Digital PCR assay detects trace leukemia residual disease below standard test limits

Researchers in South Korea have validated a next-generation digital polymerase chain reaction assay that quantifies the BCR::ABL1 fusion transcript in chronic myeloid leukemia down to a sensitivity of 0.0025% on the International Scale, equivalent to a molecular response level of MR4.6. The validation study, published in Blood Research, reports that the LQ Check BCR::ABL1 Major %IS Kit, run on the LQ Suite dPCR system, met international analytical standards and closely matched the conventional real-time quantitative reverse-transcription polymerase chain reaction (RT-qPCR) method that laboratories currently use to monitor residual disease. The work addresses a well-recognized gap: once patients with chronic myeloid leukemia reach deep molecular responses on tyrosine kinase inhibitor therapy, the standard RT-qPCR approach loses reliability precisely in the range where clinicians must decide whether treatment can be safely stopped.

Chronic myeloid leukemia is driven by the BCR::ABL1 fusion gene, which arises from the reciprocal translocation t(9;22)(q34;q11) and is present in more than 90% of patients. The rearrangement produces three principal fusion transcript classes depending on breakpoint location: Major (M-BCR), Minor (m-BCR), and Micro (μ-BCR). Tyrosine kinase inhibitors have dramatically improved prognosis, and achieving a deep molecular response has enabled treatment-free remission in selected patients. Sensitive and accurate quantification of BCR::ABL1 transcripts, which serve as key markers of measurable residual disease, is therefore central to decisions about treatment modification and potential discontinuation.

Molecular responses are reported on the International Scale, with standard thresholds of MR1 (10%IS), MR2 (1%IS), MR3 (0.1%IS), MR4 (0.01%IS), and MR4.5 (0.0032%IS). In real-world practice, reliable measurements in the deep molecular response range are considered essential for safe treatment-free remission attempts. However, RT-qPCR relies on standard curves and is subject to inter-laboratory and inter-operator variability, which is particularly high at or beyond MR4.5 and can compromise measurement reliability and increase the risk of false-negative results. Digital PCR offers an alternative by partitioning each reaction into numerous microdroplets and estimating target copy numbers using Poisson statistics, allowing absolute quantification that is robust to variations in PCR efficiency. Digital PCR can detect low levels of BCR::ABL1 transcripts in the MR4.5 to MR5 range, and its results can be converted to the International Scale through calibration against World Health Organization international standards.

The new assay simultaneously quantifies the two major chronic myeloid leukemia transcript variants, e13a2 and e14a2, along with the BCR reference gene, in a single reaction. The validation team, led by Yong-Hwi Kim and Jae-Hoon Kim with colleagues at multiple Korean institutions, evaluated sensitivity, accuracy, precision, and specificity according to Clinical and Laboratory Standards Institute guidelines EP17-A2, EP06, EP05-A3, and EP07, using BCR::ABL1-positive and -negative cell line materials and the WHO 1st International Standard. For the limit of blank, 60 replicates of RNA from the BCR::ABL1-negative HL-60 cell line were tested across two kit lots; a nonparametric 95th-percentile analysis yielded a blank limit of one positive droplet, so the positivity cutoff was set conservatively at two or more positive droplets, roughly 0.14 copies per microliter, to minimize false positives that could incorrectly influence treatment decisions.

The limit of detection was established using serial dilution panels spanning 0.0032%IS (MR4.5) to 0.0010%IS (MR5.0), with K-562 and KCL-22 cells spiked into HL-60 RNA, generating 576 measurements across two lots. Probit regression and a nonparametric criterion of at least 95% detection both converged on 0.0025%IS, where both transcript types achieved a 95.8% hit rate, with transcript-specific detection limits of MR4.5990 for e13a2 and MR4.5770 for e14a2. The limit of quantitation, defined as the lowest level meeting both 95% detection and a coefficient of variation of 76% or less, was also 0.0025%IS; at that level, coefficients of variation ranged from 37.83% to 52.06% across transcripts and lots. The authors note that this 76% threshold was adopted from a precedent involving an FDA-cleared digital PCR assay rather than being a universal CLSI-recommended cutoff, so the resulting quantitation limit should not be generalized to other BCR::ABL1 digital PCR assays.

