3-biomarker HRD score versus individual biomarker (LOH, TAI, LST) scores in platinum treated serous ovarian cancer (SOC)

Abstract

Background: Previous studies have shown that tumors with defects in the homologous recombination (HR) pathway show improved response to DNA-damaging agents. To identify tumors likely to benefit from these therapies, we developed a 3-biomarker HR deficiency (HRD) score that is the sum of three independent measures of HRD (loss of heterozygosity (LOH), telomeric-allelic imbalance (TAI), large-scale state transitions (LST)). Previous studies have shown that the HRD score is a better prognostic marker of PFS and OS relative to the individual scores in platinum treated SOC. Here we evaluate the correlation between, and specificity of, the 3-biomarker HRD and individual scores to quantify potential false positive and negative results. Method(s): An HRD threshold (>=42) was developed in a training cohort of ovarian and breast tumors using a cut-off of 95% sensitivity to detect BRCA1/2 deficient tumors. A threshold for each HRD component was determined using the same cohort and method (LOH >=8, TAI >=10, LST >=18). The correlation between the dichotomized scores (high, low), and specificity for classifying tumors as BRCA1/2 deficient, for the HRD score and component scores was retrospectively evaluated in 859 SOC tumors. Result(s): The correlation between the HRD score and the LOH score was 0.872. There were 126 discordant scores, including 102 cases with high LOH scores and low HRD scores. These represent potential false positives based on LOH alone. Similarly, 24 samples with low LOH scores had high HRD scores (potential false negatives). Similar behavior was observed for TAI (correlation coefficient 0.905, 95 discordant scores) and LST (correlation coefficient 0.941, 88 discordant scores). Specificity was highest for the HRD score in both the training and test cohorts (0.897 and 0.796) compared to any of the component scores (LOH: 0.624 and 0.668; TAI: 0.766 and 0.690; LST: 0.759 and 0.672). Conclusion(s): Here we show that the use of a single HRD biomarker may misinform treatment decisions in SOC relative to the combined 3-biomarker score. The combined HRD assay, which has been validated on FFPE SOC tumor tissue, warrants evaluation in a prospective study sample set in a rigorously validated laboratory.