Elevated Serum HER2/neu Linked to Lower Response to Hormone Therapy

March 1, 2003

HERSHEY, Pennsylvania-Elevated levels of common serum biomarkers, including HER2/neu, predict response to hormone therapy among women with advanced breast cancer, according to a study reported by Kim Leitzel, MS, senior research associate in the Experimental Oncology Research Lab, directed by Allan Lipton, MD, at Penn State/Hershey Medical Center, Hershey, Pennsylvania.

HERSHEY, Pennsylvania—Elevated levels of common serum biomarkers, including HER2/neu, predict response to hormone therapy among women with advanced breast cancer, according to a study reported by Kim Leitzel, MS, senior research associate in the Experimental Oncology Research Lab, directed by Allan Lipton, MD, at Penn State/Hershey Medical Center, Hershey, Pennsylvania.

The study was a retrospective analysis of pretreatment serum samples from women treated with either letrozole (Femara) or tamoxifen in the pivotal trial that found letrozole to be superior to tamoxifen as a first-line therapy for metastatic breast cancer (J Clin Oncol 19: 2596-2606, 2001). The researchers measured serum levels of CEA, CA 15-3, and HER2/neu and analyzed the effects of elevated levels on clinical benefit rate and time to progression.

Univariate Analysis

The intent-to-treat patient population was 907 postmenopausal women with metastatic breast cancer who were hormone-receptor positive or whose receptor status was unknown. Serum was available for 548 of the women and was analyzed with the Bayer Immuno-1 automated assay. Biomarker elevation was defined according to commonly used cutoff points: 15 ng/mL for HER2/neu, 5 ng/mL for CEA, and 38.5 U/mL for CA 15-3. Using these cutoff points, the percentages of patients with elevated biomarkers were: 28% HER2/neu, 44% CEA, and 56% CA 15-3. Among patients with elevated biomarkers, 273 patients were ultimately treated with letrozole, and 275 were treated with tamoxifen.

In univariate analysis, elevated serum CEA, CA 15-3, HER2/neu, and number of metastatic lesions were highly significant predictors of clinical benefit rate and time to progression. Among traditional clinical predictors, presence of liver metastases, Karnofsky performance status, and disease-free interval were all significant, Mr. Leitzel reported, as was treatment effect, with letrozole being superior to tamoxifen. Factors that did not prove predictive in univariate analysis were prior adjuvant hormonal therapy, hormone-receptor status unknown vs positive, and patient age.

When patient response was analyzed according to serum HER2/neu status, patients with elevated HER2/neu levels had greatly reduced objective responses, clinical benefit rates (complete or partial response plus stable disease for more than 24 weeks), and median time to progression.

Superior on All Counts

The objective response rate was 32% for women with normal HER2/neu levels compared to 15% for those with elevated HER2/neu. Similarly, the clinical benefit rate was 51% among women with normal serum HER2/neu vs 30% for those with HER2/neu elevations. Median time to progression was 9.5 months in the group with normal levels compared to 5.7 months for those with elevated HER2/neu.

"Patients that had longer intervals— with a time to progression of greater than 2 years—almost invariably had normal levels of serum HER2/neu," Mr. Leitzel said. Patients that had normal levels of serum HER2/neu and yet still had very short times to progression had other factors that predicted a shorter time to progression, such as elevated CEA levels.

When the researchers analyzed response to letrozole vs tamoxifen among patients with normal HER2/neu levels, letrozole proved superior in all measurements: objective response rate, 39% vs 27%; clinical benefit rate, 56% vs 45%; and median time to progression, 12.2 months vs 8.5 months.

CEA Levels Also Predictor

CEA levels also predicted treatment response. The objective response rate among women with normal CEA levels was 32% compared to 21% for those with elevated levels. Similarly, the clinical benefit rate was 51% for those with normal levels vs 37% for those with elevated levels, and median time to progression was 9.5 months vs 6 months.

In an analysis of outcome according to treatment, patients with elevated CEA levels did significantly better on letrozole than on tamoxifen. In this group, the objective response rate was 27% on letrozole vs 16% on tamoxifen; clinical benefit rate, 45% vs 29%; and time to progression, 8.2 months vs 3.5 months.

When the investigators analyzed time to progression according to treatment and the status of both serum HER2/neu and CEA, they found that women with normal levels of both biomarkers fared best. Time to progression was 14.4 months on letrozole and 9.3 months on tamoxifen for women with normal HER2/neu and CEA levels, compared to 6 months on letrozole and 3.3 months on tamoxifen for women with elevated levels of both. Among patients with normal HER2/neu and elevated CEA, the median time to progression was 9.7 months on letrozole vs 5.7 months on tamoxifen. For those with elevated HER2/neu and normal CEA, the median time to progression was 7.4 months on letrozole vs 3.1 months on tamoxifen.

Multivariate Analysis

On multivariate analysis of factors that are predictive of clinical benefit and time to progression, serum HER2/neu was an independent predictor of response to hormone therapy. "Patients with elevated serum HER2/neu had a 38% increased risk of progression compared to those with normal levels," Mr. Leitzel reported. "The number of metastatic lesions also was a significant independent risk factor. The two biomarkers of tumor burden, serum CEA and serum CA 15-3, were not significant in multivariate analysis."

In terms of traditional clinical factors, treatment effect was significantly in favor of letrozole, he added, with a 32% decrease in risk of progression compared to tamoxifen treatment.

"Elevated serum HER2/neu levels remain an independent predictor of relative resistance to hormone therapy, even when the clinical indicators of tumor burden are included in the multivariate model," Mr. Leitzel concluded.