NEW YORK – Within I-SPY 2.2, an adaptive, biomarker-stratified basket trial, researchers have identified promising neoadjuvant treatment strategies for breast cancer patients based on complex biomarkers that they say warrant further study.
During the European Society for Medical Oncology conference on Saturday, researchers presented results from two arms of the I-SPY 2.2 trial involving high-risk, early-stage breast cancer patients. In arm A, patients received neoadjuvant treatments involving AstraZeneca and Daiichi Sankyo's TROP2-directed antibody-drug conjugate datopotamab deruxtecan (Dato-DXd), and in arm B, patients received treatments involving Dato-DXd plus AstraZeneca's PD-1 inhibitor Imfinzi (durvalumab). The results from these two arms were published simultaneously in Nature Medicine on Saturday (see here and here).
The I-SPY 2.2 trial follows a design that investigators have dubbed SMART, short for Sequential Multiple Assignment Randomization Trial. The complex design involves identifying biomarker-based neoadjuvant treatment strategies for early-stage breast cancer patients and adjusting based on their responses. Patients must be HER2-negative, have stage II or III breast cancer, and have tumors at high risk of recurrence as determined by Agendia's 70-gene expression MammaPrint assay. The overarching goal of the Quantum Leap Healthcare Collaborative-sponsored multicenter trial is to identify treatment strategies to which patients will have a pathological complete response by the time they have their tumor surgically removed — and to get to that point with the least possible toxicities and lines of treatment.
"This trial is very personalized and very patient-centered," said Laura Esserman, a breast cancer surgeon at the University of California, San Francisco, and one of the principal investigators on the I-SPY 2.2 trial. "It's a continuous learning system, and we're learning from both the people who didn't respond and the people who did."
The complex Phase II basket trial comprises three blocks. As patients advance through the blocks, they receive sequential treatments guided by their specific disease subtypes, which the I-SPY 2.2 investigators developed, that influence treatment response. Depending on how patients respond to the intervention they receive in each block, they can either progress to surgical resection if they experience a pathologic complete response or receive another line of therapy if they don't.
In block A, patients are randomized to receive an experimental therapy based on their cancer's specific subtype. Patients then move to block B, where they receive surgery or a combination of taxane-based chemotherapy with or without Merck's checkpoint inhibitor Keytruda (pembrolizumab). Progressing to block C, patients can receive surgery if they haven't previously, or anthracycline-based chemotherapy with or without Keytruda.
I-SPY 2.2 investigators developed the response-predictive subtypes for guiding treatment within the adaptive trial based on data from around 990 patients in 10 arms of the trial's earlier iteration, I-SPY 2. These response-predictive subtypes incorporate more information than can be gleaned from a single immunohistochemistry or traditional molecular subtyping test, Fatima Cardoso, the director of the breast unit of the Champalimaud Clinical Center in Lisbon, Portugal, pointed out in a discussion of the data at the meeting.
The response-predictive subtypes are determined by patients' hormone receptor status; HER2 status; their likelihood of responding to DNA repair deficiency-targeted agents; their HER2 luminal or basal subtypes; and their immune responsive signatures, which capture their likelihood of responding to immunotherapy based on Agendia's 53-gene expression ImPrint test.
These response-predictive subtypes "allow us to understand which type of tumors respond better to each therapeutic manipulation," Cardosa said.
A sequential regimen with Dato-DXd
Katia Khoury, a breast cancer oncologist at UAB Medicine in Birmingham, Alabama, presented results from patients who'd received single agent Dato-DXd as the experimental treatment in arm A.
After patients went through the sequential courses of treatments in blocks A, B, and C, the overall pathologic complete response rate was 38.1 percent among 103 patients randomized to receive Dato-DXd in block A. Nearly half of these pathologic complete responses occurred after block A, which allowed some patients to skip more traditional chemo, Khoury noted.
Treatment with Dato-DXd alone didn't lead to a high enough pathologic complete response rate compared to the control arm, however, for it to be considered a "successful" therapy. That said, sequential treatment with Dato-DXd then the chemotherapy regimens in arms B and C did outperform the control arm in patients with a particularly tough-to-treat cancer subtype: hormone receptor-negative, HER2-negative, immune response-negative, and DNA damage repair-negative tumors.
"The Dato-DXd monotherapy arm was the first arm on the novel I-SPY 2.2 trial design," Khoury noted, pointing out that "accrual was brisk," indicating patients' interested in this adaptive trial design.
These results signaled to Khoury and Cardosa that the activity of a sequential therapy comprising Dato-DXd followed by taxane-based chemotherapy and subsequent anthracycline-based chemo should be tested in Phase III clinical trials.
Dato-DXd, Imfinzi
Meghna Trivedi, a breast oncologist at Columbia University Irving Medical Center, presented results from arm B, in which patients received Dato-DXd plus Imfinzi as the experimental therapy.
