A predictive scoring system for therapy-failure in persons with chronic myeloid leukemia receiving initial imatinib therapy.

Abstract

Data from 1,364 consecutive subjects with chronic-phase chronic myeloid leukemia (CML) receiving initial imatinib-therapy were interrogated to identify co-variates predicting therapy failure. Subjects were randomly divided into training (n = 908) and validation datasets (n = 456). In the training dataset, WBC count ≥120 × 10E + 9/L, haemoglobin concentration <115 g/L, blood basophils ≥12% and European Treatment and Outcome Study for CML Long-Term Survival (ELTS) risk score were significantly-associated with failure-free survival (FFS). Each co-variate was assigned 1 point to develop the imatinib-therapy failure (IMTF) model except ELTS high-risk category which was assigned 2 points based on multi-variable regression coefficients. Area under receiver-operator characteristic curve values in the IMTF model for 1-, 3- and 5-year FFS were 0.79-0.84 in the training dataset and 0.78-0.85 in the validation dataset. Calibration plots showed high agreement between predicted and observed outcomes. Decision curve analyses indicated subjects benefited from clinical use of this model. Cumulative incidences of imatinib-therapy failure and probabilities of FFS among the 5 risk cohorts (very low-, low-, intermediate-, high- and very high-risk) using the IMTF model were significantly different (all p values < 0.001). The IMTF model also correlated with probabilities of progression-free survival and survival (all p values < 0.001). These data should help physicians optimize TKI-therapy strategy at diagnosis in persons with chronic phase CML.