DNA-Based Digital PCR for the Quantification of Residual Disease in CML — Sensitivity or Specificity?

Background

We have previously shown that DNA based digital PCR (dPCR) is a more sensitive approach for monitoring patients with CML in deep molecular remission compared to the 3 alternatives: qPCR, RT-qPCR and RT-dPCR. In this study we compared dPCR and RT-qPCR for their ability to predict molecular relapse in the context of national study of de-escalation of TKI dose followed by cessation of therapy (DESTINY).

Methods

The DESTINY study recruited 174 patients in major (or deeper) molecular remission, stratified according to whether their molecular remission status at study entry was MR³ or ≥MR⁴. The standard dose of TKI (imatinib, dasatinib and nilotinib) was halved at study entry and stopped 12 months later provided the patient remained in ≥ MR³. Material from diagnosis, required to clone the patient specific breakpoint for dPCR, was available only in a limited subset of patients. We cloned the breakpoints in 30 patients who reflected the range of outcomes of the study: 4 relapsed within the first 12 months of de-escalation, 12 relapsed 12-24 months from study entry (0-12 months from stopping TKI) and one relapsed 26 months from study entry (14 months from stopping therapy). All 30 patients were in ≥MR⁴ at study entry, of whom 7, 13 and 10 patients were in MR⁴, MR^4.5 and MR⁵, respectively. Relapse was defined as loss of MR³. Fusion specific assays were designed and validated for all patients and dPCR was performed using the RainDrop platform (BioRad^®). Multiple negative samples were used to establish the true positive quantification threshold per assay, including a pooled DNA samples from healthy controls and from other CML patients. Approximately 80,000 cells were included per reaction. RT-qPCR was performed according to standard protocols and all results expressed on the international scale (IS). Samples were collected at a minimum of 3 specified intervals, study entry, time of stopping TKI (month 12), month 24 and/or at the time of relapse. The log rank test was used to predict the risk of relapse and the Spearman correlation coefficient to compare the sensitivity of the two methods.

Results

At trial entry, patients were in MR⁴, MR^4.5 or MR⁵ as defined by RT-qPCR: the depth of remission was not associated with the risk of relapse (log rank test p = 0.18 and 0.07, 0.19, 0.59 for MR⁴ vs MR^4.5, MR⁴ vs MR⁵ and MR^4.5 vs MR⁵, respectively). However, dPCR was positive in only 5 of the 30 patients of whom 3/5 (60%) relapsed compared to 14/25 (56%) who were dPCR negative (p=0.03). At 12 months neither the depth of remission by RT-qPCR nor the detection of malignant cells by dPCR, predicted subsequent relapse.

In order to account for the fact that both methods were detecting different types of target molecules, the results were treated as a binary outcome (target detected/target not detected). 54 samples (49%) were positive by both methods, 6 (5%) were negative by both, 47 (42%) were positive only by RT-qPCR and 4 (4%) were positive only by dPCR. There was no difference in the predictive value of relapse when both methods were positive or negative. The majority of target molecules were not detected by dPCR when RT-qPCR showed transcript numbers ≤ 7 suggesting the possibility of false positivity and that dPCR is more specific in detecting true positive residual disease. Target molecules were detected by dPCR in all samples defined as > MR³ by RT-qPCR: absolute numbers of target molecules were understandably lower by dPCR (detecting cells) than by RT-qPCR (detecting transcripts).

Conclusion

Both methods can detect truly negative samples, however, dPCR is more specific in excluding false positivity. Neither dPCR nor RT-qPCR was capable of predicting relapse at the time of study entry. The ability of dPCR to predict loss of MR³ while off-therapy however requires the monitoring of serial time points leading to relapse before a final conclusion could be made. The question that remains posed here is the clinical validity of early residual disease detection given that, in the case of relapse (defined by ≥MR³ by RT-qPCR), all patients regain molecular remission after the TKI is reintroduced as it has been shown by all the STOP clinical trials.