Low vitamin D levels as a risk factor for cancerBMJ 2017; 359 doi: https://doi.org/10.1136/bmj.j4952 (Published 31 October 2017) Cite this as: BMJ 2017;359:j4952
- Despoina Manousaki, paediatric endocrinologist1,
- J Brent Richards, endocrinologist and associate professor1 2 3
- 1Centre for Clinical Epidemiology, Department of Epidemiology, Lady Davis Institute for Medical Research, Jewish General Hospital, McGill University, Montreal, Quebec, Canada
- 2Departments of Medicine, Human Genetics, Epidemiology and Biostatistics, Jewish General Hospital, McGill University Montreal, Quebec, Canada
- 3Department of Twin Research and Genetic Epidemiology, King’s College London, London, UK
- Correspondence to: J B Richards
Arguably, to date the most clinically effective way to reduce the burden of cancer has been through primary prevention.1 One promising such strategy has been to target vitamin D insufficiency as epidemiological and animal studies have found that low levels of vitamin D are associated with an increased risk of cancer.23 Furthermore, vitamin D insufficiency affects 40% of the general population,4 is easily diagnosed by a simple blood test, and can be treated safely and inexpensively. This may partially explain the 60-fold increase in the use of vitamin D supplements in the US general population between 2000 and 2014, where 18% currently take ≥1000 IU of vitamin D daily.5
Yet the most prudent way to test the efficacy of vitamin D would be through large scale randomised controlled trials because observational studies may be biased by residual confounding (vitamin D levels are confounded by known drivers of cancer risk, such as smoking, obesity, and the healthy user effect6). Yet such trials are cumbersome and rely on the public purse, as vitamin D is not patentable. In this issue Dimitrakopoulou and colleagues (doi:10.1136/bmj.j4761) present new evidence that is clinically relevant in the absence of high quality trial data.7
The authors used Mendelian randomisation, a study design that greatly limits potential bias from confounding since the genetic determinants of vitamin D level are randomised at conception and are not associated with known vitamin D confounders.8 These genetic variants provide the opportunity to vary vitamin D levels in the population free from confounding, much like randomisation does in clinical trials. Furthermore, as they are allocated at conception and remain stable over the lifetime, these variants provide an estimate of the causal effect of decades of lowered vitamin D level. The most troubling assumption of such studies is that the genetic determinants act on cancer only through vitamin D levels. Bias from this assumption is not likely here, however, since all genetic variants are in or near genes that have known biological functions with a direct impact on vitamin D level.9
In their Mendelian randomisation analysis, Dimitrakopoulou and colleagues used four common genetic alleles that were strongly associated with lowered levels of 25-hydroxyvitamin D, the commonly measured biomarker of vitamin D status in blood, in a cohort of almost 34 000 individuals.10 They examined if these alleles were associated with a reduced risk of seven different types of cancer (colorectal, breast, prostate, ovarian, lung, pancreatic, and neuroblastoma) in separate large epidemiological cohorts, together comprising 70 563 cancer cases and 84 418 controls.111213 Their study had sufficient statistical power to detect moderate effects on cancers (with minimum odds ratios between 1.15 and 1.5) for a genetically determined 25 nmol/L decrease in 25-hydroxyvitamin D concentration. In general, their results failed to provide evidence of a causal association between reduced vitamin D level and risk of cancer, for all seven types of cancer studied, as well as for some cancer subtypes. These results persisted despite sensitivity analyses probing the assumptions of Mendelian randomisation.
Although convincing, the study by Dimitrakopoulou and colleagues is limited by its insufficient power to detect smaller effects of vitamin D in cancer. This might explain the discordance between their results and those of a recent Mendelian randomisation study examining the association of genetically lowered vitamin D level with ovarian cancer on 10 065 cases and 21 654 controls,14 which showed an odds ratio of ovarian cancer of 1.27 (95% confidence interval 1.06 to 1.51) per genetically determined 20 nmol/L decrease in 25-hydroxyvitamin D concentration. Both Mendelian randomisation studies tested the linear effect of vitamin D level in the general population (what would be expected if small doses of vitamin D supplementation were used broadly in the population), but they do not give insight into non-linear effects—that is, whether correction of more profound vitamin D deficiency can prevent cancer. This is relevant because the effects of vitamin D on bone are predominant at extremely low 25-hydroxyvitamin D levels. Despite these limitations, the study by Dimitrakopoulou and colleagues provides direct insight into what has recently become common clinical practice; the use of vitamin D supplementation in people with generally normal vitamin D levels.
Can we now cross common cancers off the list of vitamin D’s beneficial effects? Current evidence suggests that we cannot do so for ovarian cancer. Also, we cannot exclude small effects and effects unique to frank vitamin D deficiency. Nevertheless, it seems unlikely that vitamin D exerts large effects on the cancers tested. This evidence suggests that trials testing these hypotheses would need enormous samples.
It thus seems plausible that some of the previously attributed effect sizes15 may be influenced by confounding. Since large scale, long duration randomised controlled trials for vitamin D supplementation are not currently feasible, the results from Dimitrakopoulou and colleagues provide useful data for the millions of people who have recently started taking vitamin D supplements. Mendelian randomisation studies in even larger cancer cohorts will provide more definitive answers—and more rapidly than clinical trials.
Competing interests: We have read and understood the BMJ policy on declaration of interests and declare the following: BR has received unrestricted funding from Eli Lilly for principal investigator initiated research. This research programme is unrelated to vitamin D or cancer, or both.
Provenance and peer review: Commissioned; not peer reviewed.