Efficacy prediction for progression-free survival (PFS) and overall survival (OS) by genomic instability score (GIS) cutoffs in patients (pts) with advanced ovarian cancer (aOC): Post hoc results from the phase 3 PRIMA/ENGOT-OV26/GOG-3012 trial.

5565 Background: GIS is a key component of a companion diagnostic assay for determining homologous recombination deficiency (HRD) status to identify pts with aOC who may benefit from maintenance therapy. While the standard GIS cutoff to classify tumors as homologous recombination-deficient (HRd) is ≥42, this study analyzed PRIMA final analysis data to see if a lower cutoff of ≥33 based on the VELIA/GOG-3005 trial could identify additional pts who experienced a PFS benefit with niraparib (nir) first-line maintenance (1LM) vs placebo (PBO). Methods: Pts with aOC were randomized 2:1 to nir or PBO 1LM. Tumor HRD status and GIS were assessed using the myChoice HRD test (Myriad Genetics). In pts with GIS data, tumor GIS distribution was assessed in the overall population and by HRD/ BRCA status. Investigator-assessed PFS at GIS cutoffs <33 vs ≥33 and <42 vs ≥42 was evaluated using Kaplan-Meier methods with Cox proportional hazards modeling (cutoff, 08Apr2024). Two statistical utility tests were performed for each cutoff: 1) to evaluate each cutoff’s potential to add incremental value to the explanation of pt response on top of treatment, and 2), in nir-treated pts, to assess each cutoff’s ability to predict long-term PFS (≥3 y). OS was assessed by increasing GIS cutoffs. Results: In the overall population (n=568), GIS ranged from 0 to 95, and 64.3% of pts had GIS ≥33 (GIS groups: 0 to <33, n=203; ≥33 to <42, n=67; ≥42, n=298). Of pts with BRCA mutation, 9.4% (21/223) of pts had GIS <42. Nir 1LM produced a significant PFS benefit vs PBO across all GIS cutoffs, with greater benefit observed in higher GIS groups (Table). The GIS ≥33 cutoff improved the Cox model fit vs the ≥42 (ie, explaining more PFS variability and showing more powerful association with PFS after treatment adjustment). In nir-treated pts with GIS data (n=387), the ≥33 cutoff increased sensitivity (0.8936 > 0.7872), identifying 10.6% more patients with PFS ≥3 y, but lowered specificity (0.4369 < 0.5734) vs the ≥42 cutoff. Higher GIS cutoffs were also associated with more favorable nir 1LM on OS effects (lower hazard ratios). Conclusions: In PRIMA, GIS existed along a spectrum, and nir 1LM resulted in a PFS benefit vs PBO across all GIS cutoffs. Pts with higher GIS experienced greater PFS benefit than pts with lower GIS. The GIS ≥33 cutoff identified more long-term PFS responders than the GIS ≥42 cutoff. Results suggest that GIS cutoffs of ≥33 used for HRd status determination may optimize identification of pts with aOC who may benefit from nir 1LM. Clinical trial information: NCT02655016 . Tumor GIS Investigator-assessed PFS <33 ≥33 <42 ≥42 Nir vs PBO, n 133 vs 65 283 vs 141 169 vs 80 247 vs 126 Hazard ratio (95% CI) a 0.69 (0.50–0.96) 0.54 (0.42–0.68) 0.67 (0.50–0.89) 0.51 (0.40–0.66) P value 0.026 <0.001 0.007 <0.001 a Stratified Cox proportional hazards model. Pts with GIS data included.