The Validation of Digital PCR–Based Minimal Residual Disease Detection for the Common Mutations in IDH1 and IDH2 Genes in Patients with Acute Myeloid Leukemia

Abstract

Accurate monitoring of minimal residual disease (MRD) is crucial for effective management of patients with acute myeloid leukemia (AML). This study aims to validate MRD detection of the seven most common IDH1 and IDH2 mutations in patients with AML using a QuantStudio 3D digital PCR platform. This assay demonstrated a high concordance for the variant allele frequencies between digital PCR and next-generation sequencing assays. Precision analysis revealed only small variation (<0.5 log10) for all mutations near or at the limit of detection level. This validation also showed a great reproducibility for interrun and intrarun comparisons (28 runs, variation ranges from 0 to 0.48 log10), ensuring comparable results for patient follow-ups. The limit of detection was determined to be 0.1% for all mutations, except the IDH2 R140Q mutation, which was 0.5%. Controls and acceptable ranges were also established for each mutation during validation. This study suggests that the QuantStudio 3D digital PCR assay is a quantitative, sensitive, and reproducible platform for monitoring MRD in patients with AML.

Acute myeloid leukemia (AML) is a heterogeneous hematologic malignancy characterized by the rapid proliferation of abnormal myeloid precursor cells in the bone marrow. AML is associated with a variety of genetic alterations that drive the disease pathogenesis and influence the prognosis and therapeutic strategies.¹ Among these, mutations in the isocitrate dehydrogenase I (IDH1) and isocitrate dehydrogenase 2 (IDH2) genes occur in approximately 15% to 20% of AML cases.^2,3 These mutations are recognized for their role in oncogenesis through the conversion of α-ketoglutarate to an oncometabolite, 2-hydroxyglutarate, which contributes to leukemogenesis via epigenetic dysregulation.^4,5

The accurate detection and quantification of IDH1/IDH2 mutations are critical for the diagnosis, risk stratification, and management of AML.^1,6 Moreover, these mutations have emerged as promising biomarkers for monitoring minimal residual disease (MRD), a key prognostic factor and determinant in the decision-making process for post-remission therapy.⁷ Most of the IDH1/IDH2 mutations were found on the mutation hot spots [ie, IDH1 (R132H, R132C, R132G, R132L, and R132S) and IDH2 (R172K and R140Q)]. Together, these seven mutations account for >95% of IDH1 and IDH2 mutations found in AML.³ The impact of these mutations on prognosis can vary depending on the specific mutation and other concurrent genomic alterations. For example, the IDH1 R132 and IDH2 R172 mutations could correlate with a better outcome after chemotherapy,⁸ whereas the IDH2 R140Q mutation has been proposed to remain as a form of clonal hematopoiesis with uncertain significance during complete remission.^1,9,10

Digital PCR (dPCR) technology has provided a high sensitivity and absolute quantification approach for MRD detection11, 12, 13, 14 compared with conventional quantitative PCR methods. Digital PCR enhances the detection of low-abundance templates by partitioning the individual nucleic acid molecules into thousands of individual microreaction vessels, therefore allowing for an absolute quantification of target nucleic acids without the need for reference standards or calibration curves.¹⁵ Despite its potential, the clinical utility of dPCR for MRD monitoring in AML has been limited by the need for thorough validation of the assay's performance characteristics, including its accuracy, specificity, sensitivity, and limit of detection (LOD).14, 15, 16, 17 Aiming to launch a laboratory-developed test for our clinical services, we validated the QuantStudio 3D digital PCR assay for the detection of IDH1/IDH2 mutations in patients with AML, with an emphasis on its clinical application for MRD monitoring. QuantStudio 3D digital PCR assay accurately and reliably monitors MRD for the common mutations in IDH1 and IDH2 genes.

