Three-Biomarker Score More Accurately Predicts Ovarian Cancer Outcomes

The score was linked with statistically significant hazard reductions of 30% to 35% for progression-free (PFS) and overall survival. The individual components of the HRD score had lesser, but still significant, associations with PFS and overall survival, and all lost statistical significance in a bivariate comparison against the HRD score, as reported at the 2016 Society of Gynecologic Oncology Annual Meeting.

“A combined three-biomarker HRD score is a superior predictor of survival of platinum-treated ovarian cancer patients than any of the individual biomarker components,” said Gordon Mills, MD, PhD, chair of Systems Biology at the University of Texas MD Anderson Cancer Center. “The HRD score is currently being evaluated, in combination with BRCA1/2 mutation status, as a predictor of response to platinum and PARP inhibitors in multiple prospective clinical studies.”

HRD is a measure of genomic instability that predicts sensitivity to platinum chemotherapy and PARP inhibitors. However, factors other than mutations may cause HRD, said Mills.

The HRD score comprises telomeric-allelic imbalance (TAI), large-scale state transitions (LST), and loss of heterozygosity (LOH). Previous studies showed the HRD score had better predictive accuracy for treatment response than did any of the individual markers.

The score was evaluated in a training cohort of 1058 patients with chemotherapy-naïve ovarian and breast cancers. Using a 95% sensitivity for detection of BRCA1/2-deficient tumors, investigators determined that an HRD cutoff of 42 yielded the best predictive results. In the same 1058-patient cohort, investigators also determined 95% sensitivity cutoffs for the individual biomarkers: LOH ≥8, TAI ³10, and LST ³18.

Mills and colleagues retrospectively applied Cox proportional hazard models to an independent study cohort of 858 ovarian tumor samples from patients involved in clinical trials of platinum-based chemotherapy. Dichotomized scores (low, high) for LOH, TAI, and LST were tested individually and then compared to the HRD score in bivariate models.

Univariate analysis of the dichotomous scores showed that the three-biomarker HRD score achieved statistical significance for PFS (HR 0.66,P= 2 x 10^-6) and overall survival (HR 0.55, 1 x 10^-8). Also by univariate analysis, each of the individual components of the HRD score achieved statistical significance for PFS (LOH: HR 0.70,P= 5 x 10^-5; TAI: HR 0.71,P= 9 x 10^-5; LST: 0.68, 1 x 10^-5) and overall survival (LOH: HR 0.64,P= 8 x 10^-6; TAI: HR 0.61,P= 6 x 10^-7; LST: HR 0.60,P= 3 x 10^-7).

“Although the individual markers were significant predictors of outcomes for platinum-treated patients, the HRD score was more significant than any of the component scores,” said Mills.

In the bivariate analysis comparing the HRD score to the individual components, the composite score remained a significant predictor of PFS and overall survival, but none of the three biomarkers retained statistical significance.

The HRD score was associated with hazard ratios of about 0.70 for PFS in all three comparisons and statistical significance ranging fromP= .045 toP= .007. The individual components were associated with hazard ratios of approximately 0.9 and statistical significance ofP= .8 toP= .3.

The bivariate analysis of overall survival yielded hazard ratios of about 0.60 for the HRD score and statistical significance of P = .008 toP= 2 x 10-4, as compared with hazard ratios of about 0.9 for the three individual markers (P= .4 for each of the three individual markers).

Analysis of potential false-positive and false-negative results showed that the individual biomarkers had an overall diagnostic accuracy of 85% to 90% versus the composite HRD score.

References

1. Mills G, Timms KM, Reid J, et al. Homologous recombination deficiency (HRD) score shows superior association with outcome compared to its individual score components (LOH, TAI, and LST scores) in platinum-treated serous ovarian cancer. Presented at: 2016 SGO Annual Meeting. March 19-22, 2016. San Diego, CA. Abstract 2.