Multisite validation of a 92-gene molecular classifier: Diagnostic accuracy and clinical utility in metastatic cancer.

Abstract

10588 Background: Accurate tumor classification is increasingly critical to individualized care as patient outcomes improve with use of targeted cancer therapies and predictive biomarkers. In metastatic cancers of uncertain or unknown origin, identification of a primary site remains equivocal in a significant number of cases. Gene expression signatures may improve accuracy and specificity of tumor classification, however, large-scale, blinded validation studies that demonstrate stable performance in metastatic and poorly differentiated cases are currently lacking. Methods: Seven hundred and ninety cases (51% female, 44% metastatic, 63% grade 2 and 3, 14% limited tissue specimens) representing 28 tumor types and 50 subtypes were processed and adjudicated between Mayo Clinic, UCLA and Massachusetts General Hospital. Blinded FFPE tumor sections were submitted and tested using a 92-gene RT-PCR assay (CancerTYPE ID, bioTheranostics Inc.). Molecular predictions were evaluated for concordance with the reference diagnosis; diagnostic accuracy between clinical subsets was compared. Results: The 92-gene assay demonstrated overall sensitivities of 87% for tumor classification and 82% for subtyping. Forty-seven (5.9%) cases were considered unclassifiable. Comparative analysis of performance between metastatic (n=329) and primary (n=414) tumors showed no statistically significant difference in accuracy (85% vs 88%, p=0.16). Similarly, no significant decreases in performance were observed across histological grades (p=0.58) and when comparing limited tissue to excisional biopsy specimens (p=0.16). Strong precision was demonstrated for accurate identification of a primary tumor in tissues biopsied from common metastatic sites, reported as positive predictive values of 100% for lung, brain and peritoneum, 92% for ovary and 80% for liver. Conclusions: Results from this multisite, blinded study validate the diagnostic accuracy of the 92-gene assay for classifying a diverse set of tumors, and support its clinical utility in the diagnosis of metastatic tumors to determine tissue of origin, and in the differential diagnosis of metastatic vs new primary disease.