Materials and Methods

Digital PCR System and Principles

This study used the QuantStudio 3D Digital PCR System (Applied Biosystems, Thermo Fisher Scientific Inc., Waltham, MA). This system uses a high-density, 10-mm² reaction plate containing 20,000 microwells and partitions the PCR solution into discrete reactions (QuantStudio 3D Digital PCR Chip version 2). Nucleic acids are randomly distributed among the microwells, and it is assumed that this conforms to a Poisson distribution for purposes of quantification.¹⁸ For mutation detection, wild-type and mutant sequences are differentiated by sequence-specific probes labeled with different fluorophores. In this study, the authors validated the clinical utility of QuantStudio 3D Digital PCR System for the detection of seven common mutations in IDH1 (R132H, R132C, R132S, R132G, and R132L) and IDH2 (R140Q and R172K) genes.

Reagents, Primers, and Probes

Custom primers and probes were specifically designed for the R132H, R132G, and R132L mutations in the IDH1 gene and the R172K in the IDH2 gene using the Custom TaqMan Assay Design Tool available from Thermo Fisher Scientific. Predesigned off-the-shelf primers/probes for IDH1 (R132C and R132S) and IDH2 R140Q were directly purchased from Thermo Fisher Scientific. Sequences of primers and probes are provided in Table 1. Probes targeting wild-type alleles were labeled with 2′-chloro-7′-phenyl-1,4-dichloro-6-carboxyfluorescein (VIC), and those for mutant alleles were labeled with fluorescein amidite (FAM), except the IDH2 R140Q mutation, for which VIC was used to label the mutant allele and FAM for the wild-type allele. The specificity of primers and probes was evaluated through cross-reaction testing among these mutations. The cross-platform validation was conducted using the variant allele frequency (VAF) obtained from next-generation sequencing (NGS) testing in the authors’ laboratory.

Table 1.

Probes Targeting Wild-Type and Mutant Alleles With Fluorophore Labels Indicated

Mutation	Vendor/catalog no.	Forward primer	Reverse primer	Wild-type probe sequence and fluorophore used	Mutation probe sequence and fluorophore used
IDH1 R132H	Thermo Fisher/4331349 ANFVZFT	5′-CTTGTGAGTGGATGGGTAAAACCTA-3′	5′-CCAACATGACTTACTTGATCCCCATA-3′	5′-CATCATAGGTCGTCATGC-3′-VIC	5′-ATCATAGGTCATCATGC-3’-FAM
IDH1 R132G	Thermo Fisher/4331349 ANZTTTY	5′-CTTGTGAGTGGATGGGTAAAACCTA-3′	5′-CACATTATTGCCAACATGACTTACTTGAT-3′	5′-AAGCATGACGACCTAT-3’ -VIC	5′-AAGCATGACCACCTATG-3′-FAM
IDH1 R132L	Thermo Fisher/4331349 ANNK6XW	5′-CTTGTGAGTGGATGGGTAAAACCTA-3′	5′-CCAACATGACTTACTTGATCCCCATA-3′	5′-CATCATAGGTCGTCATGC-3′-VIC	5′-CATCATAGGTCTTCATGC-3′-FAM
IDH1 R132C	Thermo Fisher/A44177 Hs000000037_rm	N/A∗	N/A	N/A (VIC)	N/A (FAM)
IDH1 R132S	Thermo Fisher/4351379 C_167891676_10	N/A	N/A	N/A (VIC)	N/A (FAM)
IDH2 R140Q	Thermo Fisher/4351379 C_163475618_10	N/A	N/A	N/A (FAM)	N/A (VIC)
IDH2 R172K	Thermo Fisher/4331349 ANH6NKM	5′-GCTGGACCAAGCCCATCA-3′	5′-TCCACCCTGGCCTACCT-3′	5′-ATTGGCAGGCACGCC-3′-VIC	5′-ATTGGCAAGCACGCC-3′-FAM

N/A, not available.

^∗Primers/probes are predesigned by the vendor; sequences are not available to public.

Sample Collection and DNA Extraction

Bone marrow or peripheral blood specimens were collected in EDTA or acid citrate dextrose tubes and processed within 96 hours of collection. Genomic DNA was extracted manually using the Qiagen (Hilden, Germany) DNA Mini Kit or automated using Qiagen EZ-1 or QIAsymphony systems. Extracted DNA was quantified using NanoDrop spectrophotometry (Thermo Fisher Scientific) and diluted to a working concentration of 10 ng/μL.

