P2-12-25: Blinded Multi-Site Validation of a Pathology-Based Prognostic Marker Profile for Operable Hormone Receptor-Positive Breast Cancer.

Abstract

Abstract Background We previously reported development and preliminary validation of a marker profile with a risk of recurrence algorithm for patients with operable hormone receptor (HR)-positive breast cancer using tissue microarrays. The profile includes ER, PR, HER2, EGFR, BCL2, and p53 (assessed by IHC) and MYC/8q24 (assessed by FISH). It also directly incorporates the standard clinicopathologic risk factors tumor size, tumor grade, and nodal status to the extent that their prognostic value is not replaced by the molecular markers. Here, we demonstrate validation of the profile in a blinded multi-site study conducted on whole sections. Material and methods: The study was conducted in a blinded fashion using an independent data management firm (Synteract, Carlsbad, CA). Eligible patients were females diagnosed between 1985 and 1997 with HR-positive stage I-IIIA breast cancer treated only with hormone therapy after definitive surgery for whom sufficient tumor samples were available for testing. Slides with FFPE tumor tissue sections and associated clinicopathologic, treatment, and outcome data were provided by four clinical sites through the NCI-funded Cooperative Breast Cancer Tissue Resource program: Fox-Chase Cancer Center (Philadelphia, PA) (n=106), Kaiser Permanente Northwest (Portland, OR) (n=165), University of Miami (FL) (n=55), and Washington University St. Louis (MO) (n=74). Marker assays, scoring, and calculation of risk scores using a predefined algorithm were conducted at Clarient, a CLIA-certified laboratory. The algorithm assigned a risk score to each patient on a scale of 0 to 10+, and a predefined threshold of 3.8 was used to separate patients into low and high risk groups. Results: Complete data for all seven markers was obtained in 349 of the 400 patients. The algorithm defined 28% of the patients (n=99) as high-risk and 72% of the patients (n=250) as low-risk. The high-risk and low-risk patient groups had 10-year distant metastasis rates of 34% and 9%, respectively, resulting in a hazard ratio of 4.7 (95% CI, 2.3−7.8, p<0.0001) through 10 years without an intermediate-risk group. Similar results were achieved when using disease-specific survival (death with evidence of breast cancer recurrence) as the outcome. Using a standard threshold score of 3.4, the Nottingham Prognostic Index (NPI) also stratified the patients (n=349) into low and high risk groups (p=0.03), but only 10% of the patients were in the low-risk category, and the profile further risk stratified patients in the NPI intermediate and high risk categories similar to the overall patient set (p<0.001). In multivariate Cox proportional hazards analysis with clinicopathologic factors, only the profile and tumor size were significant (Wald statistic p<0.001 for both). Discussion: This blinded study, using a predefined algorithm and threshold, validates the prognostic and clinical utility of the multi-marker profile to help guide the appropriate level of adjuvant treatment in breast cancer patients. Citation Information: Cancer Res 2011;71(24 Suppl):Abstract nr P2-12-25.