Define the causal intervention represented by parameters in discrete-time multivariable MR with repeated exposure measures

Determine the precise causal intervention on the exposure trajectory that is represented by parameter estimates in multivariable Mendelian randomization models that treat repeated measurements of the same exposure at discrete time points as separate exposures, in order to clarify the causal interpretation of these estimated parameters.

Background

The paper critiques prior time-discrete approaches to time-varying Mendelian randomization that model repeated measurements of an exposure at selected time points as separate nodes or exposures, often within multivariable MR frameworks. The authors argue that these models lack a clear data-generating mechanism in a time-varying context and raise interpretational challenges.

Specifically, they note that while such methods may yield parameter estimates, it is unclear what specific intervention on the exposure trajectory these estimates correspond to, which undermines causal interpretation. Clarifying this correspondence is necessary to make rigorous causal claims from such discrete-time multivariable MR analyses.