Marginal structural models (MSMs) are statistical methods used to examine how treatments change over time. To ensure accurate results, it is important to choose appropriate time intervals when applying these statistical models. However, the impact of selecting different time interval durations for MSMs is not well understood. Different time intervals and approaches can influence the accuracy of results. In this methodological study, we aim to investigate how diffferent time interval approaches affect MSMs by examining this issue in a specific clinical example: comparing sulfonylurea versus metformin as first line therapies and the risk of major adverse cardiovascular events (an outcome characterized by blocked arteries that includes heart attacks, strokes, and dying of heart-related causes) among patients with type 2 diabetes (a condition characterized by elevated blood sugar levels). To achieve this aim, we will use a simulated cohort that we will create using a statistical approach called a plasmode framework. The plasmode simulation framework allows to modify data while preserving the underlying relationships observed in the real data. We will create 9 simulation scenarios with different methodological aspects. We will conduct a real-world case study comparing sulfonylurea versus metformin and their associated risk of major adverse cardiovascular events between September 1st 2002 and March 31st 2022 to illustrate how results are affected by different time intervals in MSMs.
Marginal Structural Models (MSMs) are statistical models used to examine the effects of time-varying treatments or exposures in longitudinal studies. They are often used in studies that involve time-to-event data and are estimated using the inverse-probability-of-treatment weighting (IPTW) estimator. However, the choice of time intervals for MSMs can have an important impact on data granularity, computational burden, and the risk of extreme weights, which may lead to biased estimates. To better understand the impact of different time interval durations on MSMs, this study will compare the use of different time intervals and approaches in a case study comparing sulfonylurea versus metformin as first line therapy and the risk of major adverse cardiovascular events (MACE) among patients with type 2 diabetes. The study will use a plasmode simulation framework to generate data and establish 9 simulation scenarios with varying prevalence of outcome and censoring to compare the occurrence of extreme weight with different time intervals using MSMs. We will also examine other concerns: the first one is to explore how the impact of different lengths of patient's follow-up, resulting in varying numbers of time intervals, may lead to fragility with extreme weight. We will also explore potential problems that can arise from fitting pooled prediction models. Finally, we will conduct a real-world case example that evaluates the cumulative effect of sulfonylurea compared with metformin on MACE in patients with type 2 diabetes between September 1st 2002 and March 31st 2022 will be conducted to illustrate the effect of different time intervals on estimates and weights using MSMs.
• Major adverse cardiovascular events (MACE; composite endpoint of myocardial infarction, ischemic stroke, and cardiovascular death)
• Individual components of MACE
• All-cause mortality
• Methodological outcomes (relative bias, mean square error of the estimates of interest, range of extreme weight, proportion of extreme weight, and rate of censored subjects)
Samy Suissa - Chief Investigator - Sir Mortimer B Davis Jewish General Hospital
Kristian Filion - Corresponding Applicant - McGill University
In-Sun Oh - Collaborator - McGill University
Michael Webster-Clark - Collaborator - McGill University
pauline reynier - Collaborator - Sir Mortimer B Davis Jewish General Hospital
Robert Platt - Collaborator - McGill University
HES Admitted Patient Care;ONS Death Registration Data