A Genomic Predictor of Response and Survival Following Taxane-Anthracycline Chemotherapy for Invasive Breast Cancer
JAMA. 2011 May 11;305(18):1873-1881, C Hatzis, L Pusztai, V Valero, DJ Booser, L Esserman, A Lluch, T Vidaurre, F Holmes, E Souchon, H Wang, M Martin, J Cotrina, H Gomez, R Hubbard, JI Chaco´n, J Ferrer-Lozano, R Dyer, M Buxton, Y Gong, Y Wu, N Ibrahim, E Andreopoulou, NT Ueno, K Hunt, W Yang, A Nazario, A DeMichele, J O’Shaughnessy, GN Hortobagyi, WF Symmans
Tumor biopsy samples obtained prior to systemic therapy were used for genomic studies to identify potential predictors of treatment outcome. Predictors were identified using gene expression microarrays to compare expression in samples from patients with higher risk in two responder groups (pathologic complete response or minimal residual cancer burden [RCB-I] vs moderate or extensive residual cancer burden [RCB-II/III]) in ER-negative and -positive subgroups.
Biopsy samples were obtained from 310 patients in the discovery cohort and 198 patients in the validation cohort. Patients in the validation cohort had clinical response rates of 25% (pathologic complete response) and 30% (pathologic complete response or RCB-1), and distant relapse-free survival (DRFS) rates at 3 years of 79%.
After excluding patients with predicted endocrine sensitivity, the chemopredictive test algorithm had a positive predictive value (PPV) of 56% (95% CI, 31%–78%). In the 28% of patients predicted to be treatment sensitive, the 3-year DRFS was 92% (95% CI, 85%–100%), with an absolute risk reduction (ARR) of 18% (95% CI, 6%–28%). In the patients predicted to be treatment insensitive, the 3-year DRFS was 75% (95% CI, 67%–82%), with a PPV of 25% and odds ratio for relapse of 4.01 (95% CI, 1.60–20.4). There was a significant association of predicted sensitivity to treatment and improved DRFS (P = .002). Among patients predicted to be treatment sensitive, the diagnostic likelihood ratio for occurrence vs absence of 3-year distant relapse or death was 0.33 (95% CI, 0.07–0.72).
Patients predicted to be treatment sensitive had 3-year DRFS (92%) similar to that of patients who achieved pathologic complete response after completing neoadjuvant chemotherapy (93%). In addition, patients predicted to be treatment insensitive had 3-year DRFS (75%) identical to that of patients with residual disease.
In the 30% of patients with ER-positive disease who were predicted to be treatment sensitive, 3-year DRFS was 97% (95% CI, 91%–100%) and ARR was 11% (95% CI, 0.1%–21%). In patients with ER-positive disease who were predicted to be treatment insensitive, the PPV for 3-year DRFS was 14% (95% CI, 6%–21%). A total of 26% of patients with ER-negative tumors were predicted to be treatment sensitive. In these patients, 3-year DRFS was 83% (95% CI, 68%–100%), with ARR of 26% (95% CI, 4%–48%) and PPV for pathologic response of 83% (95% CI, 36%–100%). In contrast, patients who were ER-negative and predicted to be treatment insensitive had a PPV for 3-year relapse of 43% (95% CI, 28%–55%). In patients predicted to be treatment sensitive, the test had a significant diagnostic likelihood ratio for predicted occurrence vs absence of 3-year distant relapse or death of 0.27 (95% CI, 0.01-0.94) in patients who were ER-positive and 0.35 (95% CI, 0.04–0.91) in those who were ER-negative. The type of taxane therapy was not associated with predicted treatment sensitivity and DRFS.
After adjusting for standard clinicopathologic parameters, genomic predictions were independently and significantly associated with risk of distant relapse or death (sensitive vs insensitive: hazard ratio [HR], 0.19; 95% CI, 0.07–0.55; P = .002). The predictive utility of a multivariate Cox model of clinicopathologic factors (tumor stage and ER status) was increased with the addition of the genomic prediction (likelihood ratio, 13.8; P < .001). Other previously published genomic predictors were also significantly predictive of pathologic response in the discovery and validation cohorts, but, paradoxically, predicted worse DRFS in patients predicted to be treatment sensitive.
There was a clinically meaningful survival difference between patients predicted to be treatment sensitive and those predicted to be treatment insensitive. The predictive test in this study improved on the predictions based on clinicopathologic parameters.