P5-13-24: A Predictive Model of Early Systemic Disease Relapse after Standard Adjuvant Therapy for Breast Cancer.

Abstract

Abstract Background Early relapse after adjuvant therapy for breast cancer is very discouraging and remains a major problem. We sought to identify predictors of early relapse risk and build a predictive model for relapse using prospectively collected data for patients seen at the Markey Cancer Center starting 2007 to date. Methods: Of the 1098 new patients seen, 814 patients had stage I-III disease and were further analyzed for predictors of early relapse risk. Univariate analyses were performed for key variables including patient age, tumor size, grade, estrogen receptor (ER) status, progesterone receptor (PgR) status, and HER2 status. A multivariate Cox regression model was built to identify predictors of systemic relapse and model-building was performed using step-wise model selection to determine candidate models. A risk score was developed based on the linear combination of covariates in the final Cox model. Time-dependent predictive curves, a newly developed statistical methodology, were used to evaluate the predictive accuracy of the proposed risk score. Results: Median patient age was 57 years (Range 25–92) and 88% were white. Forty six (46) % had stage I disease, 36% stage II, and 18% stage III. Median follow up time was 2.3 years. Of this 814 patient cohort, 708 patients had complete baseline covariate data and were used to build the candidate models. The final Cox regression model included 5 covariates that were significantly associated with risk of early relapse: stage III disease (p = 0.0011), grade III (p = 0.0028), PgR-negative status (p = 0.0121), HER2−negative status (p = 0.0305), and node-positive status (p = 0.0360). These five covariates were then used to calculate an early recurrence risk score, which is the weighted average of these risk factors when present, with the weights being the coefficients from the Cox regression model. The 1-year, 2-year and 3-year predictive curves for this risk score decrease considerably, especially for the 2-year and 3-year curves, indicating good predictive accuracy of the risk score. The highest risk score group, which represents 4.8% of the population, has a 1-year, 2-year and 3-year relapse probabilities of 13.0% (95% CI: 4.1%, 27.3%), 39.4 % (95% CI: 20.1%, 58.3%), and 52.3% (95% CI: 28.5%, 71.5%), respectively. In comparison, for the overall population, the corresponding 1-year, 2-year, and 3-year relapse probabilities were only 1.1% (95% CI: 0.5%, 2.1%), 4.2% (95% CI: 2.7%, 6.1%) and 6.2% (95% CI: 4.2%, 8.6%), respectively. Conclusions: The developed risk score based on stage, tumor grade, PgR, HER2, and node status is highly predictive of early relapse in breast cancer patients after standard adjuvant therapy. Our model can be used to identify patients with high risk of early disease relapse who may otherwise benefit from enrollment on novel adjuvant therapeutic trials to improve their outcome. Citation Information: Cancer Res 2011;71(24 Suppl):Abstract nr P5-13-24.