Possible selection bias in Mendelian randomization studies: effect of body mass index on breast cancer as an example.

The inverse association of body mass index (BMI) with postmenopausal breast cancer (BC) risk observed in Mendelian Randomization (MR) studies contradicts the International Agency for Research on Cancer (IARC) classification of high BMI as a risk factor for postmenopausal BC. We investigated whether this discrepancy stems from survival-driven selection bias in the underlying GWAS. Using summary statistics from three different GWAS, higher genetically predicted BMI was inversely associated with BC risk in the BCAC consortium (odds ratio (OR): 0.74, 95% CI: 0.64, 0.84) and the UK Biobank (OR: 0.70, 95% CI: 0.56, 0.87), but showed no relationship with sibling BC (BCs) risk (OR: 1.00, 95% CI: 0.99, 1.01). These findings demonstrate that previous MR studies may suffer from selection bias driven by survivorship bias. This highlights a critical methodological vulnerability in MR studies when both exposure and outcome cause premature mortality, and the underlying population is not fully captured by current GWAS.