A SARC multicenter phase III study of gemcitabine (G) vs. gemcitabine and docetaxel (G+D) in patients (pts) with metastatic soft tissue sarcomas (STS)

Abstract

9514 Background: We sought to compare the response of metastatic STS to G vs. G+D in a phase III study using a novel Bayesian adaptive randomization strategy. Methods: Entry criteria included non-GIST STS diagnosis, age > 10, measurable disease, ≤ 3 prior regimens, normal organ function, and Grade (Gr) ≤ 1 neuropathy. Pts were stratified by histology (leiomyosarcoma [LMS] vs. other), and history of prior pelvic radiation (PPR) or not, yielding 4 diagnostic subgroups. Therapy was G 1200 mg/m2 (over 120 min, d1+d8) or G 900 mg/m2 (over 90 min, d1+d8) and docetaxel (100 mg/m2 d8) q21d; all pts received (peg)-filgrastim starting d9. 25% dose reductions were employed for PPR or toxicity. The primary endpoint was tumor response (CR, PR, or 24+ weeks stable), using a double-blind Bayesian adaptive randomization procedure to incrementally assign more pts to the superior treatment arm, while accounting for possible treatment-subgroup interactions. Although this study is not based on a 1:1 randomized phase III design, enrolling ∼120 pts would show a difference between a response rate (RR) of 10% for G and 30% with G+D with a power of 0.8 and two-sided alpha of 0.05. Results: 122 pts were randomized; 119 pts had evaluable outcomes. Median number of prior regimens was 1; median number of cycles was 4 (range 1–26). 49 pts were adaptively randomized to G and 73 to G+D, indicating superiority of G+D. Of 119 evaluable pts, 27% (G) and 32% (G+D) showed response as defined. The odds of pts with LMS receiving G+D instead of G increased from 1:1 at the start of the study to ∼6:1 at its completion. RR was 10% (G) vs. 16% (G+D) (p=0.15, Fisher exact test). Consistent with the RR outcomes, progression free survival (PFS) was 6.2 m (G+D, K-M 95%CI 3.6–8.8) vs. 2.6 m (G, 95%CI 2.0–3.2). Overall survival (OS) was 18.0 m (G+D, K-M 95%CI 11.7–24.1) vs. 11.2 m (G, 95%CI 6.8–15.5). Pts receiving G+D had a higher rate of discontinuation due to toxicity than those receiving G (p<0.01, log rank). Conclusions: We observed substantial clinical activity in this pre-treated STS population. G+D yields superior PFS and OS to G, but at the price of increased toxicity. Adaptive randomization is an effective method that substantively reduces the number of pts receiving inferior therapy. [Table: see text]