Calcium could decrease risk of colorectal neoplasia by binding bowel-irritating compounds and diminishing mucosal proliferation. This study quantitatively summarizes epidemiologic studies addressing this hypothesis and aims to explain heterogeneity between studies. Twenty-four articles reported 43 measures of relative risks (RRs). The weighted mean, according to a random effects model, did not indicate substantial protection by calcium [RR = 0.89; 95% confidence interval (CI) = 0.79-1.01]. Results from different studies showed substantial heterogeneity, with the "true" underlying RRs ranging from about 0.50 to 1.60. Summary RRs for cohort and case-control studies were 0.90 and 0.88, respectively. For adenomas and carcinomas, RRs were 1.13 (95% CI = 0.91-1.39) and 0.86 (95% CI = 0.74-0.98), respectively, both falling within the range of between study heterogeneity. With respect to subsites, lower RRs were observed for estimates that included proximal colon as one of the subsites (RR = 0.67), whereas the RR was close to 1.0 for distal (RR = 0.97) and rectal subsites (RR = 0.99). Stratification on study characteristics and weighted regression analysis yielded RRs slightly below 1.0, with considerable heterogeneity. These results do not support the hypothesis that calcium prevents colorectal neoplasia.