OBJECTIVES: Endothelin receptor antagonists (ERAs) are recommended for patients with pulmonary arterial hypertension (PAH), either as monotherapy or in combination with a phosphodiesterase type 5 inhibitor (PDE-5i; initially or sequentially). Head-to-head trials of ERAs are lacking. We compared the effectiveness of macitentan versus ambrisentan in monotherapy and combination therapy, utilizing Bucher’s indirect treatment comparison (ITC) method with individual patient data from the SERAPHIN study and aggregate data from the AMBITION study.

METHODS: Primary clinical efficacy outcomes in both trials were composite morbidity/mortality endpoints albeit with different components; these were harmonised using a common modified clinical failure (MCF) endpoint. An AMBITION-like analysis set (ALS) was derived from the SERAPHIN population by applying revised AMBITION exclusion criteria. Additional harmonisation measures included balancing between-study treatment exposure, applying AMBITION haemodynamic inclusion criteria to ALS (hALS), and identifying patients with time since diagnosis at baseline ≤6 months (incident). ITC analysis provided hazard ratios (HRs) and 95% confidence intervals (CIs) for macitentan versus ambrisentan.

RESULTS: Point estimates indicated reduced risk of an MCF event with macitentan versus ambrisentan, by an increasing margin as additional harmonisation steps were applied. Exposure-time–balanced HRs [95% CI] were 0.22 [0.09, 0.58] in monotherapy and 0.78 [0.38, 1.59] in combination therapy in the ALS set, decreasing to 0.18 [0.07, 0.50] and 0.54 [0.26, 1.16], respectively, in the hALS set. With ALS-incident patients, HRs were 0.41 [0.12, 1.38] and 0.38 [0.09, 1.60], respectively, indicating a 59% and 62% reduction in risk, respectively, of an MCF event with macitentan versus ambrisentan.

CONCLUSIONS: ITC analysis suggests reduced risk of an MCF event with macitentan versus ambrisentan, after application of harmonisation measures. Results in incident patients (closest to AMBITION newly diagnosed patients) showed wider confidence intervals due to fewer events, but a clinically relevant difference.