Abstract
Introduction
Treatment of chronic myeloid leukemia (CML) with imatinib (IM) has replaced stem cell transplantation as first line therapy also in pediatric patients (pts). Although CML is rare in the first 2 decades of life, pediatric studies published recently have indicated that like in adults most pts achieve prolonged molecular response (MR). In adults it could be demonstrated that IM monotherapy induces a biphasic decline of BCR-ABL1 transcript levels, characterized by an initially steep decline (α-slope) documenting the rapid initial depletion of actively cycling BCR-ABL1–positive cells, followed by a 2nd moderate decline (β-slope), most likely indicating the slow elimination of residual leukemic stem cells (rLSK) with low turnover. In adults regular monitoring of the BCR-ABL1/ABL1 ratio during IM therapy allowed detailed mathematical modelling to predict the abundance of rLSK, which is a major determinant to access the relapse risk after treatment cessation (Horn et al, Blood 2013). A similar assessment of individual relapse risk is even more important in children because long-term IM exposure is associated with longitudinal growth retardation and may increase the risk for cardiac failure at younger age. Therefore, we sought to evaluate and adapte the existing mathematical model as established for adult German pts from the IRIS and CML IV trial to children and teenagers.
Methods
For statistical analysis of IM treatment response in children (trial CML-PAED-II) and adults (German arm IRIS) we required a minimal essential data set of >=7 consecutive BCR-ABL1 level measurements over at least a follow-up interval >1.5 years. Early non-responders within that interval were excluded. Data from 56 pediatric pts (male/female: 34/22; median age 12 years, range 1-18; median follow up 30.7 mos) and 60 adult pts (male/female 43/17; median age 51.5years, range 21-69, median follow up 57.4 mos) were included. For comparison of pediatric and adult cohorts a segmented linear regression model was applied to characterise the pt-specific decline of BCR-ABL1/ABL1-ratios. Five parameters were analyzed in a comparative fashion: i) the α-slope; ii) the β-slope; iii) backwards extrapolation of the α-slope (corresponding to an estimation of the BCR-ABL1 level at diagnosis); and at the breakpoint of α- and β-curves: iv) the transcript ratio as well as, v) the time point (months on IM treatment). For the corresponding modelling approach, we use an established single cell-based model of CML dynamics under IM treatment to estimate the fraction of rLSK (Roeder et al, Nat Med 2006).
Results
In 42/56 pediatric and in 54/60 adult pts a biphasic slope of transcript decline patterns was identified. A comparison of median response curves documented a slightly higher, however, not significant reduction of transcript ratios in the pediatric cohort with a tendency to achieve lower ratios at the breakpoint of the biphasic decline (mean BCR-ABL1 ratio adults vs pediatric: 0.103 %/ 0.051 %). This effect is attributed to a significantly increased heterogeneity in the initial IM response (SD of mean ratio adults versus pediatric: 0.63/1.2 log scales). For the α- and β-slopes no statistically significant difference was observed between the cohorts. We used the available data to adapt our mathematical model for pediatric CML, in which the slightly increased reduction in BCR-ABL1 levels in children can be attributed to an increased sensitivity to IM or an increased stem cell turnover. However, the increased heterogeneity in the initial IM response in children limits the model’s ability to provide refined risk estimators for treatment cessation and calls for the integration of further parameters.
Conclusions
Statistical analysis revealed distinct features of pediatric CML with respect to the dynamic response to IM treatment. In particular, we observed an increased heterogeneity in the IM response pointing towards a subgroup of pediatric pts that achieves MR4 during the initial decline (11 out of 42 children versus 2 out of 54 adults). We will further adapt our mathematical modelling approach for pediatric CML to estimate whether relapse-free treatment cessation is achieved more likely in these pts. Thereby, the model-based analysis and predictions on CML dynamics demonstrated to become useful for optimization of treatment strategies also in pediatric pts.
Disclosures:
Hochhaus: Novartis: Consultancy, Honoraria, Research Funding, Travel Other; BMS: Consultancy, Honoraria, Research Funding; Pfizer: Consultancy, Honoraria; Ariad: Consultancy, Honoraria.
A Branching Process to Characterize the Dynamics of Stem Cell Differentiation
