Progression-free survival outcomes of PARP inhibitors in ovarian cancer: an exploratory analysis of treatment heterogeneity based on organ vulnerability.

Background: PARP inhibitors are standard maintenance therapies for ovarian cancer (OC), but evidence regarding how baseline physiological vulnerability relates to heterogeneity in comparative treatment effectiveness-beyond molecular biomarkers-remains limited. This study applied a causal inference framework, benchmarked against established molecular associations, to explore whether organ vulnerability is associated with variation in progression-free survival (PFS) between PARP inhibitors. Methods: In this retrospective study of 604 OC patients, we constructed an interpretable 0-3 organ vulnerability score (OVS) using routine pre-treatment laboratory indicators reflecting physiological reserve. A benchmark analysis in 280 patients with complete BRCA status evaluated whether the framework could reproduce established prognostic associations under real-world conditions. G-computation-based counterfactual survival standardization and causal forest analyses were applied to the full cohort to explore OVS-associated treatment heterogeneity, with directional reproducibility assessed in an independent external cohort (n = 58). Results: The estimated association between BRCA mutation status and PFS was directionally consistent with prior clinical evidence (HR = 0.685, 95% CI: 0.490-0.959; P = 0.028). In the full cohort, a significant treatment-by-OVS interaction was observed (interaction HR = 0.488, 95% CI: 0.293-0.813; P = 0.006). Among patients with low vulnerability (OVS 0-1, n = 479), olaparib was associated with a modest PFS advantage over niraparib (HR = 0.744, P = 0.030). In patients with high vulnerability (OVS 2-3, n = 125), the estimated relative benefit associated with olaparib was more pronounced (HR = 0.363, P < 0.001); however, these estimates are model-dependent and should not be interpreted as definitive comparisons. The directional heterogeneity pattern was qualitatively reproduced in the external cohort. Exploratory heterogeneity analysis further identified individual-level variation in estimated treatment response, including a subset with estimated relative benefit favoring niraparib. Conclusions: This retrospective exploratory analysis suggests that baseline physiological vulnerability may be associated with heterogeneity in the relative effectiveness of PARP inhibitors. These findings are hypothesis-generating and lay the groundwork for prospective investigation into physiological vulnerability as a correlate of heterogeneous treatment response in real-world oncology populations.