PCR Setup and Conditions

The reaction for each QuantStudio 3D Digital PCR 20K Chip comprised the QuantStudio 3D Digital PCR Master Mix version 2, TaqMan assays containing primer/probe, and genomic DNA template. A total reaction volume of 14.5 μL, incorporating a 10% excess to account for pipetting losses, was loaded onto each chip (Supplemental Table S1). The ProFlex thermal cycler, equipped with a dedicated tilt base and adaptor kit, was used for the thermal cycling process. Conditions were individually optimized for each mutation (Supplemental Table S2).

Analysis

Following PCR, microchips were imaged and analyzed using the QuantStudio 3D AnalysisSuite Software version 3.1.6-PRC-build18, providing absolute quantification of target molecules without the necessity for standard curves. Thresholds of wild-type and mutant signals were established using the auto setting in the software to eliminate the reviewer's subjectivity. A cross-platform validation of VAF was performed using the data obtained from NGS testing in the authors’ laboratory (the VAF obtained from NGS in the authors’ laboratory is comparable with other laboratories' data based on College of American Pathologists proficiency testing–internal data). The specificity of primers and probes was assessed by cross-reaction experiments. Analytical sensitivity and specificity were determined by testing known positive and negative samples. The LOD for each mutation was determined statistically by the AnalysisSuite Software based on average false positives from wild-type controls and then tested by diluting specimens with known VAF using mutation-negative DNA.

Results

The IDH1/IDH2 Mutation Spectrum

As of August 2020, 751 diagnostic specimens were received in the authors’ laboratory for myeloid NGS panel testing. A total of 60 patients were positive for IDH1 or IDH2 mutations (two patients harbor more than one mutation). The IDH1/IDH2 mutation–positive patients account for approximately 7.9% of the patient population. The IDH2 R140Q was the most frequent mutation, accounting for 63.3% of the IDH-positive patients (Table 2).

Table 2.

Mutational Spectrum of IDH1 and IDH2 Genes in Patients with AML

Gene	cDNA change	Protein change	Patients, N	Frequency, %	dbSNP ID∗
IDH1	c.395G > A	p.R132H	2	3.3	rs121913500
	c.394C > T	p.R132C	8	13.3	rs121913499
	c.394C > A	p.R132S	2	3.3	rs121913499
	c.394C > G	p.R132G	1	1.7	rs121913499
	c.395G > T	p.R132L	3	5.0	rs121913500
IDH2	c.419G > A	p.R140Q	38	63.3	rs121913502
	c.515G > A	p.R172K	4	6.7	rs121913503
Double and triple mutations	c.394C > T/c.395G > T	p.R132C/p.R132L	1	1.7	N/A
	c.394C > T/c.395G > T/c.515G > A	p.R132C/p.R132L/p.R140Q	1	1.7	N/A
Total			60	100	N/A

AML, acute myeloid leukemia; ID, identifier; N/A, not available.

^∗The IDs of single-nucleotide polymorphisms (rs-ID) were obtained from the National Center for Biotechnology Information dbSNP database (https://www.ncbi.nlm.nih.gov/snp, last accessed July 20, 2024). All data related to rs-IDs are publicly released and available.

Evaluation of the Cross-Reactivity Between Different Probes

To evaluate the cross-reactivity (hybridization specificity) between different probes, the signal level of positive samples was tested using probes for all nontargeted mutations. Each probe was confirmed to detect only its corresponding mutation without cross-reactivity (signals are similar to the negative controls; data not shown).

Analytical Sensitivity and Specificity

The analytical sensitivity and specificity of the assay were assessed using 12 positive samples tested by NGS and 83 negative controls [because the primary goal of the study was to test the LOD of this platform, these original positive and negative controls were diluted to different levels of VAF (see subsequent sections)]. The assay correctly identified all positive samples and showed no false positives among the negative controls, resulting in both sensitivity and specificity rates of 100% (negative samples were randomly selected from anonymized laboratory specimens referred for germline FII and FV testing).

Comparison of VAF Obtained by NGS and dPCR

Twelve positive samples harboring IDH1 and IDH2 mutations were tested to evaluate the VAF concordance detected by QuantStudio 3D digital PCR assay and NGS (two samples harbor more than one mutation). The dPCR assay exhibited a high degree of concordance of VAF with that of the NGS data, with a correlation coefficient (R) of 0.98, indicating a strong positive correlation between the two methods (Table 3 and Figure 1).

