FLT3mutation Assay Laboratory Cross Validation: Results from the CALGB 10603/Ratify Trial in Patients with Newly Diagnosed FLT3-Mutated Acute Myeloid Leukemia (AML)

Introduction:FLT3-mutations (FLT3^mut) are the most common molecular alteration in younger adult patients with AML. The two major variants are internal tandem duplication (ITD) mutations mainly affecting the juxtamembrane domain and point mutations (mut) located in the second tyrosine kinase domain (TKD). FLT3-ITD mut, especially those with a high allelic ratio (AR), are associated with a high rate of treatment failure. In line with this, the 2017 European Leukemia Network (ELN) recommendations categorize patients with a high AR (i.e. >0.5 mutant/wildtype; wt) with NPM1-wt within the adverse risk group. The recent approval of midostaurin for the treatment of newly diagnosed FLT3-mutant AML underscores the need for reliable detection of FLT3^mut. Several methods have been described, but most laboratories (labs) use PCR combined with capillary electrophoresis (CE), either alone (for ITD-mutations) or combined with a restriction enzyme digest (for TKD- point mutations). Despite the high clinical importance of FLT3^mut, data on the diagnostic accuracy of available methods are scarce. We here describe our experience in FLT3^mut testing gained in the context of the prospective international CALGB 10603/RATIFY study.

Methods: CALGB 10603/RATIFY was a phase 3, randomized study in patients (18-<60 yrs of age) with newly diagnosed, FLT3^mut-AML. Patients enrolled were screened for FLT3^mutprior to randomization, using a clinical cut-off for the AR of ≥0.05 forFLT3-positivity, and stratified as FLT3-TKD, FLT3-ITD^low(AR 0.05-0.7), orFLT3-ITD^high(AR >0.7). Testing was performed in 9 reference labs across 6 countries using a harmonized PCR-method based on CE-detection (Stone et al, NEJM, 2017). PCR was done in triplicate starting from the DNA, mean values of these measurements were reported. To ensure consistency between labs, a cross validation quality control (CVQC) procedure was performed every 6 months. A set of 50 test samples (whole-genome amplified DNA from patients with known FLT3status) was prepared centrally and distributed in a blinded fashion. Per round, 25 of these samples (5 ITD wt, 5 ITD AR cut-off positive ≈0.05, 5 ITD AR ≈0.7, 5 TKD wt, and 5 TKD AR cut-off positive ≈0.05)were analyzed. Each lab measured the AR of each sample with 3 reactions. For the analysis reported here, we re-evaluated all data collected in this CVQC effort to study the overall accuracy of results as well as the intra- and interlab variability.

Results:In 8 rounds of CVQC, 5700 single tests were performed between 12/2007 and 6/2011. Results were centrally scored using the criteria mentioned above. The overall consistency with respect to the qualitative information (FLT3^mutpos./neg.) was high, with more than 95% of measurements meeting the prespecified criteria. We calculated the "true" value for each sample using the median for each sample over all performed analyses. The median "true" AR values were 0.72 (range 0.61-0.93; ITD-high), 0.06 (range 0-0.09; ITD-cut-off) and 0.07 (0.04-0.12; TKD-cut-off). Comparing these true values with the individual results, we saw considerable intra- and interlab variability with respect to accuracy of the measurements. The median error (ME) from this true value was 11.2% (95%CI 10.5-11.9%), it was lower for ITD high samples (7.1%; 95%CI 6.1-8.3%) than ITD-cut-off (13.4%; 95%CI 11.5-16.7%) and TKD (16.2%; 95%CI 13.3.-17.6%). The ME was significantly reduced, when the triplicate values were used, e.g. ME ITD AR 0.7 5.7% (95%CI 4.8-6.6%; p<0.001), as was the coefficient of variation of the measurements (p=.003). Using the prespecified cut-off for AR of 0.7, 89 of 379 values were wrongly assigned to be above/below the cut-off. Incorrect allocation was more common in samples with a true value within 1-standard-deviation from the cut-off (i.e. between an AR of 0.63 - 0.77; 39% incorrect), for AR-values outside this range, assignment was significantly more accurate (9.1% incorrect; p<0.001).

Conclusions:This first report of round robin testing for FLT3^mutshows that using a standardized protocol, the qualitative assessment of FLT3^mutis feasible with high accuracy. However, the assessment of the FLT3-AR clearly shows a considerable variability, which can be reduced by using a triplicate analysis, but still has to be taken into account, especially in patients evaluated for allogeneic stem cell transplantation. Further standardization of FLT3-testing appears highly warranted.