Genetically predicted vitamin D levels show no link to head and neck cancer risk

Vitamin D has long been cast as a kind of nutritional guardian against cancer, a molecule whose deficiency seemed to lurk behind a long list of malignancies. Head and neck cancer, with its well-documented seasonal patterns, its strong association with outdoor work and sun exposure, and its demonstrated links to immune function, has been a particular focus of this hypothesis. Now a new genetic study delivers a sobering verdict: using some of the strongest evidence available in human epidemiology, researchers report that genetically predicted vitamin D levels show no causal effect on the risk of head and neck cancer or any of its major subtypes.

The study, published in Cancer Causes & Control by Gowri Sivaramakrishnan of the Bahrain Defence Force Royal Medical Services and Kannan Sridharan of Arabian Gulf University, employed a technique called Mendelian randomization, a method that has become one of the most powerful tools for distinguishing genuine causation from mere correlation in observational data. Rather than measuring vitamin D in blood samples and waiting to see who develops cancer, the researchers exploited a fundamental quirk of human genetics: that variations scattered across the genome, inherited at conception, influence how much 25-hydroxyvitamin D circulates in a person’s bloodstream. Because these genetic variants are randomly assorted at birth, they are largely immune to the confounding factors, reverse causation, and measurement error that plague conventional observational studies.

The technical machinery of the analysis was substantial. The authors drew genetic instruments from genome-wide association studies encompassing 441,291 individuals in the UK Biobank, ultimately identifying 115 independent single nucleotide polymorphisms associated with serum 25-hydroxyvitamin D levels. Together, these variants explained 5.12 percent of the total phenotypic variance in vitamin D status, a figure the authors deemed sufficient for robust causal inference. The mean F-statistic, a measure of instrument strength used to judge whether genetic variants are powerful enough proxies for the exposure of interest, was 198.3, comfortably satisfying standard Mendelian randomization assumptions and well above the conventional threshold of 10.

Cancer outcomes were drawn from the HEADSpAcE consortium, a large international collaboration focused on the genetics of head and neck cancer. The researchers examined not only overall head and neck cancer but also its clinically distinct major subtypes: oral cavity cancer, laryngeal cancer, hypopharyngeal cancer, and HPV-negative oropharyngeal cancer. The primary causal estimates were derived using inverse variance weighted Mendelian randomization, a method that effectively performs a weighted regression of the genetic associations with cancer on the genetic associations with vitamin D, treating the slope of that regression as the causal effect estimate. Supporting analyses employed MR-Egger regression, which allows for directional pleiotropy to be present and simultaneously tests for it through its intercept term; the weighted median estimator, which is robust to up to half the instruments being invalid; and mode-based methods, which estimate the effect at the peak of the distribution of individual variant ratios.

The results were unambiguous. Across every subtype examined, the inverse variance weighted estimates showed no evidence of a causal association between genetically predicted 25-hydroxyvitamin D levels and cancer risk. All p-values exceeded 0.05, with odds ratios ranging from 0.95 to 1.25, a span that hovers around the null value of 1.0 and provides no hint of either a protective or a harmful effect. Nor did the complementary estimation methods, MR-Egger, weighted median, or mode-based approaches, produce findings that would suggest a hidden effect masked by the choice of primary estimator.

Perhaps more revealing than the null result itself is what the sensitivity analyses did not find. The researchers deployed an extensive battery of tests designed to detect the failure modes that can render Mendelian randomization unreliable. Cochran’s Q statistic and related methods assessed heterogeneity among the individual variant estimates. MR-Egger intercept testing, the MR-PRESSO algorithm, and leave-one-out analyses, in which each variant is removed in turn and the analysis rerun, probed for horizontal pleiotropy, influential outliers, and directional bias. Steiger filtering assessed directionality to ensure that genetic variation was driving vitamin D levels rather than the reverse. Across all cancer outcomes, no evidence of pleiotropy, outliers, or reverse causation emerged.

The researchers also took the unusual and rigorous step of systematically screening their genetic instruments for potential confounding through pleiotropic pathways. They examined associations between their 115 variants and 682 SNP-confounder pairs spanning major lifestyle factors, ultraviolet radiation exposure proxies, and other health-related traits. Only 17 of 682 associations, roughly 2.5 percent, reached nominal significance at the conventional genome-wide threshold of 5 times 10 to the power of minus 8, a proportion consistent with pure chance. Critically, none survived Bonferroni correction for multiple testing. The authors interpret this as strong evidence that their instruments are not systematically tracking sunlight exposure, outdoor occupation, physical activity, smoking, alcohol consumption, or other behaviors that could independently influence cancer risk.

To validate the entire analytical pipeline, the study incorporated a positive control. Multiple sclerosis is a condition with a well-established inverse relationship with vitamin D in prior genetic studies, making it an ideal test case. The same 115 instruments, applied to multiple sclerosis outcomes, produced exactly what the literature predicts. Inverse variance weighted analysis yielded an odds ratio of 0.84, with a 95 percent confidence interval of 0.71 to 0.99 and a p-value of 0.038. The weighted median estimator produced an odds ratio of 0.82, with a confidence interval of 0.70 to 0.96 and a p-value of 0.015. This confirms that the instruments and methods are fully capable of detecting a genuine causal signal when one exists, lending considerable weight to the null findings for head and neck cancer.

The study arrives amid a long and often contradictory literature. Observational research has variously reported that higher vitamin D intake, higher circulating 25-hydroxyvitamin D, or specific polymorphisms in the vitamin D receptor are associated with reduced head and neck cancer risk, improved survival, or no effect at all. Meta-analyses have struggled to reconcile these findings, in part because vitamin D status is tightly correlated with sunlight exposure, diet, adiposity, physical activity, and socioeconomic position, each of which also bears on cancer risk in ways that are difficult to fully adjust for statistically. Prior Mendelian randomization efforts, including a 2018 analysis of oral and oropharyngeal cancer and more recent studies focused on laryngeal and oropharyngeal sites, have produced mixed and sometimes conflicting conclusions, often limited by smaller instrument sets or narrower subtype coverage.

By combining an expanded instrument set of 115 variants with rigorous sensitivity testing, subtype-specific analyses, and a validated positive control, this study provides what the authors describe as robust genetic evidence that circulating 25-hydroxyvitamin D is not a major causal determinant of head and neck cancer risk. The findings do not negate the biological plausibility of vitamin D’s anti-proliferative and immunomodulatory effects, which have been documented in cell culture and animal models, but they do suggest that whatever influence vitamin D exerts at the epidemiological level is either too small to detect, operates through pathways not captured by circulating 25-hydroxyvitamin D, or is fully explained by confounding in the observational literature.

The practical implications are significant for a field in which supplementation trials have been advocated on the basis of observational associations alone. For individuals at elevated risk of head and neck cancer, whether through tobacco use, alcohol consumption, HPV infection, or occupational exposure, the study suggests that correcting vitamin D status, while important for bone health and other established functions, should not be expected to meaningfully alter cancer risk. The authors note that their instruments are not confounded by major lifestyle or ultraviolet-related factors, a critical point that strengthens confidence in the null result. As with any Mendelian randomization study, the findings apply to genetically influenced variation in vitamin D levels within the population studied, and extrapolation to pharmacological doses or to populations with different ancestral backgrounds warrants caution. But within those limits, the message is clear and, for those hoping vitamin D might be a simple protective lever against head and neck cancer, likely disappointing.