Structural Equation Modeling in Mendelian Randomization for Lifespan Adiposity Effects

Overview

This study applied structural mean models (SMM) within a Mendelian randomization (MR) framework to estimate the period-specific effects of adiposity during childhood and adulthood on cardiovascular disease, type 2 diabetes, and breast cancer. Persistent adulthood adiposity increased cardiovascular and diabetes risk, while higher childhood adiposity showed a protective effect against breast cancer. The study compared SMM-MR with inverse variance weighted multivariable MR (IVW-MVMR), highlighting differing assumptions and methodological implications for lifecourse research.

Background

Mendelian randomization (MR) uses genetic variants as instrumental variables to infer causal effects of modifiable exposures on health outcomes, typically estimating lifetime average effects. However, genetic influences on exposures may vary across different life stages, complicating interpretation of MR results. Lifecourse approaches aim to disentangle early versus later life exposure effects, necessitating advanced MR methods such as structural mean models (SMM) and inverse variance weighted multivariable MR (IVW-MVMR). These methods differ in assumptions about genetic instruments and exposure timing, impacting causal inference in time-varying contexts.

Data Highlights

The study applied g-estimation of SMMs to estimate period effects of adiposity at childhood and adulthood on three outcomes: cardiovascular disease (CVD), type 2 diabetes (T2D), and breast cancer. Key findings included persistent increased risks of CVD and T2D associated with higher adulthood adiposity, and a protective effect of higher childhood adiposity on breast cancer risk. The SMM-MR approach was compared with IVW-MVMR, which also assesses time-varying effects but under different assumptions regarding genetic instruments and exposure liabilities.

Key Findings

• Higher adulthood adiposity has a persistent period effect increasing risk of cardiovascular disease and type 2 diabetes.
• Higher childhood adiposity exhibits a protective period effect against breast cancer risk.
• SMM-MR uses g-estimation and structural mean models to estimate period-specific causal effects, circumventing some parametric assumptions of other MR methods.
• IVW-MVMR estimates controlled effects of correlated exposures over time but relies on assumptions about liabilities to exposures and their genetic predictors.
• SMM-MR assumes no direct association of genetic instruments with exposures outside the period considered, whereas IVW-MVMR allows pathways via other time periods.
• Methodological assumptions critically influence interpretation of MR results in lifecourse research.

Clinical Implications

Clinicians and researchers should consider that adiposity at different life stages may have distinct causal effects on disease risk, with adulthood adiposity increasing cardiometabolic risks and childhood adiposity potentially reducing breast cancer risk. Applying MR methods that account for time-varying exposures, such as SMM-MR or IVW-MVMR, can improve causal inference and guide targeted prevention strategies. Awareness of the underlying assumptions of each method is essential for accurate interpretation and application in clinical research.

Conclusion

This study demonstrates that advanced MR methods incorporating structural equation modeling can disentangle period-specific effects of adiposity on disease outcomes across the lifespan. Careful consideration of methodological assumptions is crucial for valid causal inference in lifecourse epidemiology.

References

1. Davey Smith & Hemani 2014 -- Mendelian randomization: genetic anchors for causal inference in epidemiological studies
2. Burgess et al. 2015 -- Mendelian randomization: methods for using genetic variants in causal estimation
3. Tchetgen Tchetgen et al. 2015 -- Structural mean models and g-estimation in causal inference
4. Sanderson et al. 2022 -- Multivariable Mendelian randomization: the use of pleiotropic genetic variants to estimate causal effects