Predictive scoring systems for molecular responses in persons with chronic phase chronic myeloid leukemia receiving initial imatinib therapy.

Abstract

It is vital for physicians and persons with chronic myeloid leukemia (CML) to accurately predict the likelihood of achieving a major molecular response (MMR) and a deep molecular response (DMR; at least MR4) at the start of imatinib-therapy, which could help in decision making of treatment goals and strategies. To answer this question, we interrogated data from 1369 consecutive subjects with chronic phase CML receiving initial imatinib-therapy to identify predictive co-variates. Subjects were randomly-assigned to training (n = 913) and validation (n = 456) datasets. Male sex, higher WBC concentration, lower haemoglobin concentration, higher percentage blood blasts and larger spleen size were significantly-associated with lower cumulative incidences of MMR and MR4 in training dataset. Using Fine-Gray model, we developed the predictive scoring systems for MMR and MR4 which classified subjects into the low-, intermediate- and high-risk cohorts with significantly-different cumulative incidences of MMR and MR4 with good predictive discrimination and accuracy in training and validation cohorts with high area under the receiver-operator characteristic curve (AUROC) values. These data may help physicians decide appropriateness of initial imatinib therapy.