James V. Tricoli, Linda Zane, Robin Harrington, Laura Yee, Kneshay N. Harper, Ting-Chia Chang, Lyndsay Harris, Alice P. Chen, Keith Flaherty, Peter J. O'Dwyer, Barbara A. Conley, Cynthia Winter, Jennifer Lee, Paul M. Williams, Jeffrey Sklar, David Patton, Gregory J. Tsongalis, Stanley R. Hamilton, A. John Iafrate, and Chris Alan Karlovich
Background: NCI-MATCH is a precision medicine trial that assigns treatment to refractory cancer patients by tumor mutation profile rather than by histology. After screening fresh tumor biopsies from nearly 6000 patients many treatment arms did not meet accrual due to the low prevalence of the eligible variants. NCI MATCH developed an approach to identify patients for the remaining arms utilizing a network of academic and commercial CLIA-certified labs that perform NGS assays as routine care at MATCH participating sites.
Methods: Candidate labs were recruited through a notice in the Federal Register and posted on the NCI and ECOG ACRIN web sites. Twenty-seven labs (17 academic/10 commercial) submitted applications. After acceptance each lab analyzed a common set of 10 DNAs extracted from 8 cell lines and 2 clinical samples for concordance with the central NCI-MATCH NGS assay.
Results: For the 17 labs with concordance results, a median of 8 (range 2 – 58) copy number variants (CNVs) were evaluated by the NGS assay of each DL, with the number evaluated depending on each lab’s clinical assay panel content. CNV concordance between central and DL assays, as measured by positive percent agreement (PPA), averaged 98.7% (range 87.5% – 100%) with the central assay as referent and 94.1% (range 77.8% – 100%) with the DL assay as referent. For single nucleotide variants (SNVs) and Insertion/deletions (Indels) combined, a median of 19 variants (range 11 – 26) were evaluated by each DL for concordance. PPA between central and DL assays averaged 98.0% (range 87.5% – 100%) and 98.6% (range 90.0% – 100%) with central and DL assay as referents, respectively. Strong correlations were observed between central and DL assays for both CNVs (median r = 0.93; 0.33 – 1.00) and SNV/Indels (median r = 0.98; 0.67 – 0.99).
Conclusions: Our results suggest that different NGS assay platforms using diverse strategies for target enrichment and data analysis may still achieve high concordance if pre-analytical variables are minimized and the common genomic regions interrogated by each assay are well-understood. The designated lab network allows for a wider search for rare variants in tumors and provides a model for conducting future clinical trials. Clinical trial information: NCT02465060
VIEW THE PUBLICATION