(BSPOT28) How far can we go with Federated Meta-Analytic Approaches in Rare Diseases and Special Populations: A Case Example in Pediatric Pulmonary Arterial Hypertension for Regulatory Purposes

Background: In rare diseases and special populations, reliable, transparent, reproducible real-world evidence [RWE] generation is often hindered by very few patients scattered across regions and healthcare settings. The ability to conduct federated analyses of fit-for-purpose data sources can help to overcome this issue.

Objectives: To describe patient characteristics (age, gender, growth) and outcomes (general development: puberty, sexual maturation; other: hospitalization, transplantation, liver enzymes, death) in a cohort of pediatric pulmonary arterial hypertension [PAH] patients [group A] and a sub-cohort treated with bosentan [group B]: EU PASS ENCEPP/SDPP/4523

Methods: 4 pulmonary hypertension registries were connected into a federated analytical framework, a multinational (#A:495, #B:256), a Dutch (#A:27, #B:18), a French (#A:5, #B:4), and a US ( #A:59, #B:19) one . Individual database analysis was complemented by meta-analysis of outcomes, except hospitalization which was recorded differently across the registries. The French registry was excluded from the meta-analysis due to either zero or imprecise outcome event-rate estimates. Random effects Poisson regression was chosen for overall outcome incidence rate estimation, to ensure good representation of large and small registries. Fixed effects analysis was performed to verify the robustness of the findings. Results were interpreted considering the observed between-registry heterogeneity and variability in data collection.

Results: 581 unique group A patients were included in the meta-analysis, 283 of whom were also in group B. The findings consistently suggested that A and B were similar: Most patients were ≤ 8 years old (B: 62%; A: 65%) and female (B: 59%; A: 58%). Outcomes were similar for growth and sexual maturation, including, e.g., median estimated age of puberty onset (B: 14.2 years (SD=0.3); A: 14.3 years (SD=0.3)). The random and fixed effects meta-regression both indicated that outcome event rates were slightly, but never significantly, higher in B than in A. For instance, the event rate of transplantation was 1.7 events/100 person-years in B and 1.2 in A, with overlapping confidence intervals of [0.8-2.8] and [0.6-1.7], respectively. There were no new safety signals revealed from group B.

Conclusions: The concept of federated analytics in a network of disease-specific databases broadens the range of possibilities in rare disease RWE generation. The approach in this PASS was accepted by EMA’s Pharmacovigilance Risk Assessment Committee. In 2019, only 10 years after approval of the bosentan pediatric formulation, the PASS was concluded as fulfilled. An estimated time for a de-novo registry would have been >20 years. The study results informed regulatory dossiers in the US and China.