Background
Homologous recombination DNA repair deficiency (HRD) is found in approximately 50% of high-grade serous ovarian carcinomas and associated with sensitivity to platinum and poly (ADP-ribose) polymerase (PARP) inhibitors. HRD status is becoming an important clinical indicator, with the potential to change treatment. HRD is assessed by several methods including whole-genome copy-number alterations determined by single nucleotide polymorphism (SNP) arrays or next generation sequencing (NGS). A limitation of applying these methods into routine practice is tumour sample content and quality. We investigated HRD testing in routine clinical samples that are formalin-fixed and often collected after neoadjuvant chemotherapy, which can limit the quality of testable DNA.
Methods
Cases were selected from INOVATe1, a molecular profiling program that recruited ovarian cancer patients between 2016-2020, with testing including NGS, methylation and copy number profiling for HRD status. Of 506 eligible patients, 204 with frozen tumour tissue were selected for assessment of HRD, and of these, HRD was compared in 160 cases with both frozen and formalin-fixed paraffin-embedded (FFPE) tissue available.
Results
In cases with frozen tissue, a high confidence HRD result was obtained in 185/204 (90.6%), 92% of which were concordant with NGS and methylation results.
Cases with both frozen and FFPE tissue for testing (n=160) were examined to determine DNA quality and compare HRD status in matched samples. Mean block age was 22 months (range, 1-130 months), and 76% (n=121) of samples were chemotherapy-naïve. The average DNA integrity number (DIN) of fixed samples was 3.68 (range 1.5-6.6).
After SNP array testing, results from 101 (63%) FFPE samples passed bioinformatic processing or quality metrics, of which sensitivity for detection of HRD was 90% and accuracy was 88%. The strongest predictor of a successful HRD determination from FFPE was DNA quality, although this was not an absolute predictor of test success. Increasing block age and prior chemotherapy exposure were associated with test failure (p-value <0.05).
Conclusions
In conclusion, HRD testing on frozen samples was successful, although limited by the availability of the frozen material in routine settings. Testing on FFPE samples using whole-genome SNP arrays is possible, although there is a significant rate of sample failure. These failures are apparent by bioinformatic analysis, and test performance is excellent on samples which pass all quality metrics. HRD status can be determined in FFPE tissues to complement sequencing, although adequate tumour sampling prior to chemotherapy is needed to increase the likelihood of an interpretable result.