Among 106 patients who'd initially been randomized to receive Dato-DXd and Imfinzi in block A, the pathologic complete response rate was 50 percent after they completed the full sequence of treatments in blocks B and C. By the time patients completed block A, combined treatment with Dato-DXd and Imfinzi had already met investigators' threshold for success in patients with the immune-positive subtype. At this early stage in the trial, the pathologic complete response rate was 65 percent, well above the prespecified 40 percent threshold.
"The Dato-DXd plus durvalumab treatment strategy was extremely active, with the highest observed pathologic complete response rate in the immune-positive subtype," Trivedi said, noting that the results warrant further investigation of these treatment strategies in this specific early-stage breast cancer subtype.
After patients with immune-positive breast cancer on Dato-DXd and Imfinzi completed all three blocks of treatment, the pathologic complete response rate was 79 percent, which while high, was still not enough to meet the prespecified response rate, since the pathologic complete response was 78 percent in the control arm.
For patients negative for all biomarkers, the most difficult-to-treat subtype, the combination of Dato-DXd, Imfinzi, and the two sequential chemotherapies did outperform the control arm. The pathologic complete response rate was 46 percent after all three blocks of treatment for patients who'd started on Dato-DXd plus Imfinzi in block A and 16 percent for the control arm. Accordingly, the antibody-drug conjugate-immune checkpoint inhibitor combination graduated in this specific subtype.
"This deserves to be further evaluated," Cardosa said.
Toxicity, surrogate endpoints, lingering questions
Cardosa noted during her discussion that toxicities on the I-SPY 2.2 trial were, perhaps as expected, highest for patients who went through all three blocks including courses of both taxane- and anthracycline-based chemo. One patient died due to toxicities, which certainly raised red flags for oncologists.
"We are always very careful when we see toxic deaths in such a small number of patients treated," Cardosa said. "When we decide to move forward with a certain treatment strategy, we have to [determine] a balance between efficacy and toxicity."
The data from these two arms involving single-agent Dato-DXd and Dato-DXd plus Imfinzi also raised several questions for Cardosa. For one, she said it's still unclear how best to identify patients who respond well enough on the experimental treatments that they don't need the subsequent chemo and immunotherapies.
"There is research ongoing to try to characterize these tumors better," she said. "These would be the ones that we could move forward, avoiding the add-on strategy."
She further questioned whether oncologists can be confident that after a pathologic complete response to experimental therapy in arm A, avoiding subsequent therapies is the right choice. This question touches on a fundamental challenge when interpreting trial data based on surrogate endpoints like pathologic complete response. The entire design of the I-SPY 2.2 trial is based on the concept of moving patients through treatment sequences based not only on their biomarker status but also on whether they experienced a pathologic complete response to treatments they received.
In the past, early-stage breast cancer patients who had a pathologic complete response to neoadjuvant treatment haven't always had positive event-free survival and overall survival outcomes. "We need to remember that … the magnitude of difference that we see in pathologic complete response doesn't always, and doesn't often, translate in long-term benefits," Cardosa said. "That's why we need to evaluate these initial results in a Phase III trial."
Finally, given the trial's complex design, she noted that it's difficult to parse the contributions of the various treatment blocks to patients' outcomes. For instance, if a patient experiences a pathologic complete response after experimental treatment in block A, followed by treatment in blocks B and C, it's unclear whether she might have experienced that same outcome from the treatments in blocks B and C alone.
Looking ahead, Cardosa said that the results of I-SPY 2.2 could inform Phase III trial designs for the individual drugs that meet the graduation threshold.
According to Esserman, I-SPY 2.2 investigators are in talks with the US Food and Drug Administration to discern how the design for the next iteration of I-SPY might meet regulators' criteria for approving experimental agents should the outcomes in specific response-predictive subtypes warrant it. The progression isn't straightforward, since the trial has, so far, been designed as a Phase II safety signal-finding trial rather than a trial to pit individual personalized agents head-to-head with control arms.
"We're working on building that [Phase III] design with regulators," she said. "With I-SPY 2.3, what we hope to be able to do is figure out how to morph this into a seamless, Phase II/III design where if we see a signal, we can just keep going to confirm that signal." Investigators are also working on incorporating blood-based circulating tumor DNA testing into the adaptive trial in the hopes of more accurately characterizing patients' treatment responses based on their minimal residual disease.
Esserman said she hopes this shift into a Phase III iteration could happen within the next year or so, though she acknowledged the logistics are less straightforward than most in the field, including regulators, are used to.
"We're going to figure out how to do it, and it's OK if it's a little bit complex," she said. "We'll figure it out, and we'll do it together. It's super exciting to be in a position to be able to innovate trial designs and to focus on the data and endpoints that matter most to patients."