Genetic relationships and causality between overall and central adiposity and breast, prostate, lung and colorectal cancer

ABSTRACT OBJECTIVE Diverse measures of obesity relate to cancer risk differently. Here we assess the relationship between overall and central adiposity and cancer. METHODS We constructed z-score weighted polygenic scores (PGS) for two obesity-related phenotypes; body mass index (BMI) and BMI adjusted waist-to-hip ratio (WHRadjBMI) and tested for their association with five cancers in the UK Biobank: overall breast (BrC), post-menopausal breast (PostBrC), prostate (PrC), colorectal (CrC) and lung (LungC) cancer. We utilised publicly available data to perform bi-directional Mendelian randomization (MR) between BMI/WHRadjBMI and BrC, PrC and CrC. RESULTS PGS BMI had significant multiple testing-corrected inverse association with PrC (OR[95%CI]=0.97[0.95-0.99], P =0.0012) but PGS WHRadjBMI was not associated with PrC. PGS BMI was associated with PostBrC (OR[95%CI]=0.97[0.96-0.99], P =0.00203) while PGS WHRadjBMI had nominal association with BrC. PGS BMI had nominal positive association with LungC. MR analyses showed significant multiple testing-corrected protective causal effect of BMI on PrC (OR[95%CI]=0.993[0.988-0.998], P =4.19×10 −3 ). WHRadjBMI had a nominal causal effect on higher PrC risk (OR[95%CI]=1.022[1.0067-1.038], P =0.0053). We also report nominal causal protective effect of WHRadjBMI on breast cancer (OR[95%CI]=0.99[0.98-0.997], P =0.0068). Neither PGS nor MR analyses were significant for CrC. CONCLUSIONS Higher overall adiposity appears protective from PrC while higher central adiposity is a potential risk factor for PrC but protective from BrC. STUDY IMPORTANCE What is already known about this subject? Observational studies suggest obesity is associated with higher risk of certain cancers and at the same time is protective of other cancers. The direction of association is in part influenced by the anthropometric trait used to assess obesity. Higher BMI appears protective from prostate, breast and lung cancers but is a risk factor for post-menopausal breast, pancreatic and colorectal cancers. What are the new findings in your manuscript? We implement Mendelian randomization approach using large scale datasets and show a protective causal effect of higher BMI from prostate cancer but suggest that higher WHRadjBMI is causal for prostate cancer. We also show nominal evidence of WHRadjBMI being causally protective from breast cancer. How might your results change the direction of research or the focus of clinical practice? We demonstrate the importance of partitioning obesity into discrete types depending on the area of fat deposition rather than using an overall measure. Our results show that diverse measures of obesity relate differently to cancer risk. In fact, even for the same type of cancer, overall and central obesity measures may impact in opposite direction in terms of risk to cancer.