Predictive values of the HbF for probability of responders
Basing on the results of parameters analysis above, we demonstrated the predictive performance of the HbF at different stage in predicting probability of responders by the ROC curves. The AUC was 0.806for the HbFat the third month of treatmentin predicting probability of major responders at the sixth month treatment (as shown in Figure 3 ). Based on Youden’s index algorithm in the ROC curve, 47.298 g·L-1 was the optimal cut-off value of theHbFat the third month of treatment in predicting major responders at the sixth month treatment, with sensitivity of 67.5%, specificity of 93.3% (as shown in Table 4 ).