Gene expression profiling for prediction of tissue of origin to direct site-specific therapy in patients with CUP: Treatment outcomes by assay predictive probability.

Abstract

e21023 Background: Tumor profiling as a diagnostic technique in CUP depends on the detection of similarities in gene expression between the cancer and the tissue of origin. Assay-directed site-specific therapy may be most effective in tumors with high-confidence molecular predictions, since a) the prediction may be more often correct due to greater similarity between the gene expression profile of the unknown tumor to the assay reference tumor database, and b) the tumor would more likely share biologic characteristics/treatment sensitivity with cancers of known origin. Methods: In a prospective trial performed by the Sarah Cannon Research Institute (SCRI), a 92-gene RT-PCR assay (CancerTYPE ID; bioTheranostics, Inc.) was performed on tumor biopsies from previously untreated CUP pts. Pts were then assigned standard site-specific treatment directed by the assay prediction. Treatment responses were evaluated based on the strength of the assay predictions; probabilities ≥80% were arbitrarily regarded as high confidence results. Results: Between 10/08 and 12/11, 224 pts (26 predicted tumor types) received treatment. Median overall survival (OS) was longer in 113 pts (50%) with high confidence predictions: 11.7 months (mos) vs 9.8 mos; p=.03. 120 pts (59%) had assay predictions of responsive tumor types, known to derive substantial benefit from standard site-specific therapy. Treatment responses, irrespective of predictive probability, were better in these pts vs those with less responsive tumor types (median OS 12.8 vs 7.4 mos; p=.027). Further stratification of high vs lower confidence assay predictions resulted in 4 groups with different OS (p=.03) (Table). Conclusions: Site-specific therapy for CUP pts directed by results of the 92-gene assay was more effective when the probability of the assay prediction was higher, particularly in pts with assay diagnoses of more responsive tumor types. [Table: see text]