Table 3.

Comparison of VAFs Detected by NGS and dPCR for IDH1/IDH2 Mutations

Mutation	VAF detected by NGS, %	VAF detected by dPCR, %
R132H	39.2	41.8
R132C	46.0	44.7
R132L	18.9	17.2
R132G	40.8	43.1
R132S	39.4	42.5
R132L/R132C	17.0/15.0	21.6/19.8
R132L/R132C/R14OQ	24.0/4.0/15.0	22.5/5.1/13.3
R140Q	31.6	31.4
R172K	42.3	45.2
WT	N/A	<0.1

dPCR, digital PCR; NGS, next-generation sequencing; VAF, variant allele frequency; WT, wild type.

Figure 1.

Comparison of variant allele frequencies (VAFs) detected by next-generation sequencing (NGS) and digital PCR (dPCR). This figure illustrates the strong correlation between VAFs detected by NGS and dPCR for IDH mutations.

Serial Dilution Testing

For serial dilution testing, the authors diluted all mutations to 5% and 1% (IDH2 R140Q was diluted to 5% and 2.5%) using normal genomic DNA to achieve the desired variant allele frequency (two independent samples for each mutation except the IDH1 R132G). The average VAF obtained by dPCR is 5.85% for 5% dilution (SD, 1.35%; 99% CI, 4.91%–6.78%) and 0.87% for 1% dilution (SD, 0.36%; 99% CI, 0.61%–1.14%). The 2.5% dilution for IDH2 R140Q mutation was tested to be 2.68% (SD, 0.54%; 99% CI, 1.68%–3.67%). The individual VAF in this serial dilution testing is listed in Table 4.

Table 4.

Serial Dilution Analysis of IDH Mutations: Individual VAFs Detected by Digital PCR

Mutations	Sample ID	VAF, %
		1.0%	5.0%
R132H	UKHC2	0.92	5.56
	UKHC1	0.89	4.37
	Average	0.91	4.96
	SD	0.02	0.84
R132C	UKHC3	1.11	5.69
	UKHC4	0.93	4.99
	Average	1.02	5.34
	SD	0.13	0.49
R132S	UKHC5	0.51	4.97
	UKHC6	0.82	4.96
	Average	0.67	4.97
	SD	0.22	0.01
R132G	UKHC7	1.01	5.87
	UKHC7	0.54	5.44
	Average	0.78	5.65
	SD	0.33	0.30
R132L	UKHC8	0.59	8.78
	UKHC9	0.79	8.59
	Average	0.69	8.69
	SD	0.14	0.13
R172K	UKHC10	1.84	6.62
	UKHC12	0.54	4.36
	Average	1.19	5.49
	SD	0.92	1.59

Mutations	Sample ID	VAF, %
		2.5%	5.0%
R140Q	UKHC13	2.29	5.64
	UKHC16	3.06	6.00
	Average	2.68	5.82
	SD	0.54	0.26

Performance for testing of different VAFs
Target VAF	Average	SD	99% CI
5.0%	5.85	1.35	4.91–6.78
2.5%	2.68	0.54	1.68–3.67
1.0%	0.87	0.36	0.61–1.14

ID, identifier; VAF, variant allele frequency.

Establishment of the LOD for Each Mutation Based on Background (Wild-Type) Signal Level

The LOD can be calculated statistically by testing wild-type specimens for each mutation.¹⁹ Fourteen wild-type controls were tested to establish the background signal threshold for each mutation (Table 5). Using the LOD calculation function in the AnalysisSuite Software, the LOD is calculated to be approximately 0.1% for all mutations, except R140Q, which is approximately 0.5%.

Table 5.

