(C-164) Exploring Index Event and Channeling Biases in the Evaluation of the Impact of Prior PARP Inhibitor (PARPi) Exposure on the Efficacy of Subsequent Platinum-Based Therapy Among Patients (Pts) with Ovarian Cancer (OC)

Background: Although PARPis improve outcomes in OC, a post-hoc analysis of SOLO-2 suggested that, in pts with platinum-sensitive recurrent OC and BRCA-mutations, prior PARPi exposure, compared to standard of care (SoC), may reduce the efficacy of subsequent (subseq) platinum-based therapy (PBT) [Frenel et al. Ann. Oncol. 2022]. However, index event bias from pt selection based on the initiation of subseq treatment (tx) after progression (PD) during follow-up, along with channeling bias due to non-random assignment to PBT, may have influenced this observation [Kristeleit et al. Ann. Oncol. 2022].

Objectives: This analysis explores the presence of index event and channeling biases when evaluating the impact of prior PARPi exposure on the efficacy of subseq PBT in pts with OC, using a pooled randomized control trial (RCT) database.

Methods: Pts from < 5 OC RCTs comparing PARPi to SoC were pooled from the Medidata Clinical Cloud®. The index event was initiation of subseq tx after PD during follow-up. To explore index event bias, the cause-specific hazard ratio (csHR) for the index event (PARPi vs SoC), with death as a competing risk, was estimated. Among pts with the index event, logistic regression was used to probe channeling bias by modeling the likelihood of receiving PBT, with trial arm as the key variable and interaction terms between trial arm and specific covariates. The impact of PARPi on PBT selection was reported as the adjusted relative risk (aRR), and variation in covariate impact on PBT selection across trial arms was reported as the ratio of aRR values.

Results: Among 1618 pts (PARPi arm: 66.9%; BRCA mutation/wild-type/missing: 32.7%/36.2%/31.1%; median prior lines of therapy: 2 [range: 1-10]), 1093 (67.6%) had the index event and 207 (12.8%) died without the index event. The csHR for the index event (PARPi vs SoC) was 0.61 (95% CI: 0.54-0.69; p < 0.001), indicating that pts in the PARPi arm had a 39% lower rate of the index event compared to those in the SoC arm. Among pts with the index event (PBT: 59.8%), those in the PARPi arm were more likely to initiate PBT than those in the SoC arm (aRR: 1.24 [95% CI:1.05, 1.45]). The impact of covariates on the likelihood of PBT selection varied by tx arm; the impact of a complete response to prior platinum and age > 65 yrs was lower in the PARPi arm than in the SoC arm (ratio of aRR: 0.82 [95% CI: 0.59, 0.99] and 0.75 [95% CI: 0.55, 0.90], respectively).

Conclusions: These findings suggest that there are potential biases when evaluating the impact of prior PARPi exposure on subseq therapy outcomes. While these results are based on a broader pt population than the SOLO-2-post-hoc analysis, they emphasize the importance of carefully considering methodological biases when assessing the effect of prior PARPi exposure on subseq tx efficacy.