Linearity testing per CLSI EP06 covered approximately five orders of magnitude, from about 10%IS down to 0.0006%IS, and produced coefficients of determination between 0.9916 and 0.9972 across both transcripts and two production lots. Analytical precision, assessed by spiking K-562 into HL-60 at 1%IS, 0.1%IS, and 0.01%IS, showed within-run coefficients of variation of 1.52% or less, and reproducibility coefficients of variation between 0.20% and 2.76% across lots, sites, instruments, and operators, with particularly low between-instrument and between-operator variation of 0.20% and 0.31%. Interference testing per CLSI EP07 found no significant effects from cholesterol, bilirubin, hemoglobin, triglycerides, heparin, guanidinium-containing lysis buffer, EDTA, ethanol, or genomic DNA on positive or negative samples, and cross-reactivity testing of hematological cancer cell lines, normal human blood, and related gene RNA showed no nonspecific BCR::ABL1 detection. Accuracy testing against the WHO 1st International Standard (NIBSC code 09/138) yielded coefficients of determination of 0.9941 to 0.9983 across three lots, confirming WHO traceability that allows laboratories to report results directly on the International Scale.

After seven specimens failed RNA quality control, 72 remained for categorical comparison between the digital PCR assay and RT-qPCR, and 67 with quantifiable results from both methods were used for paired statistical analyses. Passing-Bablok regression of log10-transformed International Scale values gave a slope of 1.0189 (95% confidence interval 0.9643 to 1.0679), which included one and therefore showed no significant proportional bias, and a Pearson correlation coefficient of 0.9740. The intercept of −0.2552 did not include zero, indicating a small constant systematic difference, but it remained well within the predefined acceptable clinical limit of ±0.5 log10. Bland-Altman analysis showed a mean bias of 0.2266 (RT-qPCR minus digital PCR), with most observations falling within the 95% limits of agreement.

In the categorical comparison of 72 specimens, agreement between the two methods was moderate, with a Cohen’s kappa of 0.54 and 63.9% (46 of 72) concordance. Among the 26 discordant cases, 20 were classified by digital PCR as having a deeper molecular response than by RT-qPCR, one was classified as deeper by RT-qPCR, and five showed low-level positivity by RT-qPCR (MR3.92 to MR4.47) but negative digital PCR results under the predefined reporting criterion. The authors suggest that differences in analytical sensitivity, stochastic variation in target molecule sampling at very low concentrations followed by Poisson-based partitioning, and degradation or freeze-thaw effects in the retrospectively archived RNA may all have contributed to these discordances. Inter-laboratory reproducibility, tested with 10 specimens spanning MR 0.04 to MR 3.93 across three sites, instruments, and operators, showed a correlation of 0.99 and a mean bias of 0.0967, with all data points within the 95% limits of agreement.

The study also defined practical specimen-handling requirements. Whole blood from 15 adult patients was stored at 2–8 °C and tested at 24, 48, 72, and 96 hours, with 10 replicates per patient at each time point. BCR reference gene copy numbers declined significantly between 24 and 72 hours, and at 96 hours only a minority of samples met the assay validity threshold of at least 2,000 copies per microliter. RNA extraction within 48 hours is therefore preferred for routine testing, with 72 hours an exceptional upper limit and reporting contingent on the BCR validity threshold being met. The authors caution that the study was limited by its pilot scale of 72 specimens and the retrospective nature of the archived cohort, and that the clinical significance of the deeper molecular response classifications produced by digital PCR, particularly for treatment-free remission decisions, requires longitudinal and prospective validation before it can guide practice.

Subject of Research: Analytical and clinical validation of a digital PCR assay for BCR::ABL1 measurable residual disease detection in chronic myeloid leukemia

Article Title: Next-generation digital polymerase chain reaction assay for BCR::ABL1 in chronic myeloid leukemia: comprehensive validation for standardized measurable residual disease detection

Article References: Kim, Y.-H., Jang, M.-S., Kim, S.-H., Roh, I.-Y., Kim, B.-S., Kee, K.-M., Kim, S.-H., Kim, D.-W., & Kim, J.-H. (2026). Next-generation digital polymerase chain reaction assay for BCR::ABL1 in chronic myeloid leukemia: comprehensive validation for standardized measurable residual disease detection. Blood Research, 61(1), Article 67. https://doi.org/10.1007/s44313-026-00188-5

Image Credits: AI Generated

DOI: 10.1007/s44313-026-00188-5

Keywords: chronic myeloid leukemia, digital PCR, BCR::ABL1, measurable residual disease, RT-qPCR, molecular response, limit of detection, treatment-free remission, WHO international standard, assay validation, International Scale, deep molecular response