Propensity Score Matching Methodology in ARASEC

Propensity score matching was central to the methodological rigor of the phase 2 ARASEC trial (NCT02799602),¹ enabling a valid comparison between a prospectively enrolled US treatment cohort and a historical control arm without the ethical and logistical burdens of a large phase 3 randomized trial. Neal D. Shore, MD, FACS, introduced this element of the trial design, inviting Rana R. McKay, MD, FASCO, to explain how matching was applied and why the approach was scientifically sound.

McKay described 6 key prognostic variables used for matching: age, Gleason score, disease volume (high vs low), presence of visceral metastases, de novo versus metachronous disease, and Eastern Cooperative Oncology Group performance status. Patients were matched 1:1 using stringent calipers to minimize baseline disparities between the treatment and control populations. McKay noted that an additional sensitivity analysis was performed on the unmatched total population to confirm that the propensity score algorithm itself was not introducing bias. The eligibility criteria and assessment time points in ARASEC were deliberately designed to mirror those of CHAARTED (NCT02799602), further supporting the validity of the comparison.

Shore emphasized the efficiency of the design: Enrolling more than 200 patients in a single-arm US study with tight matching parameters avoided the need for a 1000-plus patient phase 3 infrastructure and eliminated the ethical problem of assigning patients to an inferior control. Both Shore and McKay noted that the trial accrued ahead of schedule, demonstrating the feasibility of this model and its potential as a template for future oncology trial design.