Patient-specific molecular response dynamics can predict the possibility of relapse during the second treatment-free remission attempt in chronic myelogenous leukemia

Abstract

In chronic myelogenous leukemia (CML), treatment-free remission (TFR) is defined as maintaining a major molecular response (MMR) without a tyrosine kinase inhibitor (TKI), such as imatinib (IM). Several studies have investigated the safety of the first TFR (TFR1) attempt and suggested recommendation guidelines for such an attempt. However, the plausibility and predictive factors for a second TFR (TFR2) have yet to be reported.The present study included 21 patients in chronic myeloid leukemia who participated in twice repeated treatment stop attempts. We develop a mathematical model to analyze and explain the outcomes of TFR2. Our mathematical model framework can explain patient-specific molecular response dynamics. Fitting the model to longitudinal BCR-ABL1 transcripts from the patients generated patient-specific parameters. Binary tree decision analyses of the model parameters suggested a model based predictive binary classification factor that separated patients into low- and high-risk groups of TFR2 attempts with an overall accuracy of 76.2% (sensitivity of 81.1% and specificity of 69.9%). The low-risk group maintained a median TFR2 of 28.2 months, while the high-risk group relapsed at a median time of 3.25 months. Further, our model predicted a patient-specific optimal IM treatment duration before the second IM stop that could achieve the desired TFR2 (e.g., 5 years).

Keywords

• Chronic myelogenous leukemia
• Treatment-free remission
• Model-based predictive factors