Veliparib

I-SPY 2 is an ongoing, multicenter, open-label, adaptive phase 2 master protocol or ‘platform’ trial with multiple experimental arms that evaluate novel agents combined with standard neoadjuvant therapy in breast cancers at high risk of recurrence.^{6 (link)} Experimental treatments are compared against a common control arm of standard neoadjuvant therapy, with the primary endpoint being pCR, which is defined as no residual cancer in either breast or lymph nodes at time of surgery. Patients who dropout after starting therapy (with or without withdrawal of consent) or fail to have surgery for any reason are counted as non-pCRs.
Biomarker assessments (HER2, HR, MammaPrint) performed at baseline are used to classify patients into 2×2×2 = 8 prospectively defined subtypes for randomization purposes. In addition to standard IHC and FISH assays, the protocol included a microarray-based assay of HER2 expression (TargetPrintTM). This assay has previously shown high concordance with standard IHC and FISH assays of HER2^{8 (link)}. The adaptive randomization algorithm assigns patients with biomarker subtypes to competing drugs/arms based on current Bayesian probabilities of achieving pCR within that subtype vs control with 20% of patients assigned to control. Adaptive randomization speeds the identification of treatments that perform better within specific patient subtypes and helps avoid exposing patients to therapies that are unlikely to benefit them (Figure 1A).^{9 ,10 (link)}To assess efficacy, ten clinically relevant biomarker ‘signatures’ were defined in the protocol: All; HR+; HR−; HER2+; HER2−; MP Hi-2; HER2+/HR+; HER2+/HR−; HER2−/HR+; HER2−/HR−. Experimental arms are continually evaluated against control for each of these signatures and “graduate” when and if they demonstrate statistical superiority in pCR rate. Statistical analyses are Bayesian.^{9 ,11 (link)} Graduation requires an 85% Bayesian predictive probability of success in a 300-patient equally randomized neoadjuvant phase 3 trial with a traditional statistical design comparing to the same control arm and primary endpoint, pCR, as in I-SPY 2. (see Supplementary Information). Predictive probabilities of success are power calculations for a 300-patient trial averaged with respect to the current probability distributions of pCR rates for the experimental arm and control.^{9 ,11 (link)} The modest size of this hypothetical future trial means that graduation occurs only when there is compelling evidence of an arm’s efficacy. Accrual to a graduating arm halts immediately, but all patients on the arm and its concurrent controls must complete surgery before graduation is announced. An experimental arm is dropped for futility if its predictive probability of success in a phase 3 trial <10% for all ten signatures. The maximum total number of patients assigned to any experimental arm is 120.