Letters to the Editor Artifact

in the control group undermines the conclusions of a vitamin D and cancer study Dear Sir: We call your attention to an apparent problem with the data presented by Lappe et al (1). The authors concluded that their results showed a significant reduction in cancer incidence from vitamin D supplementation. The study population was described as rural Ne-braska women with a median age of 66.7 y with an average follow-up of 4 y. Cancer incidence in the placebo control group appears to be much higher than expected on the basis of surveillance data from the Nebraska Cancer Registry (estimated crude cancer incidence rate for rural women aged 65–69 y, 2000–2005; V Filos, personal commu-nication, 2007). The cancer incidence in the study as a whole closely approximates the expected cancer rates (Table 1). If the placebo control group had experienced cancer rates comparable with the population’s expected rates, any statistical significance observed in this study would almost certainly vanish. Cancer incidence in study participants who did not receive vita-min D (ie, combination of the placebo control group and the calcium-only intervention group) is almost exactly the expected incidence (Table 2). Any attempt to isolate the impact of vitamin D should be based on a comparison of these combined groups with the vitamin D intervention group. Likewise, I am very skeptical that this would produce a statistically significant result. The most likely explanation for the results presented by Lappe et al is a simple failure of randomization—an inherent danger with such a small sample size and a caution for post hoc analyses of this kind. A chance variation of a small number of cases can create a spuriously positive result. Unfortunately, Lappe et al offer an insufficient description of the characteristics of the study groups or the success of the randomization to comment further. In the absence of a plausible explanation from the authors, this observation puts the results and conclusions of their study in serious doubt. No conflicts of interest were reported