Comparative effectiveness studies in multiple sclerosis

• Published in: Neurodegenerative disease management Vol. 10; no. 3; pp. 113 - 118
• Main Authors: Hersh, Carrie M, Hua, Le H
• Format: Journal Article
• Language: English
• Published: London Future Medicine Ltd 01.06.2020
• Subjects: Bias; Clinical medicine; comparative effectiveness studies; Decision making; disease modifying therapies; Effectiveness studies; Multiple sclerosis; Observational studies; Patients; propensity score methodology
• ISSN: 1758-2024; 1758-2032
• DOI: 10.2217/nmt-2020-0015

In this observational context, treatment groups differ systematically and are therefore noncomparable. [...]one must account for such systematic differences in baseline characteristics between groups of interest when estimating the effect of treatment on desired outcomes. [...]when building a PS model for an MS study evaluating treatment A versus treatment B, important variables to incorporate into the model include demographics (e.g., age, sex and race), baseline disease characteristics (e.g., prior relapses and gadolinium-enhancing lesions, disease course, disease duration, prior number and type of DMTs) and comorbidities (e.g., vascular comorbidities, such as hypertension, hyperlipidemia, diabetes mellitus, tobacco exposure, chronic heart and lung disease). Because the PS is a function of covariates rather than outcomes, the estimated effect is therefore not biased by the desired outcome. The most common implementation of PS matching is one-to-one or pair matching, in which pairs of treated and untreated subjects are formed, such that matched subjects have similar values of the PS. [...]in a set of subjects with the same PS, the distribution of observed baseline covariates between treated and untreated groups is the same. Within each PS stratum, treated and untreated subjects will have approximately similar values of the PS. [...]when the PS has been correctly specified, the distribution of measured baseline covariates will be roughly similar between treated and untreated subjects within the same stratum (3,4).