New Score Helps Predict Late Distant Metastases in ER+ HER2- Breast Cancer

[Hopeful]

IMNG Medical Media. 2013 Jan 18, S London

SAN ANTONIO (IMNG) - A new multigene score called EndoPredict improves on clinical measures for predicting whether estrogen receptor-positive breast cancer will metastasize, especially in the long term. The score therefore may help identify patients who can skip extended antihormonal therapy.

Investigators with the Austrian Breast and Colorectal Cancer Study Group (ABCSG) studied more than 1,700 postmenopausal patients with early estrogen receptor (ER)-positive, HER2-negative breast cancer who underwent surgery and received hormonal therapy for 5 years.

Using formalin-fixed, paraffin-embedded tumor tissue, the team derived a pretreatment EndoPredict score (Sividon Diagnostics) for each patient. The score reflects expression levels of 12 genes involved in proliferation, ER signaling, and differentiation.

Patients with a low vs. high EndoPredict score were 20%-30% more likely to be free of distant metastases during the first 5 years of follow-up and also thereafter, lead investigator Dr. Peter Dubsky reported at the San Antonio Breast Cancer Symposium.

And an EndoPredict clinical score, which combined the score with clinical features, showed even better prediction, especially in the long term: Patients having a low vs. high EndoPredict clinical score at 5 years of follow-up were five times as likely to remain free of distant metastases thereafter. In absolute terms, more than 98% of this low-risk group was still metastasis free at 10 years.

"The EndoPredict score identifies early and late recurrences, and offers independent prognostic information beyond what can be achieved with all common clinical parameters," Dr. Dubsky maintained.

"Why would this data be important?" he asked. "We currently have around 20,000 women included in ongoing extended/late endocrine therapy clinical trials. Speaking from our trial, ABCSG-16, also known as SALSA [Secondary Adjuvant Long-Term Study With Arimidex], we see very low event rates, and the efficacy data of these trials is unlikely to make individual decisions for women very simple," he explained. "We believe that gene expression data may help establish patient subgroups with a very excellent prognosis and thus facilitate the therapeutic choice."

Dr. Laura Esserman of the University of California, San Francisco, asked how EndoPredict might stack up against other similar tools, such as the sensitivity to endocrine therapy (SET) index.

"I'd love to run this side by side in the same biomarker sample. That would be great," Dr. Dubsky commented. "We have had a fantastic presentation on the BCI (Breast Cancer Index), but this is a score that runs only in node-negative. And I think what is similar between the EndoPredict score and the BCI is that the BCI also has the ER signaling incorporated in the algorithm of genes."

Another attendee likewise asked whether the investigators had compared EndoPredict with tools such as Adjuvant Online.

"I'd be very careful with prediction power and comparing, because this is always going to depend on the biomarkers that you are using," Dr. Dubsky replied. "We have a fantastic sensitivity with this test, but how would another test from another company perform in the same biomarker sample, I cannot say."

"Would you not enrich your gene signature for late recurrence by simply examining patients who recurred late, and compare them with those that did not?" a third attendee wondered.

"The women assigned to the low-risk group have a very, very low incidence of both early and late recurrence; the high-risk group has a high risk of both early and late recurrence. The signature does in no way select specifically for the late recurrence," Dr. Dubsky explained. "I think what is special about the signature is that it can predict the late recurrences better than others do because it is not relying solely on proliferation."

Giving some background to the study, he noted, "The first-generation multigene signatures have largely been trained to predict early recurrences and not late recurrences, and commonly fail to identify the late events."
EndoPredict may be the first multigene test for breast cancer that can be used in a decentralized setting. The test has recently received the European CE mark as an in vitro diagnostic test, and is being used in Germany, Austria, and Switzerland.

The 1,702 postmenopausal women studied were participants in a pair of randomized trials (ABCSG-6 and ABCSG-8). One-third had node-positive disease. None received adjuvant chemotherapy, but all received hormonal therapy (tamoxifen alone or some sequence of tamoxifen and an aromatase inhibitor) for 5 years. Their median age was 64 years.

"This is a very homogeneously treated cohort with low to intermediate risk," Dr. Dubsky summarized.

A total of 998 women were still at risk for distant metastases after 5 years, and their median duration of follow-up was 7.1 years.

Overall, 49% of all patients had a low-risk EndoPredict score. After multivariate adjustment, these patients were significantly more likely to be free of distant metastases in the first 5 years of follow-up (hazard ratio, 1.20; P less than .001) and thereafter (HR, 1.28; P = .001). Fully 96.3% of patients with a low-risk score were metastasis free between 5 and 10 years.
"The EndoPredict is clearly an independent prognostic parameter both in the years 0 to 5 and after 5 years," Dr. Dubsky maintained.

When the C-index for discrimination was calculated, EndoPredict significantly improved on prediction of distant metastases after 5 years of follow-up when combined with the Adjuvant Online score (P less than .001) and other factors.
However, the best prediction was achieved with the predefined EndoPredict clinical score, which combined the EndoPredict score, nodal status, and tumor size, and achieved a C-index of nearly 0.8.

Fully 64% of patients still at risk for distant metastases after 5 years of follow-up had a low EndoPredict clinical score. These patients were dramatically more likely to remain free of distant metastases thereafter (HR, 5.11; P less than .001). In absolute terms, 98.2% were free of distant metastasis at 10 years.

"Risks and side effects of extended therapy should be weighed against this outcome," Dr. Dubsky recommended.

In a final analysis teasing apart the role of various genes included in the EndoPredict score, the proliferation genes added independent negative prognostic information for early recurrence, whereas the genes associated with ER signaling added independent positive prognostic information for late recurrence.

Dr. Dubsky disclosed that he is an adviser to Sividon, Agendia, Genomic Health, and AstraZeneca; receives grant support from Sividon and Agendia; and is on the speakers bureau for Sividon.

Hopeful