Quantification of BCR-ABL with Digital PCR Results in a Significantly Lower Rate of Deep Molecular Responses When Compared to RT-qPCR in CML Patients Treated in the ENEST1st Trial

Digital PCR (dPCR) generates an absolute read out that is largely tolerant to variations in PCR efficiency, reducing the requirement for standardisation like the conversion of the BCR-ABL/ABL ratio to international scale (IS). The aim of this study was to compare the results of dPCR to qPCR in blinded samples from two independent laboratories with respect to the observed rates of molecular response (MR) in CML patients (pts) having undergone 18 months of nilotinib treatment in the ENEST1st trial.

A total of 230 cDNA samples from CML pts treated within the ENEST1st trial with e13 or e14/a2 BCR-ABL fusion genes were analysed in Leipzig (L, n=75) or Mannheim (M, n=155) with qPCR between 2012 and 2013. BCR-ABL levels were determined relative to those of ABL and standardization was achieved using plasmid DNA. Both labs are accredited by the European Treatment Outcome Study (EUTOS) collaboration. The cDNA samples were blinded for the qPCR results and re-analysed in L with a duplex dPCR using a QX200 Droplet Digital PCR System (BIO-RAD). In line with the manufacturer's recommendations, samples yielding a minimum of 3 positive droplets in duplicates from the 12-19.000 routinely analysed were scored as positive (+). Depth of MR was scored using the EUTOS definitions used in the ENEST1st trial.

For the whole cohort, the median copy number (CN) of BCR-ABL and ABL was 12 and 59350 by dPCR and 10 and 53537 by qPCR, respectively. Both methods detected similar numbers of BCR- ABL+ samples (dPCR 186, qPCR 189) with a median % BCR-ABL 0.022 by dPCR compared 0.013 by qPCR after conversion to IS. 90% of the BCR-ABL+ samples with dPCR were within an deviation of 4.06 -fold (median 1.22 fold) from qPCR for BCR-ABL, 1.77 fold (median 1.06 fold) for ABL and 6.43 fold (median 1.72 fold) for %BCR-ABL^IS.

Samples from L showed median CN with dPCR and qPCR for BCR-ABL (9 and 10) and ABL (29670 and 30734) with a correlation R² = 0.95 and 0.84. The median % BCR-ABL was 0.02% by dPCR and 0.03% before and 0.01% after conversion to IS with qPCR. 90% of the BCR-ABL+ samples by dPCR were within a range of 2.9 -fold deviation (median 0.66 fold) from qPCR for BCR-ABL, 1.9 fold (median 0.91 fold) for ABL, 2.6 fold (median 0.74 fold) for %BCR-ABL and 7.5 fold (median 1.96 fold) for %BCR-ABL^IS.

Samples from M had higher median CN for BCR-ABL and ABL (16 and 80000) by dPCR compared to qPCR (10 and 66570). Correlation was better for BCR-ABL compared to ABL with R²=0.95 and R² = 0.74. The median % BCR-ABL was 0.022 by dPCR compared to 0.017 and 0.015 with qPCR after conversion to IS. 90% of the BCR-ABL positive samples with dPCR were within a deviation of 4.7 -fold (median 1.6 fold) from qPCR for BCR-ABL, 1.7 fold (median 1.1 fold) for ABL, 4.9 fold (median 1.5 fold) for %BCR-ABL and 5.6 fold (median 1.4 fold) for %BCR-ABL^IS.

The cumulative rates of MR3, MR4 and MR4.5 or better @ 18 months of treatment in the ENEST1st trial were 83, 43 and 29% with qPCR. The distribution in MR classes was significantly different between dPCR and qPCR (p<0.001). MR scoring by dPCR resulted in decreased cumulative rates of MR3 (-6%), MR4 (-10%) and MR4.5 or better (-11%). Significantly fewer pts achieved >=MR4 when analysed by dPCR compared to qPCR (n=77 vs. 100, p<0.05). Of the 91, 33, 44 and 23 samples scored MR3, MR4, MR4.5 or MR5 with qPCR, 71 (37%) were one (n=49), two (n=21) or three (n=1) MR classes higher with dPCR. In contrast, only 13% were scored one (n=20) or more (n=5) MR classes deeper, most likely as a result of the cut-off of 3 positive droplets. 51% of the samples were concordant (Table 1). Comparison of samples for which MR was deeper by dPCR against those for which MR was worse by dPCR than by qPCR showed no difference in ABL levels (p=0.6).

Conclusions: dPCR tends to read out higher levels of BCR-ABL/ABL than standard qPCR, resulting in the placement of pts in worse MR classes. This effect does not appear to be associated with the amount or quality of material and was observed in two independent pt cohorts. Therefore, dPCR should not be used without careful evaluation and comparison to RT-qPCR.