Large-scale Searching for Causal Relationships of Body Mass Index Bridged Modifiable Lifestyle Factors on Colorectal Cancer Risk: A phenome-Wide Mendelian Randomization Study.

Abstract

Currently, mendelian randomization (MR) has played an important role in studying causality. However, due to the unique role of third-party variables, false positive conclusions may arise in the process of causal inference. We aimed to assess the body mass index (BMI)-mediated effect of modifiable lifestyle factors on colorectal cancer (CRC) risk. The MR phenome-wide association study (MR-pheWAS) was initially performed to identify BMI-relevant lifestyle factors. Heritability estimation, genetic correlation, inverse-variance weighted MR analysis, and mediation analysis were conducted to evaluate the causal relationships among lifestyle, BMI, and CRC in the Genetics and Epidemiology of Colorectal Cancer Consortium (GECCO). The FinnGen database was used to replicate the above results on genetic results and the UK Biobank cohort provided population-level validation for the causality. We initially analyzed the causal relationship between 32,760 phenotypes and BMI through MR-pheWAS. Five lifestyles (i.e., alcohol intake, time spent watching television, walking, walking pace, and smoking status) were found to causally affect BMI. Our results supported a causal association between genetically proxied BMI and CRC [odds ratio (OR)_GECCO = 1.192 (95% confidence interval (CI): 1.038-1.369); OR_FinnGen = 1.567 (95%CI: 1.398-1.756)]. Notably, alcohol intake [β_GECCO = 0.252 (95%CI: 0.002-0.502), β_FinnGen = 0.285 (95%CI: 0.085-0.484)] and time spent watching TV [β_GECCO = 0.470 (95%CI: 0.015-0.924), β_FinnGen = 0.421 (95%CI: 0.084-0.758)] had a direct causal effect on increasing CRC risk. The following effect decomposition indicated that the effect of alcohol intake on CRC was marked by BMI [proportion_GECCO = 5.556% (95% CI: 0.397-9.518%) and proportion_FinnGen = 8.421% (95% CI: 4.912-12.982)]. The proportion of the effect of time spent watching television on CRC mediated by BMI was 9.722% (95% CI: 0.198-19.246%) in GECCO and 16.936% (95% CI: 11.650-22.977%) in FinnGen. In contrast, the full effects of walking, walking pace, and smoking status on CRC were entirely mediated by BMI. A similar mediation pathway was observed in the UK Biobank cohort. This study, combining genetics-driven causal inference, provides evidence for the relationship between lifestyle factors and the risk of CRC ,helping to prioritize strategies for CRC prevention in populations.

Supplementary information: The online version contains supplementary material available at 10.1007/s43657-025-00270-5.