Comparative Analysis of Persistent Mutations at Pre- and Post-Transplantation in Patients with Acute Myeloid Leukemia Undergoing Allogeneic Stem Cell Transplant

Introduction Selected gene mutations are known valuable diagnostic and prognostic markers at diagnosis in AML. However, persistence of these genetic alterations after induction chemotherapy and peri-allogeneic stem cell transplantation (allo-HSCT) are less well-characterized. Objective We evaluate the prognostic impact of persistent mutations pre- and post-allo-HSCT on overall survival from transplant (OS) and relapse-free survival (RFS). Methods We conducted a retrospective study of patients who underwent allo-HSCT between January 2014 to June 2020. Next-generation sequencing using a 23 gene panel was performed on bone marrow biopsies at diagnosis, pre-transplant, and post-transplant. Survival analyses were conducted using several methods including cox regression models and Kaplan-Meier curves. Results Our cohort consisted of 101 patients with 68 having one or more mutations at diagnosis. After induction therapy, 37% of patients had complete clearance of mutations. Among the patients who had persistent mutations after induction therapy, 53% were cleared after allo-HSCT.In a univariable analysis, persistence of FLT3-ITD and NRAS at pre-and post-transplant were associated with worse OS with a HR 4.21 (1.25 - 14.14, p = 0.02), HR 6.72 (0.88-51.41, p = 0.066) and HR 7.07 (2.13 - 23.39, p = 0.001), HR 3.58 (1.10-11.6, p = 0.034) respectively. In addition, CBL (HR 15.29 (1.84-127), p = 0.012), ETV6 (HR 2. 91 (0.69-12.32), p = 0.146), IDH1 (HR 4.89 (1.15-20.85), p = 0.032) and FLT3-TKD (HR 4.16 (0.54-31.91), p = 0.171) were associated with worse OS at pre-transplant. ASXL1 (HR 3.21 (1.28-8.09), p = 0.013) was associated with worse OS at post-transplant.In our multivariable cox regression models, addition of mutational data using a purposeful selection method to the clinical and demographic data consistently improved the c-statistic for both OS and RFS at pre- and post-transplant. Persistence of gene mutations including FLT3-ITD at pre-transplant (HR 6.99 [1.00 - 49.01], p = 0.050) and post-transplant (HR 8.76 [1.71 - 44.88], p = 0.009) remained statistically significant for OS.We also constructed multivariable models using a backward variable selection method. The presence of ASXL1, CBL, ETV6, FLT3-TKD, and FLT3-ITD at pre-transplant and the presence of ASXL1, NRAS, JAK2, and FLT3-ITD at post-transplant were associated with worse OS.We analyzed the dataset using multiple machine learning algorithms. The random forest method provided the best AUC and accuracy. Age and ELN risk were the most impactful clinical predictors for OS at pre- and post-transplant. The presence of ASXL1, DNMT3A and TET2 at pre-transplant and FLT3-ITD, CBL and TET2 mutations at post-transplant were the most important predictors of OS. Conclusion Persistence of FLT3-ITD mutations post- allo-HSCT may have prognostic value in patients with AML using several statistical methods and machine learning algorithms.