AbstractThere is strong evidence for a protective association of vitamin D on head and neck cancer (HNC) incidence and prognosis from mechanistic and observational epidemiological studies, with the most recent estimate suggesting a 30% reduction in odds of HNC (95% confidence interval (CI): 44%, 12%) even after correcting for smoking and alcohol use. This thesis aims to identify the causal effect of vitamin D on HNC risk and prognosis by triangulating estimates from the most robust observational sources and Mendelian randomization (MR).
I have undertaken the most recent observational study of vitamin D on HNC risk and prognosis in a case-cohort study in UK Biobank. Incident oral, oropharyngeal and laryngeal cancer cases were identified from cancer registry data and participants with no cancer diagnosis were used as controls. In survival analyses, death was identified from linked death registry data. The association between pre-diagnosis serum 25 hydroxyvitamin D (25(OH)D), the commonly measured blood metabolite, and HNC risk and prognosis was assessed using logistic and Cox proportional hazards regression respectively. A doubling in vitamin D was related to a reduction in oral (OR = 0.61 (95% CI: 0.50, 0.76)) and oropharyngeal (OR = 0.75 (95% CI: 0.61, 0.91)) cancer risk as well as a reduction in hazards of death for all HNC sites combined (HR = 0.64 (85% CI: 0.53, 0.77)).
As part of a small team, I also performed genome-wide association study of 25(OH)D in 17 cohort studies, including two with whole genome sequencing data, and meta-analysed. This study identified three novel associations including an association with a low-frequency genetic variant (rs117913124) with large effects on 25(OH)D (-0.43 standard deviation change per A allele). Genetic variants identified as being robustly related to 25(OH)D were combined into a genetic proxy for serum 25(OH)D. This genetic proxy was compared to the serum 25(OH)D measure in a phenome-wide association study of 25(OH)D to assess the confounding structure of the two measures.
Further, I used MR to assess the causal association of 25(OH)D on HNC risk in the GAME-ON study and HNC prognosis in Head and Neck 5000 using the genetic proxy identified above. For risk the MR estimate for a doubling of 25(OH)D was OR = 0.74 (95% CI: 0.50, 1.10) and 1.12 (95% CI: 0.52,2.39) for oral and oropharyngeal cancer respectively. For HNC prognosis, a doubling of 25(OH)D corresponded to a HR of 0.90 (95% CI: 0.51, 1.57) for all HNC sites combined.
In summary, this thesis cannot rule out small effects of 25(OH)D on cancer risk but suggests that the effects as large as those reported in observational studies are unlikely. The genetic proxy for 25(OH)D has been improved by the work of this thesis but questions remain about how appropriate a genetic proxy for 25(OH)D may be in studies of cancer.
|Date of Award||28 Nov 2019|
|Supervisor||Richard M Martin (Supervisor) & Nicholas John Timpson (Supervisor)|