Alpelisib plus olaparib failed to beat chemotherapy in platinum-resistant ovarian cancer

What was tried

Novartis ran EPIK-O, also registered as ENGOT-OV61 (NCT04729387), a Phase 3, multicenter, randomized, open-label, active-controlled study of the alpha-selective PI3K inhibitor alpelisib (BYL719) combined with the PARP inhibitor olaparib, compared against single-agent cytotoxic chemotherapy, in platinum-resistant or refractory high-grade serous ovarian cancer with no germline BRCA mutation detected (ClinicalTrials.gov, NCT04729387). Patients were randomized 1 to 1, with alpelisib given at 200 mg orally once daily and olaparib at 200 mg orally twice daily, against investigator choice of weekly paclitaxel or pegylated liposomal doxorubicin (ClinicalTrials.gov, NCT04729387). The primary endpoint was progression-free survival by blinded independent review committee using RECIST 1.1 (ClinicalTrials.gov, NCT04729387). The study opened on 2021-07-22 and reached an actual enrollment of 358 participants, 180 to the combination and 178 to chemotherapy (ClinicalTrials.gov, NCT04729387). The primary completion date was 2023-04-21, and the record was later set to terminated with the posted reason "Sponsor Decision" (ClinicalTrials.gov, NCT04729387).

The biological hypothesis

The rationale rested on documented crosstalk between the PI3K pathway and homologous recombination repair. PI3K inhibition lowers homologous recombination capacity and can sensitize BRCA-proficient tumors to PARP inhibition, which is the basis for combining alpelisib with olaparib in cancers that lack a BRCA mutation (Lancet Oncology, Konstantinopoulos et al., 2019, doi:10.1016/S1470-2045(18)30905-7). PIK3CA is a frequently altered oncogene, and in Open Targets its association with ovarian carcinoma reaches an overall score of 0.3478, built mainly from affected-pathway evidence at 0.5499 and somatic mutation evidence at 0.4559 rather than from germline genetic association (Open Targets Platform, PIK3CA ENSG00000121879, EFO_0001075). A Phase 1b dose-escalation and expansion study set the recommended doses and reported preliminary antitumor activity for the alpelisib plus olaparib combination in epithelial ovarian cancer, which motivated the Phase 3 test (Lancet Oncology, 2019, doi:10.1016/S1470-2045(18)30905-7). The same combination was carried into triple-negative breast cancer on the identical synthetic-lethality logic (Clinical Cancer Research, Batalini et al., 2022). The EPIK-O design was published before the readout (Future Oncology, 2022).

What actually happened

The combination did not separate from chemotherapy. At a median follow-up of 9.3 months, median progression-free survival by blinded review was 3.6 months with alpelisib plus olaparib against 3.9 months with chemotherapy, with a hazard ratio of 1.14 (95% CI 0.88 to 1.48) and P = .84 (JCO, Konstantinopoulos et al., 2025, ASCO Post, August 2025). The posted results record the same medians, 3.6 months (95% CI 3.35 to 4.27) and 3.9 months (95% CI 3.68 to 5.36) (ClinicalTrials.gov, NCT04729387). Confirmed overall response rate was 15.6 percent (95% CI 10.6 to 21.7) for the combination and 13.5 percent (95% CI 8.8 to 19.4) for chemotherapy, a difference inside the noise (ClinicalTrials.gov, NCT04729387). The point estimate for the primary endpoint favored chemotherapy, so the study missed on both direction and significance. Novartis then terminated the program under the posted reason "Sponsor Decision" (ClinicalTrials.gov, NCT04729387). The safety profile of alpelisib was consistent with its class. Across 6822 openFDA FAERS reports for PIQRAY, the leading terms were rash at 1186 and hyperglycaemia at 1161, the expected consequences of systemic PI3K-alpha inhibition (openFDA FAERS, queried 2026-06-06).

Failure mechanism, best guess

The most likely failure mode was a translational mismatch between a preclinical sensitization signal and an unselected, heavily pretreated population. The PI3K to homologous recombination link is real in models, but platinum-resistant high-grade serous ovarian cancer is genomically chaotic and dominated by platinum and PARP resistance mechanisms that the combination did not address. Enrolling patients without a germline BRCA mutation removed the context where PARP inhibition has its strongest effect, and adding alpelisib was asked to manufacture a synthetic-lethal vulnerability that the tumor biology did not reliably present. The hazard ratio of 1.14 shows the combination added toxicity without converting the preclinical signal into clinical benefit (JCO, 2025). The genetic anchoring supports this read. The Open Targets association for PIK3CA in ovarian carcinoma rests on somatic and pathway evidence rather than causal germline genetics, so the program leaned on a pathway rationale rather than human evidence that PIK3CA activity drives progression in this resistant setting (Open Targets Platform, PIK3CA ENSG00000121879, EFO_0001075).

