A nomogram based on clinical features and molecular abnormalities for predicting the prognosis of patients with acute myeloid leukemia

Background: The high clinical and molecular heterogeneity of acute myeloid leukemia (AML) has led to an unsatisfactory clinical prognosis, thus we sought to incorporate both clinical features and molecular abnormalities to construct a new prognostic model. Methods: A database search of the Gene Expression Omnibus (GEO) revealed 238 cases of adult AML. The independent risk factors were assessed using both univariate and multivariate Cox regression, as well as least absolute shrinkage and selection operator (LASSO) regression. The predictive accuracy, discriminatory power and clinical applicability of the nomogram were determined by the consistency index (C-index), calibration curves and decision curve analysis (DCA). In addition, a single-centre cohort of 135 cases was used for external validation. Results: Multivariate Cox regression analysis showed that the independent influences on overall survival (OS) were age, type of disease, DNMT3A, IDH2 and TP53 mutations. The area under the curve (AUC) values for the training set were 0.755, 0.745 and 0.757 at 1, 2 and 3 years respectively; the AUC for the validation set were 0.648, 0.648 and 0.654 at 1, 2 and 3 years; and the AUC for the northwest China set were 0.692, 0.724 and 0.689 at 1, 2 and 3 years. The calibration and DCA indicated good consistency and clinical utility of the nomogram. Finally, younger (age <60 years) and elderly (age ≥60 years) patients were each divided into two risk groups with significantly different survival rates. Conclusions: A nomogram consisting of five risk factors was developed for forecasting the prognosis of AML with guaranteed reliability.