Molecular classification of cancer with the 92-gene assay in cytology and limited tissue samples

Abstract

Background: Detailed molecular evaluation of cytology and limited tissue samples is increasingly becoming the standard for cancer care. Reproducible and accurate diagnostic approaches with reduced demands on cellularity are an ongoing unmet need. This study evaluated the performance of a 92-gene assay for molecular diagnosis of tumor type/subtype in cytology and limited tissue samples.

Methods: Clinical validation of accuracy for the 92-gene assay in limited tissue samples such as cytology cell blocks, core biopsies and small excisions was conducted in a blinded multi-institutional study (N = 109, 48% metastatic, 53% grade II and III). Analytical success rate and diagnostic utility were evaluated in a consecutive series of 644 cytology cases submitted for clinical testing.

Results: The 92-gene assay demonstrated 91% sensitivity (95% CI [0.84, 0.95]) for tumor classification, with high accuracy maintained irrespective of specimen type (100%, 92%, and 86% in FNA/cytology cell blocks, core biopsies, and small excisions, respectively; p = 0.26). The assay performed equally well for metastatic versus primary tumors (90% vs 93%, p = 0.73), and across histologic grades (100%, 90%, 89%, in grades I, II, and III, respectively; p = 0.75). In the clinical case series, a molecular diagnosis was reported in 87% of the 644 samples, identifying 23 different tumor types and allowing for additional mutational analysis in selected cases.

Conclusions: These findings demonstrate high accuracy and analytical success rate of the 92-gene assay, supporting its utility in the molecular diagnosis of cancer for specimens with limited tissue.

Keywords: biological markers; clinical oncology; cytology; gene expression profiling; molecular targeted therapy.

Figure 1. Hematoxylin and eosin stained specimens from limited tissue sample types

A. & B. Fine needle aspirate (FNA)/cell block showing breast adenocarcinoma (10X and 400X). C. & D. Core biopsy showing colorectal adenocarcinoma metastatic to the liver (10X and 400X). E. & F. Small excisional biopsy showing renal cell carcinoma metastatic to the lung (10X and 400X).

Figure 2

A. Distribution of biopsy sites from clinical validation cohort. *Biopsy sites with three or fewer cases were combined into a single category, which includes: Breast (3), Adrenal (2), Mediastinum (2), Prostate (2), Retroperitoneum (1). B. Distribution of molecular diagnoses as determined by the 92-gene assay in the limited tissue data set. Molecular diagnoses labeled on the y-axis. Number of unique biopsy sites for each molecular diagnosis is labeled to the right of the graph. C. Sensitivity of the 92-gene assay for each specimen type in the limited tissue data set (p = 0.26). Specimen type labeled along the x-axis. Sensitivity expressed as a percentage is shown by each column. Exact values per specimen type are shown in white. Overall sensitivity for the 92-gene assay was 91%.

Figure 3

A. Distribution of biopsy sites within the clinical cytology case series that passed analytical QC (N = 558). *Biopsy sites with three or fewer cases were combined into a single category, which includes: Mediastinum (3), Bowel (2), Breast (2), Kidney (2), Adrenal (1), Periaortic (1), Thyroid (1). B. Distribution of molecular diagnoses as determined by the 92-gene assay in the clinical cytology data set that passed analytical QC (N = 558). Molecular diagnoses labeled on the y-axis. Number of unique biopsy sites for each molecular diagnosis is labeled to the right of the graph.

Figure 4. Proposed tissue-based diagnostic algorithm

Adapted from (Schnabel & Erlander, Expert Opin. Med. Diag. 2012). Abbreviations: IHC = Immunohistochemistry, FISH = Fluorescence in-situ hybridization, NGS = Next generation sequencing.