Abstract P1-06-11: Prediction of oncotype DX® results based on local gene expression measurements by MammaTyper®

Abstract

Abstract Background: Oncotype DX® recurrence score (RS) has emerged as a recommended risk classifier for patients with ER+/HER2- early-stage breast cancer. While RS is one of the most rigorously studied risk scores, it is also one of the most expensive tests, thus remaining beyond reach for a many patients. The necessity for an affordable method for estimating risk of recurrence has motivated investigations on the correlation between RS and traditional parameters such as IHC for ER, PR and Ki67. However, semi-quantitative IHC lacks standardization across different laboratories especially for Ki67. In this study we therefore investigated whether the standardized assessment of HER2, ER, PR, and Ki67 on mRNA level could better serve for prediction of low risk RS cases. Methods: ERBB2, ESR1, PGR and MKI67 mRNA expression was measured by RT-qPCR in extracts from FFPE breast cancer samples using the MammaTyper® test. Complete data for RS, IHC, grading and mRNA measurement was available for 198 samples. Tumor subtypes according to St Gallen surrogate definition from 2013 were assigned based on binary mRNA marker classification (pos/neg) according to pre-defined cut-offs. Subtype results were compared to RS risk classes based on commercial and TAILORx-trial cut-offs. RS low risk classification (RS ≤25) based on four IHC markers and grading was estimated using the online tool breastrecurrenceestimator.onc.jhmi.edu and compared to observed RS classes. Finally, the prediction of continuous RS values by mRNA or semi-quantitative IHC measurement was compared by linear regression and subsequent ROC analysis of prediction models. Results: The distribution of RS risk classes in the set of samples with full data was 21% RS 0-10, 39% RS 11-17, 27% RS 18-25, 7% RS 26-30 and 7% RS >30. MammaTyper® called 38% (76) of the samples as Luminal A-like. From these samples 70% and 99% had RS values below 18 and 25 respectively. Only 1 MammaTyper® Luminal A-like sample had an RS >30. Estimation of RS according to the online tool resulted in classification of 61% (121) of the samples as low risk (RS ≤25). Of these 74% and 98% of samples had observed RS values below 18 and between 18 and 25 respectively. 2 and 1 samples called as low risk by the online tool had an RS of 26-30 and >30 and, respectively. In linear regression analysis of IHC against RS only PR and Ki67 were significant predictors (p-values <0.0001 and 0.0128) while when using mRNA values ESR1, PGR and MKI67 were found as predictors of RS in the multivariate model (all p-values <0.0001). On a training set (67% of samples) the IHC based prediction model was correlated to the observed RS with an R2 of 0.305 whereas the mRNA based model achieved an R2 of 0.489. When the models were applied to training and validation dataset (33% of samples) for prediction of an RS >25 result, the IHC based model had AUCs of 0.887 and 0.836, respectively, while the mRNA based model achieved AUCs of 0.909 and 0.899, respectively. Conclusion: mRNA based prediction of RS was considerably better than prediction based on IHC. As Ki67 IHC standardization is reaching its limits, local gene expression measurements with their high degree of standardization could serve as a safer way for prediction of Oncotype low risk results. Citation Format: Lehr H-A, Aulmann S, Laible M, Etzrodt A, Hartmann K, Gürtler C, Sahin U, Varga Z. Prediction of oncotype DX® results based on local gene expression measurements by MammaTyper® [abstract]. In: Proceedings of the 2017 San Antonio Breast Cancer Symposium; 2017 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2018;78(4 Suppl):Abstract nr P1-06-11.