Principal Data Scientist Hoffman- La Roche, United Kingdom
Background: Spinal muscular atrophy (SMA) is a rare genetic disorder for which three disease-modifying therapies (DMT) have been approved, including Evrysdi® (risdiplam), the first orally administered targeted therapy for patients with SMA. The post-authorization long-term prospective effectiveness study (European Medicines Agency requirement) entails the comparison of disease progression in patients treated with risdiplam versus. DMT-naïve patients; stratified by symptomatic/ pre-symptomatic status and number of copies of the SMN2 gene.
Objectives: To describe the challenges of this study and to provide design, methodological and operational solutions to researchers facing similar constrictions.
Methods: To address the design and operational challenges, targeted literature review, brainstorming sessions, multi-stakeholder consultations, and discussions within a multidisciplinary team were conducted.
Results: Three challenges were identified: 1) overall small sample size due to the rare disease (especially for pre-symptomatic and the 4 - SMN2 copies subgroups), 2) sample size and follow-up time imbalance between risdiplam treated and naïve cohort (low number of treatment-naïve patients given the treatment options available), and 3) treatment-naïve patients cross-over to the risdiplam cohort resulting in censoring bias as the decision of starting or switching treatment might be related to disease progression.
To overcome the first challenge, patient-level data will be pooled across global registries to maximise the sample size, especially for the SMN2 subgroup analyses. Additionally, data can be complemented by aggregated data from other registries, using appropriate privacy-protecting analytical methods.
To minimise the imbalance between the two study cohorts in terms of number of patients and duration of follow-up, the start of the eligibility period for untreated patients was defined in a way to enhance the sample size and the follow-up time while minimising the impact of clinical practice changes.
Patients in the treatment naïve cohort can only be followed up until crossing over to risdiplam cohort or censorship (because of the initiation of other DMTs). To address censoring bias related to cross-over or treatment switching, the Inverse Probability of Censoring Weighting (IPCW) method will be considered.
Conclusions: We present the design and analytical solutions to optimize the available data in a rare disease. Further analytical methods might be considered to address future challenges that arise due to the nature of a long-term prospective comparative study.