FIGURE 1.

An approximate comparison of the sensitivity (left axis) and fraction of monitorable cases for commonly used AML MRD methods. Approximate sensitivity of molecular methods was converted from VAFs to number of mutated cells (double the VAF) to facilitate comparisons across methods. Cytogenetic analysis assumes that 50% of AML cases have at least one cytogenetic abnormality (Haferlach et al., 2012) with a detection sensitivity of one clonal cell in 20 metaphases. Morphologic evaluation assumes that a blast count of > 5% is considered relapse. FISH assumes a sensitivity of two clonal events in 200 evaluated metaphases. Panel-based NGS assumes a 40-50 gene panel capable of detecting at least one mutation in 85% of AML patients with a minimum sensitivity of 0.1% VAF (Cancer Genome Atlas Research Network et al., 2013). NPM1 qPCR assumes a sensitivity of 0.001% VAF and an NPM1 prevalence of 27% (Jongen-Lavrencic et al., 2018). qPCR for translocations assumes monitoring of inv(16), t(8;21), and t(15;17) with a prevalence of 13% (Papaemmanuil et al., 2016). Comparisons to AML MRD flow are complicated as the assay sensitivity depends to a large extent on the exact phenotype of the leukemia. The exact fraction of cases amenable to flow-based MRD is uncertain as indicated by the hashed line, but is modeled at 100% with a sensitivity of 0.0002% (Loken et al., 2012; Zhou and Wood, 2017; Schuurhuis et al., 2018). We also note that there may be discrepancies between MRD flow-based methods and molecular MRD (Jongen-Lavrencic et al., 2018).