I-SPY 2: Separating Contenders From Pretenders in Breast Cancer

Get Permission

While identifying appropriate agents for testing in larger randomized trials, the pressure to generate data that can lead to drug approval may make it more difficult to explore alternative regimens in the hope of improving efficacy and/or reducing toxicity and could result in rejection of a useful agent.
— William M. Sikov, MD, FACP

Tweet this quote

With the expansion of our understanding of signaling pathways in normal cells and how they are co-opted or corrupted in malignancy, the number of potential antitumor agents to be tested has exploded, exposing the limitations of traditional antineoplastic drug development and challenging us to develop novel strategies to distinguish those with real promise from those with insufficient activity or excessive toxicity. Given its prevalence and heterogeneity, the treatment of localized breast cancer is an appealing target for these efforts.

I-SPY 2 is an elegantly designed trial that uses the intermediate endpoint of pathologic complete response to assess the addition of a series of novel agents to a standard neoadjuvant chemotherapy regimen. To enhance accrual to promising treatments, and minimize exposure to less effective ones, the study utilizes adaptive randomization, in which new patients are more likely to be assigned to a treatment arm if past pathologic complete response results for their subset—defined by hormone receptor and HER2 subtypes and results of a 70-gene expression assay—are encouraging. If outcomes continue to favor an investigational arm over the control, the agent will “graduate” when Bayesian analysis suggests a > 85% probability that a 300-patient randomized phase III trial with the same design would demonstrate its statistical superiority (again using pathologic complete response as the primary endpoint) in one or more patient subsets. Accrual to an investigational arm is also closed (the term “flunk” would appear to be appropriate for this) if statistical analysis suggests that an agent would not cross the 85% probability threshold for success in any patient subgroup. Imaging and correlative studies embedded in I-SPY 2 are also expected to yield valuable findings.

Veliparib, Neratinib, and Others

The two articles that recently appeared in The New England Journal of Medicine—summarized in this issue of The ASCO Post—represent the first published results from I-SPY 2.1,2 In the first, the addition of the combination of the chemotherapeutic agent carboplatin and the PARP (poly [ADP-ribose] polymerase) inhibitor veliparib was studied in patients with HER2-negative cancers. The combination required only 60 patients, randomized between the investigational and control arms, to “graduate” in the triple-negative patient subset, due to a large differential in observed pathologic complete response rates.

The investigational treatment was associated with much higher rates of hematologic toxicities and early treatment discontinuation than the control arm. The authors noted there is no way to determine how much of the efficacy, or toxicity, of the investigational regimen is due to carboplatin vs the veliparib/carboplatin combination. Thus, the successor phase III study ( NCT02032277), which has completed accrual (n = 624) and from which results for the pathologic complete response endpoint are expected within a year, includes an arm on which patients receive standard neoadjuvant chemotherapy plus carboplatin, with a veliparib, placebo, as well as a control arm and one on which patients receive both carboplatin and veliparib.

The second article reported results for the oral pan-HER family irreversible tyrosine kinase inhibitor neratinib. Although adding this agent to weekly paclitaxel had no impact on pathologic complete response rates in HER2-negative patients, when administered with weekly paclitaxel in place of trastuzumab (Herceptin) in HER2-positive patients (n = 87), the estimated pathologic complete response rate rose, and the difference between the investigational and control regimens in the hormone receptor–negative/HER2-positive cohort met the study’s “graduation” criteria. Although the increment in the pathologic complete response rate was less impressive in the hormone receptor–positive/HER2-positive cohort, the investigators favored including these patients in any subsequent phase III trial, stratifying for this variable.

Treatment with paclitaxel plus neratinib was associated with a 38% incidence of grade ≥ 3 diarrhea (compared with 4% with trastuzumab), although this diminished with prophylactic loperamide. Complicating plans to conduct a simple phase III study replicating the I-SPY 2 randomization is the widespread use of dual HER2-targeted therapy in the neoadjuvant setting following accelerated approval of pertuzumab (Perjeta) for this indication.

In addition to these published manuscripts, data from I-SPY 2 presented in abstract form have demonstrated that the addition of the oral Akt inhibitor MK-2206 increased the estimated pathologic complete response rates in multiple patient subsets, with the highest likelihood of demonstrating superiority over the control regimen in the hormone receptor–negative/HER2-positive cohort. The data also show that although the addition of the angiopoietin-1/2–neutralizing peptibody trebananib increased the estimated pathologic complete response rates in a number of patient cohorts, in no group was the increment sufficient to yield a ≥ 85% probability of success in a subsequent phase III study.3,4 The trial continues to study a number of agents, and others are likely to be added going forward.

Concerns About the I-SPY 2 Design

Given these intriguing findings, what concerns might be raised about the I-SPY 2 design? While identifying appropriate agents for testing in larger randomized trials, the pressure to generate data that can lead to drug approval may make it more difficult to explore alternative regimens in the hope of improving efficacy and/or reducing toxicity and could result in rejection of a useful agent. For example, might a lower weekly dose of carboplatin be more effective and better tolerated given in combination with veliparib than the higher every-3-week dose currently employed? Is 50 mg a day the optimal dose and schedule for veliparib?

