Carcinosarcomas (CS) are rare and highly aggressive gynecological carcinomas with poor outcome. The median Progression-Free-Survival (PFS) in relapse after platinum-based chemotherapies (CT) is less than 4 months. CS were excluded from Keynote755 trial. Since CS show high DNA damage response activity and potentially a high tumor mutational load resulting in neo-antigens, a synergy between PARPi and anti-PD1 is expected. This is an open-label (NCT03651206) study with a two-stage design. In step 1, patients (pts) were randomized (2:2:1) to receive either niraparib (N), niraparib and dostarlimab (ND) or chemotherapy (CT). Randomization was stratified by number of previous CT lines, FIGO stage, localization, and performance status. The primary objective was to select the best experimental strategy between N and ND using Response Rate (RR) at 16 weeks (RECIST1.1). Secondary endpoints included best Objective RR, disease control rate (DCR), safety, OS & QoL. After an interim analysis at the end of phase II, data will be reviewed by an Independent Data Monitoring Committee, and eventually allow the enrolment in subsequent phase III (part 2). 64 pts with recurrent or progressing endometrial or ovarian CS after at least one line of platinum-based CT were randomized in the phase II (N=26, ND=25, CT=13). Median age was 70 years (range 34-84). The 16w-RR was 3.8%, 12.0% and 15.4% respectively. The ORR was 3.8%, 20% and 15.4% and the 8w-DCR was 26.9%, 52% and 30.8% in arm N, ND and CT, respectively. With a median follow-up of 11.2 months, median PFS (months) was 2.0 (95%CI, 1.9-2.2), 2.7 (95%CI, 1.9-3.7), 1.9 (95%CI, 1.7-3.6) and median OS (months) was 6.7 (95%CI, 3.8-9.6), 6.3 (95%CI, 3.9-12.4), 4.5 (95%CI, 3.0-NE) in arm N, ND and CT respectively. %AE grade >3 were 69.2%, 68% and 69.2% respectively. ROCSAN step 1 did not met primary endpoint for 16W RR (> 20%), however the DCR, median OS and safety suggest some benefice for ND compared to CT in this very rare and poor prognostic population.