Time Dependence of Gene-Expression Signatures to Predict Breast Cancer Recurrence

• Activity Number: 253
• Type: Contributed
• Date/Time: Monday, July 30, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Kit Fun Lau*+
• Companies: VitaPath Genetics
• Keywords: Gene expression signature ; time-dependent Cox model ; annualized hazard rates ; breast cancer

Abstract

Breast cancer signatures, namely Recurrence Score (RS) of Oncotype, 70-gene profile (SGP) of Mammaprint, Intrinsic Subtype (IS) and Wound Signature (WS), were analyzed for their performances over time using a single dataset of 295 samples from Fan et al.1 Time-varying coefficient Cox models were used to estimate the time-dependent hazard ratios (HR) for recurrence-free survival (RFS). Annualized hazard rates were estimated using cubic splines. Concordance between RS and SGP was calculated for different IS.

For the high risk groups of RS, SGP, IS and WS, hazard rates peaked around the 2nd year from surgery and decreased rapidly over time. For the low risk groups, hazard rates also peaked around the 2nd yr but remained fairly constant over time. Hazard ratios of high vs. low risk groups of RS at yr 2, 5 and 10 were 17.0 (8.4-34.0), 4.0 (2.0-8.0), 1.7 (0.83-3.4). Poor vs. good groups of SGP had HRs of 7.4 (4.5-12.1), 3.6 (2.2-5.9), 2.0 (1.2-3.3) at yr 2, 5, 10. Concordance rates of RS and SGP varied from 1.0 in Basal-like to 0.62 in Normal-like of IS.

Ref. 1: Fan et al. N Engl J of Med 2006; 355:560-9.