Prediction model for detecting circulating tumor DNA (ctDNA) in metastatic colorectal cancer (mCRC).

Allan Andresson Lima Pereira Hospital Sírio-Libanês, Brasília, Brazil info_outline Allan Andresson Lima Pereira, Aparna Raj Parikh, Emily E. Van Seventer, Jingquan Jia, Jonathan M. Loree, Preeti Kanikarla Marie, Kanwal Pratap Singh Raghav, Van K. Morris, Michael J. Overman, Victoria M. Raymond, Richard B. Lanman, AmirAli Talasaz, John H. Strickler, Ryan Bruce Corcoran, Scott Kopetz

Authors Allan Andresson Lima Pereira Hospital Sírio-Libanês, Brasília, Brazil info_outline Allan Andresson Lima Pereira, Aparna Raj Parikh, Emily E. Van Seventer, Jingquan Jia, Jonathan M. Loree, Preeti Kanikarla Marie, Kanwal Pratap Singh Raghav, Van K. Morris, Michael J. Overman, Victoria M. Raymond, Richard B. Lanman, AmirAli Talasaz, John H. Strickler, Ryan Bruce Corcoran, Scott Kopetz Organizations Hospital Sírio-Libanês, Brasília, Brazil, University of California, San Francisco, San Francisco, CA, Massachusetts General Hospital, Boston, MA, Duke University Medical Center, Durham, NC, BC Cancer, Vancouver, BC, Canada, University of Texas MD Anderson Cancer Center, Houston, TX, Department of Gastrointestinal Medical Oncology, The University of Texas MD Anderson Cancer Center, Houston, TX, The University of Texas MD Anderson Cancer Center, Houston, TX, Guardant Health, Inc., Redwood City, CA, UT MD Anderson Cancer Center, Houston, TX Abstract Disclosures Research Funding Pharmaceutical/Biotech Company Background: While tissue-based assays have yields above 90% in solid tumors, there is less known about factors that influence the sensitivity of ctDNA for detecting mutations. Met hods: We retrospectively evaluated mCRC patients (pts) who had plasma-derived NGS utilizing a highly-sensitive targeted 68-73-gene ctDNA assay. In a case-control design, pts with a known mutation on tissue and radiologic evidence of metastatic disease but no detectable ctDNA mutation were matched 1:3 with randomly selected pts with detectable mutations and compared according to clinical, laboratory, and radiologic characteristics. A prediction score for ctDNA detection was built using a binary logistic backward stepwise regression analysis and tested in two independent data sets from different institutions. Area under the curve (AUC) from receiver operating characteristics curves (ROC) were used for internal and external validation. Results: From 416 pts who met inclusion criteria, plasma-derived NGS did not find tumor mutations in 66 cases (15.9%); 198 pts with detectable alterations were selected as controls. After multivariate analysis, the detection of ctDNA was associated with increasing age (OR 1.05; 95%CI 1.02-1.09; p = .001), presence of liver (OR 5.82; 95%CI 2.55-12.49; p < .001) and lymph node metastases (OR 3.28; 95%CI 1.51-7.60; p = .004), archival TP53 mutations (OR = 2.88; 95%CI 1.37-6.17; p = .006). A key determinant was timing of collection relative to disease status: plasma collected in newly diagnosed metastatic disease or after evidence of progression was substantially more likely to have detectable alterations (OR 9.24; 95%CI 4.11-22.40; p < .001); The simplified prediction model performed well in internal (AUC = 0.88) and external validation (AUC = 0.95; 163 pts). Conclusions: Our validated prediction model provides clinicians and researchers with a tool to screen for patients in whom ctDNA testing can outperform tissue-based testing in detecting genomic alterations.