In addition, despite the statistical rigor with which the I-SPY 2 results were analyzed, they still involve relatively small numbers of patients. Data from another randomized phase II trial suggest that combining neratinib with trastuzumab is more effective at inducing pathologic complete response in hormone receptor–negative/HER2-positive cancers than either drug alone, suggesting that, despite the I-SPY 2 prediction of success, a phase III study utilizing single-agent neratinib could fail to confirm its value.5

Another concern is the use of the dichotomous variable pathologic complete response as the primary means to define the success of treatment. Data suggest that the prognosis for patients who achieve minimal residual disease, such as those with residual cancer burden class I, is equivalent to that with pathologic complete response and that even in patients with more extensive residual disease, there is a relationship between its extent and prognosis.6 Thus, an investigational treatment that shifts more patients toward less residual disease may have a greater impact on long-term outcomes than apparent based on pathologic complete response results alone.

The ‘Neoadjuvant Paradigm’

However, the biggest challenge to the I-SPY 2 model comes from questions being raised as to the applicability of what has been called the “neoadjuvant paradigm.” More effective neoadjuvant therapy allows a higher percentage of patients to be offered breast-conserving surgery, and the possibility of avoiding a full axillary dissection in patients with clinical N1 disease who respond to treatment, but the ultimate endpoints for assessing efficacy remain its impact on recurrence-free and overall survival. Although achievement of a pathologic complete response is clearly a useful prognostic factor for individual patients, some have questioned whether trial-level data demonstrating a significant increase in the pathologic complete response rate can be relied upon as a surrogate for improvements in long-term outcomes.

Many factors contribute to the complexity of translating higher pathologic complete response rates into improved recurrence-free and overall survival, starting with the realization that although the prognosis for patients who achieve a pathologic complete response is generally very good, many patients who do not achieve a pathologic complete response will also be cured, especially those with lower-risk disease by stage and subtype. Adjuvant treatments, ranging from standard endocrine and HER2-targeted therapies to additional chemotherapy to investigational agents, can only benefit patients with occult residual disease, potentially reducing outcome differences between patients who achieve a pathologic complete response and those who do not.

Assume that the I-SPY 2 analysis is correct, and a subsequent 300-patient randomized neoadjuvant trial demonstrates a 20% to 30% absolute increase in the pathologic complete response rate with addition of the investigational agent. This could be expected to translate into a 5% to 7% improvement in recurrence-free and/or overall survival, which many would consider clinically relevant in a high-risk patient subset. However, to confirm this with statistical rigor would likely require a study with thousands of patients. Are we certain that doctors would support, and patients would agree to participate in, a study in which they could be randomized not to receive the promising agent, especially if it has received accelerated approval? For example, although there is as yet no proof that the addition of pertuzumab improves long-term outcomes in HER2-positive cancers, its listing as an option in the adjuvant setting in the National Comprehensive Cancer Network® (NCCN®) guidelines illustrates how difficult it can be to control the use of an agent perceived to be of value once it is available.

The optimal situation may be to identify a patient subset with a poor prognosis with standard treatment and hope that the pathologic complete response increment achieved with the investigational therapy in a randomized neoadjuvant trial is large enough to translate into significant recurrence-free and/or overall survival benefits, and thus avoid having to do the successor study.

Although these and other issues challenge the advancement of novel agents in the neoadjuvant setting, the first step is to identify those agents most likely to make a major difference, to separate contenders from pretenders, which is the primary goal of the I-SPY 2 investigators. They are to be congratulated for the work accomplished thus far, especially their success in convincing a number of pharmaceutical and biotech companies to make their novel agents available to be studied in this trial, and encouraged to continue their efforts. ■

Disclosure: Dr. Sikov is on the steering committee for the veliparib study which is sponsored by AbbVie Pharmaceuticals. He is compensated for his time spent at its meetings; however, he does not receive any other funding or travel reimbursement.


1. Rugo HS, Olopade OI, DeMichele A, et al: Adjuvant randomization of veliparib-carboplatin treatment in breast cancer. N Engl J Med 375:23-34, 2016.

2. Park JW, Liu MC, Yee D, et al: Adaptive randomization of neratinib in early breast cancer. N Engl J Med 375:11-22, 2016.

3. Tripathy D, Chien AJ, Hylton N, et al: Adaptively randomized trial of neoadjuvant chemotherapy with or without the Akt inhibitor MK-2206: Graduation results from the I-SPY 2 trial. 2015 ASCO Annual Meeting. Abstract 524.

4. Albain KS, Leyland-Jones B, Symmans F, et al: The evaluation of trebananib plus standard neoadjuvant therapy in high-risk breast cancer: Results from the I-SPY 2 trial. 2015 San Antonio Breast Cancer Symposium. Abstract P1-14-03.

5. Jacobs SA, Robidoux A, Garcia JMP, et al: NSABP FB-7: A phase II randomized trial evaluating neoadjuvant therapy with weekly paclitaxel plus neratinib or trastuzumab or neratinib plus trastuzumab followed by doxorubicin and cyclophosphamide with postoperative trastuzumab in women with locally advanced HER2-positive breast cancer. 2015 San Antonio Breast Cancer Symposium. Abstract PD5-04

6. Symmans WF, Peintinger F, Hatzis C, et al: Measurement of residual breast cancer burden to predict survival after neoadjuvant chemotherapy. J Clin Oncol 25:4414-4422, 2007.

Related Articles

Neoadjuvant Veliparib/Carboplatin and Neratinib ‘Graduate’ From Adaptive Randomization I-SPY 2 Trial in Early Breast Cancer

Veliparib-carboplatin added to standard therapy resulted in higher rates of pathological complete response than standard therapy alone, specifically in triple-negative breast cancer.
— Hope S. Rugo, MD, and colleagues

Tweet this quote

As reported by Hope S. Rugo, MD,...