Clinical assessment of HRD scoring in relation to PARPi response in terms of statistical guidance on reporting results for CDx tests evaluation.

e17597 Background: Homologous recombination deficiency (HRD) is a tumor characteristic that can be assessed via different biomarkers such as individual mutations in BRCA1 and BRCA2 genes, and/or other HRR genes and genomic instability. We assessed a combined HRD score in three measures; BRCAs mutations, HRR mutation (HRRm) status, and genomic instability scoring via a specific bioinformatic algorithm of the testing panel to establish clinical validation. Methods: We tested 157 ovarian cancer patients’ FFPE samples to define HRD score, BRCA1/2 mutation and other HRRm status, to camper the responses to PARP inhibitors. For the HRD scoring, all the samples that were previously tested with alternative HRD tests and the cases had an indication of PARPi also tested via HRDInfo (CE-IVD) Panel for clinical diagnostics in this study. The PPA, NPA and OPA have been measured according to the FDA regulations of overall agreement for a follow-on CDx to show clinical evidence. Results: Depending on the three levels of testing strategy, BRCA1-2 mutation status in relation to the clinical response for PARPi has the lowest positive percent agreement (PPA) with 63.41% while HRD score has the highest with 97.62% that is higher than the any alternative HRD tests. All the PPA, NPA and OPA measures showed in the Figure for BRCA1-2 status and Genomic Instability stand alone and all BRCA1-2 plus other HRR genes plus Genomic Instability Status (GIS) to define HRD scoring. Conclusions: Even though the testing rate of HRD tests are still low all around the world and limited to specific cancers, there is an increasing requirement for different types of cancers. However, most alternative HRD tests have high overall concordance but fail to reach FDA CDx test requirements. Thus, this study showing the clinical evidence in the treatment of PARPi response data together with high PPA, NPA and OPA measures of HRDInfo panel and its bioinformatics algorithm. [Table: see text]