Correcting for confounding in longitudinal experiments: positioning NLME as implementation of standardization using latent exchangeability

Non-linear mixed effects modeling and simulation (NLME M& S) is evaluated to be used for standardization with longitudinal data. Non-linear mixed effects modeling (NLME) is a particular implementation of standardization that conditions on individual parameters – the random effects of the mixed effects model. The present work is motivated by the fact that in pharmacometrics NLME M& S is routinely used to analyze clinical trials and to predict and compare hypothetical outcomes of the same patient population under different treatment regimens. Such a comparison is a causal question sometimes referred to as causal prediction. Nonetheless, NLME M& S is rarely positioned as a method for causal prediction. As example a simulated clinical trial is used that assumes treatment confounder feedback in which early outcomes can cause deviations from the planned treatment schedule and are correlated with the final outcome. Being interested in the outcome for the planned treatment schedule, we evaluate possibilities to correct for the confounding using NLME M& S based on latent conditional exchangeability or implementations of other causal inference g-methods based on sequential conditional exchangeability (inverse probability weighting (IPW), standardization or g-estimation). All the methods can correct for the confounding, as long as assumptions required for identification of the estimand hold, including in particular, positivity, no unobserved confounding, and correct specification of the models used for the analyses.