.

Milk may be associated with improved survival among patients with colorectal cancer because it is a rich source of dietary calcium and vitamin D. In addition, milk is a primary dietary source of conjugated linoleic acid, which was found to inhibit colorectal cancer cell growth in vitro.^46,47 Other potentially beneficial components in dairy products include butyric acid, lactoferrin, and fermentation products.⁴⁶

The strengths of our study include its large sample size, prospective design, and detailed pre- and postdiagnosis questionnaire information. We were also able to examine cause-specific mortality. Limitations include the lack of information on adverse effects from treatment and tumor recurrence. FFQs may underestimate diet-disease associations compared with more objective biomarker measurements because of nondifferential misclassification. For large cohort studies, however, FFQs offer a feasible method to detect potential associations (especially when using energy-adjusted nutrients) in the absence of biomarker measurements.⁴⁸ As in most studies of this type, estimates of the effects of prediagnosis exposures are potentially biased because of selecting patients who survived until the occurrence of colorectal cancer or the first postdiagnosis questionnaire.^49–51

In conclusion, higher intakes of total calcium and milk after, but not before, colorectal cancer diagnosis may be associated with lower overall mortality. Our findings, if replicated in future observational studies and randomized trials, will provide important guidance for cancer survivors who are actively seeking diet and lifestyle changes to improve their prognosis.

The American Cancer Society funds the creation, maintenance, and updating of the Cancer Prevention Study-II (CPS-II) cohort. We thank the CPS-II participants and Study Management Group for their invaluable contributions to this research. The authors also acknowledge the contribution to this study from central cancer registries supported through the Centers for Disease Control and Prevention National Program of Cancer Registries and cancer registries supported by the National Cancer Institute SEER program.

The authors indicated no potential conflicts of interest.