Background: Little is known regarding the effectiveness of statins for primary prevention among older adults, and analysis of observational data can add crucial information on the benefits of actual patterns of use. Latent class growth models (LCGM) are increasingly proposed as a solution to summarize the observed longitudinal treatment in a few distinct groups. When combined with standard approaches like Cox proportional hazards models, LCGM can fail to control time-dependent confounding bias because of time-varying covariates that have a double role of confounders and mediators.
Objectives: To propose a valid approach that makes use of LCGM for estimating the effect of treatment trajectories (statin adherence patterns) on an outcome (first cardiovascular event).
Methods: We propose to use LCGM to classify individuals into a few latent classes based on their medication adherence pattern, then choose a working marginal structural model that relates the outcome to these groups. Our approach allows to estimate the parameters of interest using inverse probability of treatment weighting (IPTW) and conservative inferences can be obtained using a standard robust variance estimator. Simulation studies are used to illustrate our approach and compare it with unadjusted, baseline covariates-adjusted, time-varying covariates adjusted and inverse probability of trajectory groups weighting adjusted. Scenarios with varying types of outcomes (continuous, binary or time to event), number of follow-up times (3, 5, 10) and number of trajectory classes (3, 4, 5) are considered.
Results: For all explored scenarios, we find that our proposed approach yield estimators with little or no bias. As expected, when using stabilized IPTW, all confidence interval coverages are close to 95% (between 90% and 98%). Opposingly, regardless of the number of follow-up times and number of trajectory classes, alternative LCGM analyses are highly biased with low coverage of their confidence interval (between 2% and 67%).
Conclusions: The combination of LCGM with MSM is a convenient way to describe treatment adherence and can effectively control time-dependent confounding. Moreover, it can be implemented using standard software such as SAS or R.