Plasma cell-free DNA (cfDNA) sequencing is a compelling diagnostic tool in solid tumors and has been shown to have high positive predictive value. However, limited assay sensitivity means that negative plasma genotyping, or the absence of detection of mutation of interest, still requires reflex tumor biopsy.
Methods: We analyzed two independent cohorts of patients with advanced non-small-cell lung cancer (NSCLC) with known canonical driver and resistance mutations who underwent plasma cfDNA genotyping. We measured quantitative features, such as maximum allelic frequency (mAF), as clinically available measures of cfDNA tumor content, and studied their relationship with assay sensitivity.
Results: In patients with EGFR-mutant NSCLC harboring EGFR T790M, detection of driver mutation at > 1% AF conferred a sensitivity of 97% (368/380) for detection of T790M across three cfDNA genotyping platforms. Similarly, in a second cohort of patients with EGFR or KRAS driver mutations, when the mAF of nontarget mutations was > 1%, sensitivity for driver mutation detection was 100% (43/43). Combining the two NSCLC patient cohorts, the presence of nontarget mutations at mAF > 1% predicts for high sensitivity (> 95%) for identifying the presence of the known driver mutation, whereas mAF of ≤ 1% confers sensitivity of only 26%-54% across platforms. Focusing on 21 false-negative cases where the driver mutation was not detected on plasma next-generation sequencing, other mutations (presumably clonal hematopoiesis) were detected at ≤ 1% AF in 14 (67%).
Conclusion: Plasma cfDNA genotyping is highly sensitive when adequate tumor DNA content is present. The likelihood of a false-negative cfDNA genotyping result is low in a sample with evidence of > 1% tumor content. Bioinformatic approaches are needed to further optimize the assessment of cfDNA tumor content in plasma genotyping assays.
© 2021 by American Society of Clinical Oncology.