[A scoring system to predict molecular responses in patients with chronic myeloid leukemia in the chronic phase...

Objective: To develop a scoring system to predict molecular responses in patients with chronic myeloid leukemia in the chronic phase (CML-CP) receiving initial imatinib therapy. Methods: Data from consecutive adults with newly diagnosed CML-CP treated by initial imatinib was interrogated and subjects were distributed randomly into training and validation cohort, in a ratio of 2∶1. Fine-gray models were applied in the training cohort to identify co-variates of predictive value for major molecular response (MMR) and MR4. A predictive system was built using significant co-variates. The predictive system was then tested in the validation cohort and the area under the receiver-operator characteristic curve (AUROC) was used to estimate accuracy of the predictive system. Results: 1 364 CML-CP subjects receiving initial imatinib were included in this study. Subjects were distributed randomly into training cohort (n=909) and validation cohort (n=455) . In the training cohort, the male gender, European Treatment and Outcome Study for CML (EUTOS) Long-Term Survival (ELTS) intermediate-risk, ELTS high-risk, high WBC (≥130×10(9)/L or 120×10(9)/L, MMR or MR4) and low HGB (<110 g/L) at diagnosis were significantly related with poor molecular responses and were given points based on their regression coefficients. For MMR, male gender, ELTS intermediate-risk and low HGB (<110 g/L) were given 1 point; ELTS high-risk and high WBC (≥130×10(9)/L) , 2 points. For MR4, male gender was given 1 point; ELTS intermediate-risk and low HGB (<110 g/L) were given 2 points; high WBC (≥120×10(9)/L) , 3 points; ELTS high-risk, 4 points. We divided all subjects into 3 risk subgroups according to the predictive system above. Cumulative incidence of achieving MMR and MR4 in 3 risk subgroups was significantly different in both training and validation cohort (all P values <0.001) . In the training and validation cohorts, the time-dependent AUROC ranges of MMR and MR4 predictive systems were 0.70-0.84 and 0.64-0.81, respectively. Conclusions: A scoring system combining gender, WBC, HGB level and ELTS risk was built to predict MMR and MR4 in CML-CP patients receiving initial imatinib therapy. This system had good discrimination and accuracy, which could help phsicians optimize the selsction of initial TKI-therapy.