Simpson's Paradox and the Impact of Different DNMT3A Mutations on Outcome in Younger Adults With Acute Myeloid Leukemia.

PURPOSE To evaluate the impact of DNMT3A mutations on outcome in younger patients with cytogenetic intermediate-risk acute myeloid leukemia. PATIENTS AND METHODS Diagnostic samples from 914 patients (97% < 60 years old) were screened for mutations in DNMT3A exons 13 to 23. Clinical outcome was evaluated according to presence or absence of a mutation and stratified according to type of mutation (R882, non-R882 missense, or truncation). RESULTS DNMT3A mutations (DNMT3A(MUT)) were identified in 272 patients (30%) and associated with a poorer prognosis than wild-type DNMT3A, but the difference was only seen when the results were stratified according to NPM1 genotype. This example of Simpson's paradox results from the high coincidence of DNMT3A and NPM1 mutations (80% of patients with DNMT3A(MUT) had NPM1 mutations), where the two mutations have opposing prognostic impact. In the stratified analyses, relapse in patients with DNMT3A(MUT) was higher (hazard ratio, 1.35; 95% CI, 1.07 to 1.72; P = .01), and overall survival was lower (hazard ratio, 1.37; 95% CI, 1.12 to 1.87; P = .002). The impact of DNMT3A(MUT) did not differ according to NPM1 genotype (test for heterogeneity: relapse, P = .4; overall survival, P = .9). Further analysis according to the type of DNMT3A mutation indicated that outcome was comparable in patients with R882 and non-R882 missense mutants, whereas in those with truncation mutants, it was comparable to wild-type DNMT3A. CONCLUSION These data confirm that presence of a DNMT3A mutation should be considered as a poor-risk prognostic factor, irrespective of the NPM1 genotype, and suggest that further consideration should be given to the type of DNMT3A mutation.