Accounting for antihypertensive medication in Mendelian randomization studies of blood pressure: methodological considerations in the Canadian Longitudinal Study on Aging

Abstract

Background: Mendelian randomization (MR) studies investigating determinants of blood pressure (BP) do not account for antihypertensive medication consistently, which may explain discrepancies across studies. We performed an MR study of the association between body mass index (BMI) and systolic BP (SBP) using five methods to account for antihypertensive medication and evaluated their impact on the estimation of the causal effect and on the assessment of the invalidity of the instruments used in MR. Methods: Baseline and follow-up data on 20 430 participants from the Canadian Longitudinal Study on Aging (CLSA) Comprehensive cohort (2011–2018) were used. The five methods to account for antihypertensive medication in the MR study were: no correction, adjustment for antihypertensive medication as a covariate in models, exclusion of treated individuals, addition of a constant value of 15 mmHg to measured values of SBP in treated individuals, and using hypertension as a binary outcome. Results: The magnitude of the estimated MR causal effect for SBP (mmHg) varied across the methods of accounting for antihypertensive medication effects ranging from 0.68 (effect per 1 kg/m2 increase in BMI) in scenario adjusting MR models for medication covariate to 1.35 in that adding 15 mmHg to measured SBP in treated individuals. Conversely, the assessment of the validity of the instruments did not differ across methods of accounting for antihypertensive medication. Conclusions: Methods to account for antihypertensive medication in MR studies may affect the estimation of the causal effects and must be selected with caution.