Comparing the Prognostic Significance of Early Predictors of Survival in Chronic Myeloid Leukemia (CML) Treated with Imatinib - an Analysis of the Randomized CML-Study IV

Introduction: Early prediction of outcome using response-related predictive landmarks has become a major paradigm in the clinical management of chronic myeloid leukemia (CML). Several studies have shown the predictive impact of 10% BCR-ABL^IS at 3 and 6 months for different tyrosine kinase inhibitors. The question, which landmark should define treatment failure and determine treatment intervention has been discussed vividly. However, an objective analysis of quality criteria for different early prognostic landmarks is lacking up to now. Here we compare sensitivity, specificity and the proportion of later disease progressions predicted by 3-month and 6-month landmarks in imatinib-treated patients of the CML-study IV.

Methods: A total of 1,303 newly diagnosed patients were assigned to an imatinib-based treatment arm of CML-Study IV by April 2010. Median follow-up was 7.1 years. The number of molecular assessments was as follows: n=789 (at 6 months), n=692 (at 3 months) and n=301 (at 3 months and at diagnosis, without pretreatment). Gene expression levels were determined by quantitative RT-PCR. At 3 and 6 months, a BCR-ABL ratio was calculated using ABL as reference gene and standardized according to the international scale (BCR-ABL^IS). In addition, at 3 months and at diagnosis a BCR-ABL ratio was calculated using beta-glucuronidase (GUS) as reference gene in order to ensure linearity of measurement at diagnosis. The log reduction at 3 months was calculated from the BCR-ABL ratio at 3 months and at diagnosis. Due to the time-dependent nature of censored survival data, the sensitivity and specificity at eight years were calculated using the method by Heagerty et al. (Biometrics 2000). Overall survival (OS) is defined by the absence of death from any reason, progression-free survival (PFS) is defined as survival in the absence of progression to accelerated or blastic phase. Landmark analyses were performed to compare survival outcomes according to Kaplan-Meier.

Results:Comparing the 10% BCR-ABL^IS landmark at 3 and 6 months, 8-year OS and PFS rates are equal or comparable (table). In contrast, sensitivity and specificity differ substantially with an advantage in favor of sensitivity for the 3-month landmark and in favor of specificity for the 6-month landmark. This difference is paralleled by a smaller proportion of high-risk patients and less progressions identified by the 6-month landmark. From a clinical point of view the 6-month landmark is not only less than half as sensitive, moreover a treatment intervention at 6 months might also prevent less progressions due to the delay of 3 months. The half-log reduction landmark at 3 months is as sensitive as 10% BCR-ABL^IS at the same time. However, it shows improved specificity and defines the smallest proportion of high-risk patients.

Conclusion: The 10% BCR-ABL^IS landmark, which is currently defining treatment failure at 6 months according to European LeukemiaNet (ELN) criteria, fails to detect the majority of patients with later disease progression. Less than a half-log reduction of individual baseline BCR-ABL transcript levels at 3 months on treatment identifies patients with later progressions as sensitive but with higher specificity as compared to 10% BCR-ABL^IS.