Introduction: Optimise:MS is an observational pharmacovigilance study aimed at characterising the safety profile of disease-modifying therapies (DMTs) for multiple sclerosis (MS) in a real world population. The study will categorise and quantify the occurrence of serious adverse events (SAEs) in a cohort of MS patients recruited from clinical sites around the UK. The study was motivated particularly by a need to establish the safety profile of newer DMTs, but will also gather data on outcomes among treatment-eligible but untreated patients and those receiving established DMTs (interferons and glatiramer acetate),. It will also explore the impact of treatment switching. Methods: Causal pathway confounding between treatment selection and outcomes, together with the variety and complexity of treatment and disease patterns observed among MS patients in the real world, present statistical challenges to be addressed in the analysis plan. We developed an approach
for analysis of the OPTIMISE:MS data that will include disproportionality-based signal detection methods adapted to the longitudinal structure of the data and a longitudinal time-series analysis of a cohort of participants receiving second-generation DMT for the first time. The time-series analyses will use a number of exposure definitions in order to identify temporal patterns, carryover effects and interactions with prior treatments. Time-dependent confounding will be allowed for via inverse probability-of-treatment weighting (IPTW). Additional analyses will examine rates and outcomes of
pregnancies and explore interactions of these with treatment type and duration. Results: To date 13 hospitals have joined the study and over 2000 participants have been recruited.
A statistical analysis plan has been developed and is described here. Conclusion: Optimise:MS is expected to be a rich source of data on the outcomes of DMTs in real world conditions over several years of follow-up in an inclusive sample of UK MS patients. Analysis is complicated by the influence of confounding factors including complex treatment histories and a highly variable disease course, but the statistical analysis plan includes measures to