PROPAN: Bayesian dose-finding of propranolol combined with metronomic fixed-dose cyclophosphamide based on bivariate efficacy-tolerability outcomes in patients with locally advanced or metastatic angiosarcoma (LAMA): A phase I-II sequential trial by the French Sarcoma Group.

Abstract

e23542 Background: Angiosarcoma is a rare, aggressive vascular malignancy with limited treatment options. Preclinical data suggest beta-adrenergic blockers like propranolol may have anti-tumor effects. Only one window-of-opportunity trial (Embaby et al., 2024) has prospectively evaluated the activity of propranolol in angiosarcoma. Propranolol's efficacy is believed to fluctuate dose-dependently. We used an innovative phase I-II dose-finding method integrating toxicity, efficacy, and a time-to-event parameter (Rivière et al., 2018), tailored to small samples and optimizing recruitment. The PROPAN trial aimed to determine the optimal dose of propranolol combined with metronomic cyclophosphamide in patients with LAMA, using a sequential dose-finding design based on a bivariate continual reassessment method. Methods: Patients were enrolled in cohorts of 2, with sequential dose adaptations made for the next cohort among three dose levels (80, 120, or 160 mg/d) based on toxicity (grade ≥3 at 1 month per NCI-CTCAE v4.0) and efficacy (non-progression rate at 3 months per RECIST 1.1). The safe most successful dose (sMSD) was determined using a Bayesian model (dfmta package), balancing an non progression efficacy of 15, 30 and 50 % for respectively 80, 120 and 160 mg/j and ≤25% toxicity). This approach determined the next adapted dose based only on toxicity assessed early compared to efficacy, reducing waiting time between cohorts. The trial ended early if all doses failed efficacy or exceeded toxicity targets, with a maximum of 24 patients. Results: Fifteen patients were enrolled, with 12 used for dose optimization. 2 patients were treated in the second line, and the others in the third lines Efficacy data were unavailable for 3 in the final model. One grade 3 toxicity (dyspnea at 120 mg/d) was reported, and 4 patients met the non-progression endpoint at 3 months (3 at 80 mg/d, 1 at 120 mg/d). The final model identifies 80 mg/day as the optimal dose minimizing toxicity, with no efficacy gain at higher doses. Median progression-free survival was 3 months (0.7-6). Conclusions: This trial confirms a favorable safety profile but limited efficacy. This innovative method demonstrates feasibility for rare diseases. However, dose adjustment based on early toxicity assessment optimizes patient recruitment but risks over-reliance on toxicity data for dose optimization. Clinical trial information: NCT02732678 .