Determining the rise in BCR-ABL RNA that optimally predicts a kinase domain mutation in patients with chronic myeloid leukemia on imatinib

Abstract

In imatinib-treated chronic myeloid leukemia (CML), secondary drug resistance is often caused by mutations in the BCR-ABL kinase domain (KD). As alternative therapies are available for imatinib resistance, early identification of mutations may prevent disease progression. Because most patients are routinely monitored by BCR-ABL quantitative polymerase chain reaction (PCR), it is important to define the optimal increase in BCR-ABL that should trigger mutation testing. Expert panels have provisionally recommended a 10-fold BCR-ABL increase as the trigger for mutation screening, acknowledging the lack of consensus. To address this question, we monitored 150 CML patients by quantitative PCR and DNA sequencing. Thirty-five different mutations were identified in 53 patients, and, during 22.5 months (median) of follow-up after sequencing, mutations were significantly predictive of shorter progression-free survival. An unbiased receiver operating characteristic analysis identified a 2.6-fold increase in BCR-ABL RNA as the optimal cutoff for predicting a concomitant KD mutation, with a sensitivity of 77% (94% if including subsequent samples). The 2.6-fold threshold approximated the analytic precision limit of our PCR assay. In contrast, transcript rise cutoffs of 5-fold or greater had poor diagnostic sensitivity and no significant association with mutations. We conclude that the currently recommended 10-fold threshold to trigger mutation screening is insensitive and not universally applicable.

Figure 1

ROC curve for optimally predicting a kinase domain mutation by a rise in BCR-ABL RNA. The quantitative increase in BCR-ABL RNA levels was determined on 233 samples (from 132 patients) with a readable kinase domain (KD) DNA sequence, and a numeric BCR-ABL RNA level on both the sequenced sample and the immediately prior sample. Sensitivity was defined as the number of mutation-bearing samples with a transcript level rise above a moving (fold-change) cutoff threshold divided by the total number of samples with a mutation. Specificity was defined as the number of wild-type samples with a transcript level rise below the same cutoff threshold divided by the total number of samples without a mutation. The Youden index (J) is the vertical distance from each point on the receiver operating characteristic (ROC) curve to the diagonal “chance” line (from 0,0 to 1,1). The maximal J value (J_max, vertical dotted line), defining the optimal cutoff threshold (2.6-fold transcript level rise) for predicting a concomitant mutation, is denoted, as are the ROC points associated with a 2-, 3-, 5- and 10-fold transcript level rise.

Figure 2

The presence of a kinase domain mutation predicts shorter progression-free survival after sequencing. The cumulative rate of progression-free survival (Kaplan-Meier method) is shown for patients with, versus patients without, a detectable KD mutation.

Figure 3

The presence of a kinase domain mutation predicts shorter progression-free survival after imatinib initiation. The cumulative rate of progression-free survival (Kaplan-Meier method) is shown for patients with versus patients without, a detectable KD mutation, detected by 30 months after initiation of imatinib therapy. This 30-month landmark analysis excludes patients with sequencing performed after 30 months, and those with disease progression or last follow-up occurring before 30 months.

Figure 4

Time to disease progression after mutation discovery. The time from the initial discovery of a KD mutation to subsequent disease progression (y-axis) is plotted against the location of the mutation (kinase domain amino acid number; x-axis). The 37 data points reflect the presence of multiple mutations in some of the 29 patients with disease progression after a mutation was discovered. Shaded bars depict average values for each mutation location. Underlined codons represent the mutation location of 5 patients with no subsequent disease progression during follow-up. For these 5 patients, the follow-up time from mutation discovery to the last monitoring visit is indicated in parentheses.