Triangulating evidence to evaluate selection bias in lifecourse Mendelian randomization studies: a practical...

Higher adiposity in early-life has consistently been associated with a reduced risk of breast cancer in later life, with Mendelian randomization (MR) studies supporting a causal effect. However, concerns have been raised that selection bias, particularly collider stratification due to selective participation or survival, may induce spurious protective MR estimates.

We triangulated across empirical analyses and simulations to evaluate whether selection-induced bias could plausibly explain the inverse effect estimate of early-life adiposity on breast cancer risk. First, we analysed proxy-genotype Mendelian randomization (MR) analyses of breast cancer in relatives, in which participant genotype is used as a proxy for relatives' genotype, and conducted family-based simulations to assess whether attenuation in relative-based estimates could arise without selection bias. Second, we performed multivariable MR analyses of parental survival to evaluate survival-related selection mechanisms. Third, we conducted extensive simulations to quantify the magnitude of bias introduced by selection under a range of plausible and extreme scenarios, including interaction-driven selection.

The weaker proxy-genotype MR estimates of breast cancer in relatives, compared with MR estimates for an individual's own breast cancer, were reproduced in family-based simulations without selection bias, indicating that this pattern does not provide evidence for selection bias. Multivariable MR analyses of parental survival indicated that survival differences are primarily driven by mid-to-late adulthood, not early-life, adiposity, providing little support for survival-related selection acting through early-life adiposity. In simulation analyses, additive selection produced minimal bias, while interaction-driven selection generated increasing distortion; however, even under extreme scenarios, the magnitude of bias was insufficient to replicate the observed protective effect estimate. In simulations where selection depended on mid-to-late adulthood rather than early-life adiposity, bias was expressed primarily in mid-to-late adulthood MR estimates, with little distortion of early-life MR estimates. Across all simulated scenarios, the combined pattern of empirical univariable and multivariable MR findings was not reproduced by selection alone.

Although selection bias can influence MR estimates, our findings suggest that plausible selection mechanisms are unlikely to substantively explain the observed inverse effect estimate of early-life adiposity on breast cancer risk. These results support a causal interpretation of the strong protective effect estimate of early-life adiposity on breast cancer risk and highlight the value of triangulating evidence across complementary approaches when evaluating bias in lifecourse MR.

Previous studies have found that having a larger body size in early life is linked to a lower risk of developing breast cancer later in life. Mendelian randomization studies, which use genetic variation to investigate possible causal effects, have supported this finding. However, it has been suggested that the result could be explained by selection bias arising from who survives or participates in the studies analysed. We used several complementary approaches to evaluate this possibility. Together, the findings provide little evidence that the selection mechanisms examined explain the protective effect estimate of early-life adiposity on breast cancer. This work also provides a practical framework for investigating selection bias in lifecourse Mendelian randomization studies.