Enhanced Validation of the FLT3-like Gene Expression Signature As a Predictive Biomarker for Quizartinib Response in FLT3-ITD Negative Acute Myeloid Leukemia: Expanded Cohort and Extended Follow-up from the Pethema Quiwi Trial

Background

The identification of predictive biomarkers is crucial for guiding treatment decisions in acute myeloid leukemia (AML). Previously, we identified a FLT3-like gene expression signature in FLT3 wild-type AML patients, which clustered a proportion of patients with FLT3-ITD and TKD mutated cases. The QUIWI trial, a randomized, placebo-controlled, phase II study (n=273), showed a significant increase in overall survival (OS) in FLT3-ITD negative AML patients treated with the FLT3 inhibitor quizartinib (Quiza) plus intensive chemotherapy (3+7). Our initial study including 161 randomized patients in the QUIWI trial demonstrated that the FLT3-like signature could predict responses to Quiza, highlighting its potential as a biomarker for personalized treatment in AML.

Methods

In this final report, we analyzed a total of 248 patients randomized in the QUIWI clinical trial, including an extended follow-up of the initial patient set (N=161) and an additional set of 87 patients included in the trial. RNA was extracted using standard methods, followed by assessment of nucleic acid integrity (TapeStation) and quantification (Qubit). Total mRNA sequencing was performed using polyA RNAseq with TruSeq technology. The sequences were aligned to the GRCh37 reference genome using the Hisat algorithm. Gene expression quantification was performed using the Bioconductor workflow, and gene expression estimates (FPKM) were obtained and log2 normalized. The original FLT3-like signature (595 genes) was selected for downstream analysis. The method to assign new patients to predefined clusters involved calculating the distances between the new samples and the centroids of the existing groups, then assigning the new samples to the group with the nearest centroid. OS was defined as time from start of screening to death.

Results

The first step was to evaluate the outcomes of patients in the original training set (N=161) with longer patient follow-up (median follow-up 3.20 years and median OS was not yet reached). Within the FLT3-like cluster (N=80), a significant benefit for Quiza over placebo was observed with a p-value of 0.002 and a hazard ratio (HR) of 0.34 (95% CI: 0.17-0.68). No benefit for Quiza over placebo was observed in the non-FLT3-like cluster (p-value 0.43, HR 1.35, 95% CI: 0.64-2.84).

Next, we used the FLT3-clusterization to identify FLT3-like patients among the 87 newly sequenced cases. 24 cases (27.6%) were identified as FLT3-like, and these patients showed a significant benefit for Quiza over placebo (p-value 0.02, HR 0.08, 95% CI: 0.01-0.71). Among these, 50% of cases harbored NPM1 mutations, and 29% had concurrent DNMT3A mutations, but a tendency for better response to Quiza was observed independently of these mutations.

Finally, when merging the two cohorts (N=248), a significant benefit for Quiza over placebo was observed in the FLT3-like group (N=104) with a p-value <0.001, an HR of 0.27, and a 95% CI of 0.14-0.52. This effect was independent of NPM1 and DNMT3A mutations (p-value <0.001).

Conclusion

Our study confirms the predictive value of the FLT3-like signature for identifying FLT3-ITD negative AML patients who benefit from Quiza. The extended follow-up reinforces the significant survival benefit observed in FLT3-like patients treated with Quiza. Furthermore, the identification of new FLT3-like patients who exhibited superior response to Quiza over placebo validates the robustness of the FLT3-like clusterization method. This predictive strategy was effective regardless of NPM1 or DNMT3A mutation status, supporting its potential use in personalized treatment approaches for FLT3-ITD negative AML.