Systematic assessment finds confounding gap in multiple sclerosis trial emulations

A systematic assessment of evidence on treatments for relapsing-remitting multiple sclerosis found a substantial mismatch between factors that influence clinical decisions and those considered in studies intended to estimate treatment effects. Researchers assessed 53 observational studies and 27 randomized trials, interviewed eight clinical experts and surveyed eight clinicians. The analysis identified 160 potential confounders relevant to designing a target trial emulation comparing two multiple sclerosis drugs, raising concerns about the validity of observational treatment-effect estimates when important sources of confounding remain unmeasured.

The study combined a systematic literature analysis, qualitative interviews and a quantitative clinician survey. The literature search was last updated on November 2, 2023, and focused on both observational research and randomized trials concerning the comparison of two drugs for relapsing-remitting multiple sclerosis. All potential confounders found through these sources were categorized and assessed for clinical importance. Researchers also estimated the resources required for each stage of the process and for identifying each potential confounder.

Most of the 160 candidate confounders—125, or 78%—appeared in the published literature, while interviews with clinical experts contributed 35 additional factors, representing 22% of the total. This result suggests that published research alone did not capture the full range of variables that may influence treatment selection and outcomes. Interviews were especially valuable for identifying patient-perspective factors: every potential confounder related to the patient perspective in the final assessment was found through interviews rather than through the literature review.

Clinician judgments also exposed gaps between the literature and routine treatment decisions. Of 17 potential confounders that clinicians rated as very important for treatment decisions, nine, or 53%, were identified through interviews. Across the combined evidence sources, the five factors considered most frequently were age, cited in 71 studies; gender or sex, in 62; disability status, in 62; disease duration, in 53; and relapses, in 44. Yet these five factors appeared in only 16 of the 53 observational studies, or 30%. The findings therefore point to a disconnect not merely between expert knowledge and published research, but also between factors clinicians consider relevant and variables routinely handled in observational analyses of treatment effects.

The resource estimate provides a benchmark for future target trial emulation studies: the researchers judged that 62 person-days were needed to identify the 160 potential confounders. Even with that level of effort, the assessment concluded that the key assumption of no unmeasured confounding is unlikely to be met in observational studies of these treatments. Because unmeasured confounding can distort comparisons between treatment groups, the authors warned that credible real-world evidence on treatment effects may remain difficult to produce. The work is methodological rather than a comparison of the two drugs, and it does not establish that either treatment is more effective or safe. Its central finding is that designing a target trial emulation requires systematic searches across evidence and clinical expertise, particularly to capture patient-centered and decision-relevant factors that may otherwise be overlooked.

Subject of Research: Identification and assessment of potential confounders for target trial emulation in relapsing-remitting multiple sclerosis

Article Title: Identification of known potential confounders to design target trial emulation studies: a systematic assessment

Article References: Hirt, J., Woelfle, T., Janiaud, P., & Hemkens, L. G. (2026). Identification of known potential confounders to design target trial emulation studies: a systematic assessment. BMC Medicine. https://doi.org/10.1186/s12916-026-05328-3

Image Credits: AI Generated

DOI: 10.1186/s12916-026-05328-3

Keywords: multiple sclerosis, target trial emulation, observational studies, confounding, treatment effects, relapsing-remitting MS, clinical decisions, systematic assessment, patient perspective, Identification, known, potential