How to prevent this next time

Two quantitative tools would have reframed the go decision before a 358-patient Phase 3.

First, a historical base-rate adjustment. Oncology carries the lowest likelihood of approval of any therapeutic area, with a Phase 1 likelihood of approval near 5.3 percent and the steepest attrition at the efficacy-defining stages across the 2011 to 2020 record (Clinical Development Success Rates 2011-2020, BIO, Informa Pharma Intelligence, QLS Advisors). A small single-arm Phase 1b activity signal in an all-comers resistant population should have been discounted hard against that base rate before committing to a randomized Phase 3, and the prior should have demanded a biomarker that identifies tumors with an actual PI3K-driven repair deficit.

Second, a Bayesian predictive-probability framework with an explicit interim futility gate. The predictive probability of success integrates the probability of a positive final result over the posterior for the treatment effect given interim data.

PP(success)=∫θ​P(final success∣θ,interim data)⋅π(θ∣interim data)dθ

A prior centered on the modest and uncertain incremental effect of adding PI3K blockade to PARP inhibition in non-BRCA disease, combined with a pre-specified stop boundary, would have halted enrollment at the first interim once the predictive probability fell below a useful threshold rather than randomizing all 358 patients. The same machinery supports a biomarker-enriched interim that tests the sensitization claim where it is biologically strongest.

The single highest leverage change would have been to enrich enrollment for tumors with a measurable PI3K-pathway-driven homologous recombination deficit and to gate continuation on a pre-specified Bayesian predictive probability rather than testing the combination in an unselected non-BRCA population.

What this means for similar programs

The result is a caution for combination programs that justify a randomized trial on a pathway-crosstalk rationale rather than a patient-selection biomarker. PI3K and PARP combinations remain mechanistically plausible, but EPIK-O shows that plausibility does not survive an unselected resistant population. The same alpelisib plus olaparib pairing was pursued in triple-negative breast cancer on identical logic (Clinical Cancer Research, 2022), and any continuation should carry a prospective enrichment strategy. In the Claidex graph PIK3CA now carries a Mechanism Risk Score of 42 (yellow), reflecting a Phase 3 efficacy failure layered on a target whose ovarian association is somatic rather than germline (Claidex failure graph, 2026-06-06).

Open questions

Would a biomarker-selected population, for example tumors with PIK3CA activation plus a functional homologous recombination deficit signature, have separated from chemotherapy? Did the open-label design and the use of physician-choice chemotherapy as comparator affect the progression-free survival reads? Was the 200 mg alpelisib dose, constrained by hyperglycaemia and rash, sufficient to suppress PI3K signaling in tumor tissue? These cannot be resolved from the posted summary data alone.

Sources

1. • ClinicalTrials.gov, NCT04729387 (protocol, posted results): https://clinicaltrials.gov/study/NCT04729387 - Konstantinopoulos PA et al. Primary Analysis of EPIK-O/ENGOT-ov61. Journal of Clinical Oncology 2025.- ASCO Post. Alpelisib Plus Olaparib vs Single-Agent Chemotherapy in Platinum-Resistant or Refractory HGSOC. August 2025. - Konstantinopoulos PA et al. Olaparib and alpelisib in epithelial ovarian cancer, Phase 1b. Lancet Oncology 2019.- EPIK-O/ENGOT-OV61 study design. Future Oncology 2022.- Batalini F et al. Alpelisib plus olaparib in advanced triple-negative breast cancer, Phase 1b. Clinical Cancer Research 2022.- Open Targets Platform, PIK3CA (ENSG00000121879), ovarian carcinoma (EFO_0001075): https://platform.opentargets.org/target/ENSG00000121879 - ChEMBL v34, alpelisib CHEMBL2396661, olaparib CHEMBL521686 - openFDA FAERS, PIQRAY, queried 2026-06-06 - BIO, Informa Pharma Intelligence, QLS Advisors. Clinical Development Success Rates 2011-2020.