Date of Award
Open Access Thesis
Medical Doctor (MD)
Gene signature based prognostic tests can help improve adjuvant treatment decisions in early stage estrogen receptor positive (ER) breast cancers. Available tests in the clinic include Oncotype DX recurrence score (RS), PAM50 molecular class, and the Genomic Grade Index (GGI), which can identify high risk tumors that are likely to recur and have less favorable survival when treated with surgery and endocrine therapy alone. These high risk patients are recommended to also receive chemotherapy to improve their chance of survival. A subset of these "high risk" tumors is highly sensitive to adjuvant chemotherapy due to their high proliferation rates, and will be cured. We hypothesized that a new gene signature test ACES, which predicts treatment sensitivity to both endocrine therapy and chemotherapy and identifies tumors with excellent distant relapse free survival (RFS), could further stratify the currently "high risk" ER positive cancers into two groups: ACES predicted low and high residual risk after chemotherapy.
This is a retrospective cohort study, and samples size and power are limited by the number of available specimens. Three independent ER positive breast cancer cohorts - ACES Discovery Cohort (n=176), ACES Validation Cohort 1 (n=123), and a new Validation Cohort 2 (n=127) - were used to assess the ability of ACES to identify patients who were initially considered to be high risk for recurrence (by high RS, Luminal B subtype by PAM50, or high GGI) but became low risk after receiving adjuvant chemotherapy. The ACES algorithm was applied to the baseline high risk groups and cases were re-stratified into ACES predicted treatment sensitive and treatment insensitive groups. RFS and absolute risk reduction (ARR) of relapse were the main outcome measures compared between the ACES stratified groups.
In all three cohorts, cases that were high risk at baseline but predicted to be treatment sensitive by ACES showed a trend toward improved RFS. Cases with high risk by Oncotype DX high RS showed significant difference in RFS by ACES risk strata (p=0.048 and p=0.033) in validation cohort 1 and combined validation cohorts. Among these high RS tumors, n=11-13 (28-35%) were predicted to be treatment sensitive, which had RFS of 92-100% (95% CI: 54-100%) at 4-years. The ARR at 4-years was 0-41% (95% CI: -21-60%) and increased by 10-years to 19% (95% CI: 3-30%) favoring the treatment sensitive groups. Cases with high GGI in the discovery cohort also showed significant differences in RFS by ACES risk strata (p=0.004); the 45 (50%) high GGI cases who were predicted to be treatment sensitive had a RFS of 81% (95% CI: 60-92%) with ARR of 23% (95% CI: -2-51%). For these high RS and high GGI tumors, ACES remained an independent predictor of RFS in multivariate Cox regression analysis including age, T-stage, and lymph node involvement at diagnosis (p=0.072 and 0.017 respectively). Among Luminal B cancers, ACES was significantly associated with RFS only in the multivariate model of both validation cohorts (p=0 and 0.013).
This analysis provides evidence to suggest that ACES may further risk stratify high RS and high GGI tumors into low and high residual risk groups after adjuvant chemotherapy and endocrine therapy. The clinical relevance is that if ACES is adequately validated: (i) patients with low residual risk by ACES can be safely treated with current adjuvant chemotherapies and reassured, (ii) patients with high residual risk despite best current adjuvant chemotherapies could be encouraged to enter clinical trials that aim to improve the efficacy of current adjuvant therapies. Before ACES can be adopted for routine use it would require validation in an adequately powered prospective trial, and the results presented in this thesis suggest that future validation of the ACES algorithm as residual risk prediction tool should be pursued.
Khan, Sabrina, "Genomic Predictor Of Residual Risk Of Recurrence After Chemotherapy In High Risk Estrogen Receptor Positive Breast Cancers" (2014). Yale Medicine Thesis Digital Library. 1890.