OBJECTIVES:

Inclusion of treatment sequencing allows economic models to represent treatment pathways more accurately. However, data limitations and additional model complexity pose challenges for modelling treatment sequencing. A targeted literature review was conducted to understand the prevalence and methods of treatment sequence modelling in relapsing multiple sclerosis (RMS), a chronic condition with many available therapies that serves as an informative case study.

METHODS:

A previous systematic literature review of economic evaluations in RMS was updated by searching MEDLINE, Embase, and EBM Reviews databases from 1^st January 2019 to 8^th May 2023. Studies were included if published in English and presenting cost-effectiveness outcomes for adult patients with RMS. Data extraction was performed using Microsoft Excel.

RESULTS:

79 publications were included, reporting on Markov models (n=63), microsimulations (n=5), trial-based analysis (n=1), or a combination (n=1). 8 studies did not report or use a formal model and 1 was a review. Treatment sequencing was reported in 16 (20%) publications; the number of sequences in a given model ranged from 2 to 445, 2 being the most common (n=7). Most models included 2 treatments in sequence (n=5); the longest sequence comprised up to 9 treatments (n=1). Of the publications that reported their assumptions (n=7), subsequent treatment effectiveness was either directly observed (n=1), determined using simulated decision-making (n=2), assumed equal (n=2), or assumed waning at each subsequent line (n=2).

CONCLUSIONS:

Despite the numerous RMS treatments available, treatment sequence modelling remains limited and there is no consistent method for modelling treatment effectiveness at later lines. Resultantly, the patient experience is frequently not accurately represented in RMS economic evaluations, which could lead to misinformed decision-making. Published economic evaluations in RMS should more routinely provide justification if treatment sequencing is not modelled; modelling sequencing in scenarios would help to address the impact of this modelling choice.