PARP Inhibitor Sensitivity in Tumors Harboring Non-BRCA Homologous Recombination Gene Alterations: Current Evidence Across Ovarian, Breast, Prostate, and Pancreatic Cancers.

Synthetic lethality: pathogenic variants in homologous recombination genes other than BRCA1/BRCA2 — PALB2, RAD51C, RAD51D, ATM, CHEK2, CDK12 — impair homologous recombination repair to unequal degrees, hence a gene- and tumour-type-dependent sensitivity to PARP inhibitors.

This narrative review assesses the predictive value of pathogenic variants in homologous recombination genes other than BRCA1/BRCA2 as biomarkers of PARP inhibitor sensitivity in ovarian, breast, prostate and pancreatic cancers. The search, conducted between October 2023 and December 2025, first identified pivotal PARP inhibitor trials, then covered PubMed, Embase, Web of Science and Google Scholar; 17 clinical studies and multiple preclinical reports were analysed for genomic frequency, association with homologous recombination deficiency and treatment response. Preclinical models consistently showed increased PARP inhibitor sensitivity, but clinical evidence was heterogeneous: PALB2 showed the strongest and most consistent association, RAD51C and RAD51D showed clinically meaningful activity particularly in ovarian cancer, while evidence remained limited or inconsistent for ATM, CHEK2, CDK12 and several other genes. The authors attribute this variability to differences in gene function, biallelic inactivation, variant type and current limitations of homologous recombination deficiency companion diagnostics.

Synthesis written by Geno'X. For the full original abstract, please refer to the source publication.

The review usefully breaks the convenient equation between "homologous recombination gene" and "PARP inhibitor eligibility": PALB2 behaves like BRCA1/BRCA2, ATM and CHEK2 much less so, and presenting these genes as a homogeneous group in the cancer genetics clinic is hard to defend. Its limitation is the format: a narrative review without a registered protocol, resting on 17 clinical studies of disparate size and design, from which no pooled estimate of benefit can be drawn. In practice it helps rank what we tell patients, not decide treatment gene by gene.

Analysis by Dr Thibaut Benquey