Adaptive Randomization of Veliparib-Carboplatin Treatment in Breast Cancer

Abstract

Background: The genetic and clinical heterogeneity of breast cancer makes the identification of effective therapies challenging. We designed I-SPY 2, a phase 2, multicenter, adaptively randomized trial to screen multiple experimental regimens in combination with standard neoadjuvant chemotherapy for breast cancer. The goal is to match experimental regimens with responding cancer subtypes. We report results for veliparib, a poly(ADP-ribose) polymerase (PARP) inhibitor, combined with carboplatin.

Methods: In this ongoing trial, women are eligible for participation if they have stage II or III breast cancer with a tumor 2.5 cm or larger in diameter; cancers are categorized into eight biomarker subtypes on the basis of status with regard to human epidermal growth factor receptor 2 (HER2), hormone receptors, and a 70-gene assay. Patients undergo adaptive randomization within each biomarker subtype to receive regimens that have better performance than the standard therapy. Regimens are evaluated within 10 biomarker signatures (i.e., prospectively defined combinations of biomarker subtypes). Veliparib-carboplatin plus standard therapy was considered for HER2-negative tumors and was therefore evaluated in 3 signatures. The primary end point is pathological complete response. Tumor volume changes measured by magnetic resonance imaging during treatment are used to predict whether a patient will have a pathological complete response. Regimens move on from phase 2 if and when they have a high Bayesian predictive probability of success in a subsequent phase 3 neoadjuvant trial within the biomarker signature in which they performed well.

Results: With regard to triple-negative breast cancer, veliparib-carboplatin had an 88% predicted probability of success in a phase 3 trial. A total of 72 patients were randomly assigned to receive veliparib-carboplatin, and 44 patients were concurrently assigned to receive control therapy; at the completion of chemotherapy, the estimated rates of pathological complete response in the triple-negative population were 51% (95% Bayesian probability interval [PI], 36 to 66%) in the veliparib-carboplatin group versus 26% (95% PI, 9 to 43%) in the control group. The toxicity of veliparib-carboplatin was greater than that of the control.

Conclusions: The process used in our trial showed that veliparib-carboplatin added to standard therapy resulted in higher rates of pathological complete response than standard therapy alone specifically in triple-negative breast cancer. (Funded by the QuantumLeap Healthcare Collaborative and others; I-SPY 2 TRIAL ClinicalTrials.gov number, NCT01042379.).

Figure 1A. I-SPY 2 Adaptive Design

Figure 1A illustrates the steps within the I-SPY 2 adaptive process. When new patients are enrolled, their subtypes are assessed. As patients are randomized, their outcomes are used to update the Bayesian co-variate adjusted model which computes the predictive probability of success in phase 3 for each signature. Pre-defined termination rules are applied for each experimental arm to determine whether it should stop for futility, graduate, or continue, adding on additional experimental arms if accrual permitting. As the trial continues, for each experimental arm, the probability of superiority over control within each subtype is updated; and the randomization probabilities for each subtype into the various experimental arms are adapted (such that new patients entering the trial will be more likely to be randomized to an agent showing activity within their subtype).

Figure 1B. I-SPY 2 Study Design

Patients are screened for I-SPY 2 eligibility. Eligible patients are adaptively randomized to 12 weekly paclitaxel (and trastuzumab if HER2+) cycles (control) or in combination with one of several experimental agents followed by doxorubicin/cyclophosphamide (AC) × 4, with serial biomarkers (biopsies, blood draw and MRI scans) assessed over the course of their therapy. Only patients with HER2− disease were randomized to the veliparib/carboplatin arm.

Figure 1C. I-SPY 2 Consort Diagram for veliparib/carboplatin arm and its concurrent control

Only patients with HER2-negative disease were eligible for randomization to the VC arm. Patients were categorized as received allocated invention if they received at least one dose of experimental (or control) therapy.

Figure 2

Estimated pCR Rate for the signatures evaluated for V/C vs. concurrent HER2-negative control. 2A. Probability distribution for all patients with HER2-negative disease; 2B. Probability distribution for patients with TNBC (HR-negative/HER2-negative); 2C. Probability distribution for patients with HR-positive, HER2-negative disease. The red curves represent patients treated with V/C plus paclitaxel followed by AC, and the blue curves represent concurrent controls. The corresponding 95% probability distributions (represented by the width of the curve) are shown for each. The mean of each distribution is the estimated pCR rate.