Background NPM1 mutated (mut) AML is considered favourable risk unless associated with FLT3 internal tandem duplication (ITD) or adverse karyotype. However many other genetic and clinical risk factors have been identified, including high white cell count (WCC), secondary disease, co-mutations and presence of a “triple hit” genotype ( NPM1mut, FLT3 ITD and DNMT3Amut). Although alternative classification systems have been proposed, few studies have been large enough to robustly evaluate clinical and molecular subgroups within NPM1mut, and none have incorporated the effects of measurable residual disease (MRD) status or treatment intensity. Therefore, clinical decision making remains challenging and inconsistent. Here, we performed comprehensive genotyping and molecular MRD assessment in two consecutive prospective randomised studies to clarify these issues. Methods NCRI AML17 (2009-2014) and AML19 (2015-2020) were large randomised trials for younger adults with newly diagnosed AML. Induction regimens included DA +/- etoposide (ADE) and FLAG-IDA, with or without Gemtuzumab. Patients underwent sequential MRD assessment by RT-qPCR, and diagnostic DNA samples were analysed by targeted NGS. Overall survival (OS) was calculated with the Kaplan-Meier method, while cumulative incidence of relapse (CIR) considered death as a competing risk. Relapse events included MRD relapse. All reported hazard/odds ratios are from multivariable analyses including age, prior haematological malignancy, prior chemo/radiotherapy, WCC (log transformed), cytogenetic risk and mutation profile ( FLT3 ITD, DNMT3A, PTPN11, N/KRAS, TET2, IDH1, IDH2, WT1, TP53 MDS-related genes). Results 1357 patients with NPM1mutAML were identified, 888 in AML17 and 469 in AML19. 1100 patients had DNA available for panel sequencing, and 981 underwent MRD monitoring of whom 737 had an evaluable post course 2 (PC2) peripheral blood (PB) sample. Morphological remission (CR+CRi) was achieved in 1264/1357 patients (93%) and PC2 PB MRD negativity in 594/737 (81%). We confirmed the adverse impact of PB PC2 MRD positivity: 3y CIR was 65% vs 29% in those MRD- (p<0.001), and 3yr OS was 40% vs 79% (p<0.001). For PB PC2 MRD+ patients, CIR was >50% in all subgroups. There was a lower rate of PC2 PB MRD negativity in patients with high presenting WCC (OR 1.14, p=0.05), FLT3 ITD (OR 2.58, p<0.001), DNMT3Amut (OR 2.2, p<0.001) and WT1mut (HR 2.04, p=0.02). There was no effect of FLT3 allelic ratio, while patients with both FLT3 ITD and DNMT3A mutations had the lowest rate of MRD- (64%, Figure A). Of the 594 PC2 MRD- patients, we observed higher 3y CIR in patients with high WCC (HR 1.14, p=0.01), DNMT3Amut (HR 1.83, p=0.001), WT1mut(HR 1.96, p=0.01) and IDH1mut (HR 1.9, p=0.002), and lower CIR in patients with N/KRASmut(HR 0.53, p=0.01). Although presence of FLT3 ITD alone was not associated with CIR (HR 0.88, p=0.7), patients with both FLT3 ITD and DNMT3A mutations had 3y CIR from MRD- of 40% (HR 1.74, p=0.03). MDS-related gene mutations had no impact on achieving MRD-, nor relapse from MRD-. Considering all patients, OS was poorer with increasing age (HR 1.03, p<0.001), WCC (HR 1.07, p<0.001), adverse karyotype (HR 2.48, p=0.01), FLT3 ITD (HR 1.34, p=0.01), DNMT3Amut (HR 1.56, p<0.001) and WT1mut(HR 1.71, p<0.001). 3y OS was 52% in patients with both FLT3 ITD and DNMT3Amut, 63% with FLT3 ITD only, 67% with DNMT3A only and 75% with neither. Considering MRD- patients, only age (HR 1.02, p=0.04), DNMT3Amut(HR 1.75, p=0.007) and WT1mut (HR 1.8, p=0.048) remained associated with OS. There was no benefit to CR1 allogeneic transplant in MRD- patients (HR 0.83, 95%CI 0.51-1.34, p=0.4), including in any subgroups (heterogeneity p=0.9). Compared with other treatments, FLAG-Ida increased MRD- (87% vs 79%, p=0.009), reduced relapse from MRD- (16% vs 34%, p<0.001) and improved 3y OS (79% vs 63%, HR 0.52, p<0.001). The OS benefit of FLAG-Ida was greatest in patients with at least one of the features associated with failure to achieve MRD- and relapse from MRD- (high WCC, triple hit or WT1mut, HR 0.39 95%CI 0.26-0.58, heterogeneity p<0.01). Conclusion The “triple hit” ( NPM1, DNMT3A and FLT3 ITD) genotype, WT1mut and high WCC were associated with poor outcomes, due to a lower probability of achieving MRD-, and a higher relapse risk from MRD- remission. Survival was improved for patients receiving intensified induction with FLAG-Ida including patients in these high-risk groups.