SAN ANTONIO – The results of the EndoPredict test appear to predict tumor response in patients with early hormone receptor–positive, HER2-negative breast cancer given neoadjuvant therapy, based on results of a study conducted by the Austrian Breast & Colorectal Cancer Study Group (ABCSG).
“Very good tumor shrinkage in estrogen receptor–positive, HER2-negative disease is going to happen only in a minority of patients, and biomarkers that would predict excellent tumor shrinkage are an unmet medical need,” commented lead investigator Peter Dubsky, MD, PhD, who is head of the Breast Center at Hirslanden Klinik St. Anna, Lucerne, Switzerland. “As a surgeon, that would help me to predict breast conservation at diagnosis, but as a surgical oncologist, I would also recognize that tumor response is an important component of future survival.”
SABCS/Scott Morgan 2017
Dr. Peter Dubsky
The ABCSG findings suggest expanded utility for EndoPredict. The test’s molecular score is currently used along with tumor size and nodal status to predict the 10-year distant recurrence rate, and whether patients may safely forgo chemotherapy or are at high risk and may need adjuvant chemotherapy in addition to endocrine therapy.
Dr. Dubsky and his coinvestigators assessed performance of the EndoPredict test among 217 patients treated on ABCSG 34, a randomized phase 2 neoadjuvant trial. Findings showed that among patients given neoadjuvant endocrine therapy because they had less aggressive disease features, an EndoPredict high-risk result was associated with poor response (negative predictive value of 92%), defined as a residual cancer burden (RCB) of II or III, he reported at the San Antonio Breast Cancer Symposium.
On the other hand, among patients given neoadjuvant chemotherapy because they had more aggressive disease features, a low-risk result was associated with poor response (negative predictive value of 100%).
“Clinicians really gave us two distinct cohorts within ABCSG 34. In the luminal A–type patients who were treated with neoendocrine therapy, a high EndoPredict score predicted a low chance of tumor shrinkage. In the more aggressive ER-positive tumors, so-called luminal B type, treated with neoadjuvant chemotherapy, there was absolutely no excellent response in the low-risk group,” Dr. Dubsky summarized. “We believe that this molecular score may contribute to patient selection for biomarker-driven studies, especially in the neoadjuvant setting.”
Session attendee Steven Vogl, MD, a medical oncologist with the Montefiore Medical Center in New York, commented, “I have trouble correlating an RCB of 0 or I with what you as a surgeon do for the patient, because you are talking about pathologic complete response or just a few cells there. That’s not what determines how much breast you take off: It’s determined by the total size of the tumor and the size of the breast. So if it’s less than a few centimeters, I’m sure you can do a lumpectomy in every patient. Tell me why I should care that you are getting an RCB of 0 or I in these endocrine patients.”
“Because it’s more likely that these patients will have a smaller tumor and better tumor shrinkage,” Dr. Dubsky replied. “You are of course right, RCB 0 or I was not designed to help surgeons. But it helps me as a translational scientist to have a surrogate and an exact classification for good tumor shrinkage. That’s how I used it.”
C. Kent Osborne, MD, codirector of SABCS and director of the Dan L. Duncan Cancer Center at Baylor College of Medicine in Houston, asked, “We see it in the clinic, and I’m sure you have as well, patients whose tumor doesn’t shrink very much, but the Ki-67 really drops. And that may or may not be a better factor than the actual tumor shrinkage. So how many patients who had tumors that didn’t shrink, which was your endpoint, had a reduction in Ki-67 that was, say, 5%?”
“We haven’t looked at that specifically, but we will do so as we carry on with the follow-up of these patients. Then we can learn more about the prognosis,” Dr. Dubsky replied.
Study details
ABCSG 34 was a randomized phase 2 trial testing addition of the cancer vaccine tecemotide (Stimuvax) to neoadjuvant standard of care among patients with HER2-negative early breast cancer.
Dr. Dubsky and coinvestigators restricted analyses to patients with hormone receptor–positive disease who, depending on clinical and pathologic factors, received neoadjuvant chemotherapy (eight cycles of epirubicin-cyclophosphamide and docetaxel) or neoadjuvant endocrine therapy (6 months of letrozole [Femara]) as standard of care. They were then randomized to additionally receive tecemotide or not before undergoing surgery.
Overall, 25% of the 134 patients in the neoadjuvant chemotherapy group had a good tumor response, defined as pathologic complete response in both breast and nodes (RCB of 0) or minimal residual disease (RCB of I).
Higher EndoPredict score was associated with greater likelihood of good response to chemotherapy. EndoPredict risk group (high vs. low) had a negative predictive value of 100%, a positive predictive value of 26.4%, a true-positive rate of 100%, and a true-negative rate of 8.9% for predicting response (P = .112).
Area under the receiver operating characteristic curve was 0.736.
In a multivariate model, EndoPredict score as a continuous variable was not an independent predictor of response. “The good response was largely driven by covariates that included cell proliferation, and it was Ki-67 that was significant,” Dr. Dubsky noted.
Overall, 18% of the 83 patients in the neoadjuvant endocrine therapy group had a good tumor response (RCB of 0 or I). Here, lower EndoPredict score was associated with greater likelihood of good response. EndoPredict risk group (high vs. low) had a negative predictive value of 92.3%, a positive predictive value of 27.3%, a true-positive rate of 80.0%, and a true-negative rate of 52.9% for predicting response (P = .024). Area under the curve was 0.726.
In a multivariate model here, EndoPredict score as a continuous variable, its estrogen receptor–signaling/differentiation component, and Ki-67 did not independently predict response. “It was maybe a bit surprising that T stage was the strongest factor, possibly indicating that we should have simply treated those women longer than 6 months,” Dr. Dubsky commented. The EndoPredict proliferation component was also a significant predictor.
“Possibly, the very narrow distribution of Ki-67 [among patients given neoendocrine therapy] may have prevented this factor from playing a bigger role in this particular model,” he speculated.
Dr. Dubsky disclosed that he receives consulting fees from Myriad, the maker of EndoPredict, and from Cepheid, Nanostring, and Amgen.