Matching with time‐dependent treatments: A review and look forward

Abstract

Observational studies of treatment effects attempt to mimic a randomized experiment by balancing the covariate distribution in treated and control groups, thus removing biases related to measured confounders. Methods such as weighting, matching, and stratification, with or without a propensity score, are common in cross-sectional data. When treatments are initiated over longitudinal follow-up, a target pragmatic trial can be emulated using appropriate matching methods. The ideal experiment of interest is simple; patients would be enrolled sequentially, randomized to one or more treatments and followed subsequently. This tutorial defines a class of longitudinal matching methods that emulate this experiment and provides a review of existing variations, with guidance regarding study design, execution, and analysis. These principles are illustrated in application to the study of statins on cardiovascular outcomes in the Framingham Offspring cohort. We identify avenues for future research and highlight the relevance of this methodology to high-quality comparative effectiveness studies in the era of big data. K E Y W O R D S longitudinal matching, new-user design, real-world evidence, time-dependent confounding, time-varying treatment This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.

Long‐term effect of first‐line injectable multiple sclerosis treatments: Input of a time‐dependent propensity score

“…In this study, we investigated how to assess long-term treatment effectiveness in real-life settings. As discussed previously, 8,10 this issue entails considerable challenges, such as the choice of the appropriate control group. The novelty of the present study was the F I G U R E 1 Flowchart of the study population (n = 2383) 26,27 or at the first consultation at the MS expert center.…”

“…In this study, we investigated how to assess long-term treatment effectiveness in real-life settings. As discussed previously, 8,10 this issue entails considerable challenges, such as the choice of the appropriate control group. The novelty of the present study was the The use of a time-fixed PS (i.e., classical PS) to compare treated patients to never treated ones would have been an intuitive and wellknown methodology.…”

“…In this study, we investigated how to assess long‐term treatment effectiveness in real‐life settings. As discussed previously, 8,10 this issue entails considerable challenges, such as the choice of the appropriate control group. The novelty of the present study was the application of a time‐dependent PS to compare patients with MS who had received β‐IFN or GA treatment to patients who had not yet initiated treatment but who had the same probability of being treated at that time.…”

“…The use of a time‐dependent PS may solve these issues, 10 as it enables control patients to be identified who are comparable to the treated patients when the latter start treatment. In the present study, we propose to assess the effectiveness of β‐IFN and GA in MS using a matching method originally devised by Lu 11 …”

Debiasing the estimate of treatment effect on the treated with time-varying counfounders

“…The main issue of these analyses is to handle such confounding. Specific methods address this issue by taking into account time-confounding and thus avoid biased estimates [1,2,3]. The most common method for dealing with time-dependent confunding is the inverse probability weighted marginal structural Cox model [4].…”