(B-208) Leveraging multiple-country data sources for a comparative drug safety study: complexities and rewards in pharmacoepidemiology

Background: In cases of rare drug exposure and rare outcomes, pooling data from various data sources can increase precision in risk estimation. However, heterogeneity may exist in drug exposed populations.

Objectives: Using a study that compared mortality risk in patients with multiple sclerosis (MS) treated with alemtuzumab vs those on other highly efficacious disease modifying therapies (HE-DMTs), we highlight the opportunities and challenges of multi-country studies. We focused on understanding differences in underlying drug exposed populations and potential biases, before outcome analyses.

Methods: The cohort study included MS patients initiating alemtuzumab or other HE-DMTs after alemtuzumab reimbursement in each country. The outcome was all-cause mortality. We used routine health care data from 5 countries: UK (chart review data from three academic centres); Denmark, Sweden, Czech Republic(CR) (national MS registries); Germany (administrative claims data). A common data model was used for data management (except Germany). Propensity score (PS) weighting was used to address confounding by MS disease severity within each country. The success of PS weighting was evaluated using standardized mean differences, PS overlap in graphical outputs, percentages of balanced variables and the effective sample size (ESS) of the comparator cohort.

Results: We included 1,183 patients treated with alemtuzumab (56 (Sweden) to 457 (UK) patients) and 27,573 comparator patients treated with HE-DMT (503 (CR) to 10,511 (Sweden)). Considerable heterogeneity in clinical and demographic characteristics was observed across countries for alemtuzumab patients. E.g., in the UK, 94% alemtuzumab patients were naïve to other HE-DMTs, 52.5% had low levels of disability (Expanded Disability Status Scale [EDSS] < 2.25), and mean MS disease duration was 5.3yrs. In contrast, for Denmark, few alemtuzumab patients (19.4%) were naïve to other HE-DMTs, 20% had low levels of disability (EDSS < 2.25), and mean MS disease duration was 12.2yrs. Despite PS weighting, core covariates remained imbalanced in UK and multiple non-core variables were imbalanced in all countries. The ESS of the HE-DMT cohort was 23% of its original size.

Conclusions: Heterogeneity is often viewed as a challenge. However, in this study heterogeneity led to deep understanding of the drug exposed populations in each country, and the influence of local treatment guidelines. Consequently, crude mortality rates in each country could be better contextualised. Such understanding is critical to informing whether a formal comparative analysis should be conducted.