Response to therapy in patients with chronic myeloid leukemia (CML) is monitored by both cytogenetic assessment of bone marrow metaphases and quantitative reverse transcriptase polymerase chain reaction (RT-PCR) of peripheral blood samples. Although standardization of BCR-ABL quantification according to the international scale (IS) provides results comparable between different laboratories, current treatment recommendations are still predominantly based on cytogenetic analyses. The aim of this work was to determine a BCR-ABL expression level according to the IS (BCR-ABLIS) which corresponds to complete cytogenetic remission (CCyR).
1,329 paired cytogenetic and molecular assessments of 557 chronic phase CML patients on imatinib-based therapies from the German CML study IV, all during follow-up, have been considered. Median number of samples per individual was 3 (range 1–24). Only molecular evaluations derived from peripheral blood and from patients expressing b2a2 and/or b3a2 BCR-ABL transcripts were considered. Participating laboratories provided results standardized according to the international scale. For cytogenetic analyses at least 20 metaphases, evaluated by conventional chromosome banding analysis, were required. Paired molecular and cytogenetic analyses were performed within an interval of <3 days.
Data was split randomly at the rate of 2:1 into learning and validation sample. 10,000 subsamples that contained no more than one observation per patient were repeatedly drawn randomly from the original learning sample, according to the bootstrap method Using these subsamples, the BCR-ABLIS expression level providing the best discrimination between CCyR and no CCyR was determined choosing the smallest p-value of the exact Fisher test. Only cut-off levels with p-values which were significant after adjustment for multiple testing were considered.
To confirm these results, again 10,000 subsamples were randomly drawn from the original validation sample. For the cut-off points discovered in the learning sample, sensitivity and specificity for predicting a CCyR from the molecular data and the p-values from the Fisher test were calculated within each of these subsamples.
1329 cytogenetic analyses using a median of 25 metaphases (range, 20–30) of which 75% were in CCyR were compared with 1329 molecular data with a median BCR-ABLIS expression level of 0.055% (range, 0–237%). In the learning sample the most frequently found BCR-ABLIS value was 0.35%, but all values between 0.2% and 1.1% were similarly well-suited for separating between patients in CCyR and not in CCyR. Based on this range, two potential cut-off levels were chosen for validation, 0.35% and 1%. While the first one represents a value, where one would rather expect to underestimate the number of CCyRs, the latter represents a value, where one would rather expect to overestimate the number of CCyRs. For the level of 0.35% the p-value was below 0.05 for 67.2% of the 10,000 subsamples, the median p-value was 0.017. With a median value of 88.4%, CCyR cases were correctly identified while cases that were not in CCyR were correctly assessed with a median rate of 71.4%. For the 1% BCR-ABLIS level, the p-value was below 0.05 for 72.0% of the 10,000 subsamples with a median p-value of 0.011. At a median value of 96.0%, CCyR cases were correctly identified while cases that were not in CCyR were correctly assessed with a median rate of 60.0%. When restricting on BCR-ABLIS values between 0.1 and 10.0%, the median concordance rate was 76%.
Compared to molecular analyses, cytogenetic assessment is imprecise and limited with regard to sensitivity. Thus, it is not surprising to find a range of molecular data equivalent to CCyR but not a one-to-one cut-off point. However, for practical reasons it seems desirable to have such a cut-off point. Since lack of CCyR is considered a failure criterion according to the ELN treatment recommendations, it is suggested to keep the proportion of patients small that are considered erroneously not to be in CCyR. Additionally, a considerable amount of patients in major molecular remission or even better is present in the sample, so there might be a slight bias towards lower cut-off values. Thus, BCR-ABLIS expression of <1% is considered to be equivalent to CCyR in chronic phase CML patients treated with imatinib with a median concordance rate of 89%.
Haferlach:MLL Munich Leukemia Laboratory: Equity Ownership. Schnittger:MLL Munich Leukemia Laboratory: Equity Ownership. Hochhaus:Novartis, BMS, MSD, Ariad, Pfizer: Consultancy Other, Honoraria, Research Funding. Müller:Novartis: Honoraria, Membership on an entity's Board of Directors or advisory committees, Research Funding; Bristol-Myers Squibb: Honoraria, Membership on an entity's Board of Directors or advisory committees, Research Funding.
Asterisk with author names denotes non-ASH members.