Establishment of the Background Signal of IDH1/IDH2 Mutations in 3D Digital PCR Assay

Sample ID	Digital PCR assay test (target/total)
	IDH1, %					IDH2, %
	R132H	R132C	R132S	R132G	R132L	R172K	R140Q
WT1	0.0167	0.0476	0.0208	0.0311	0.0241	0.0209	0.102
WT2	0.0387	0.0462	0.0000	0.0459	0.0149	0.0244	0.166
WT3	0.0323	0.0442	0.0927	0.0143	0.0000	0.0695	0.707
WT4	0.0167	0.0794	0.0419	0.0418	0.0000	0.0317	0.166
WT5	0.0474	0.0795	0.0383	0.0282	0.0258	0.0582	0.214
WT6	0.0125	0.0723	0.0000	0.0355	0.0751	0.0674	0.184
WT7	0.0559	0.0451	0.0122	0.0311	0.0000	0.0608	0.248
WT8	0.0092	0.0567	0.0526	0.0183	0.0634	0.0677	0.380
WT9	0.0278	0.0389	0.0105	0.0253	0.0076	0.0150	0.123
WT10	0.0000	0.0330	0.0000	0.0518	0.0000	0.0533	0.146
WT11	0.0000	N/A	0.0000	0.0109	N/A	N/A	0.248
WT12	0.0178	N/A	0.0164	0.0312	N/A	N/A	N/A
WT13	N/A	N/A	N/A	0.0513	N/A	N/A	N/A
WT14	N/A	N/A	N/A	0.0517	N/A	N/A	N/A
Average	0.0229	0.0543	0.0238	0.0405	0.0154	0.0469	0.2395
SD	0.0178	0.0169	0.0282	0.0231	0.0232	0.0215	0.1744
99% CI	0.0097–0.0361	0.0405–0.0681	0.0028–0.0448	0.0246–0.0564	0.0035–0.0189	0.0294–0.0644	0.104–0.374
LOD	0.102	0.116	0.078	0.086	0.045	0.118	0.521

3D, three dimensional; ID, identifier; LOD, limit of detection; N/A, mutation was not tested for corresponding specimen; WT, wild type.

LOD Validation Using Diluted Specimens and Precision Testing at the LOD Level

To test the robustness of the dPCR platform for the detection of statistically established LOD (section above), the authors diluted VAF to 0.1% (all mutations except IDH2 R140Q) and 0.5% (IDH2 R140Q) for samples harboring corresponding mutations using normal genomic DNA. dPCRs were performed for these samples at the LOD level. In addition to the VAF generated by the software, a minimum of three mutant-only dots by visual inspection are required for the final VAF determination (Figure 2). Specimens were also tested for precision at LOD level by interruns and intraruns (Table 6).²⁰ The variant allele frequencies detected by the dPCR assay were consistent with the expected dilution values, with a 99% CI of 0.090% to 0.134% for the 0.1% LOD samples and 0.473% to 0.777% for the 0.5% LOD samples. This demonstrated the assay's capability for accurate MRD quantification. In addition, interrun reproducibility was tested by running the same samples across different runs, and intrarun reproducibility was tested by running multiple replicates of the same sample within the same run. Only minimal variation was observed (<0.5 log10) in VAF for interruns and intraruns for IDH1/IDH2 mutation samples at the level of LOD, indicating excellent consistency and reliability of the assay (Table 6).

Figure 2.

Scatterplot of the QuantStudio 3D digital PCR data. Data points represent the fluorescence signals obtained from individual microwells of the chip. The x axis (VIC) and y axis (FAM) indicate fluorescence intensities for wild-type and mutant alleles, respectively. [Red dots: VIC-positive wells; blue dots: FAM-positive wells; green dots: wells with both VIC- and FAM-positive signals; yellow dots: wells negative for both VIC and FAM signals (without DNA template)]. Minimum three mutant dots are used as the cutoff for calls close to the limit of detection level.

Table 6.

Assessment of Interrun and Intrarun Reproducibility for IDH Mutation Detection Using Diluted Specimens to the LOD Level [Log Fold Change = Log₁₀(Highest VAF/Lower VAF of Same Sample in Different Runs)]²⁰

Test type	Mutation	Sample ID	Allele frequency, %		Mean, %	SD, %	Maximum log fold change
			1	2
Interrun	R132H	UKHC 1	0.17	0.18	0.175	0.0071	0.02
		UKHC 2	0.06	0.08	0.070	0.0141	0.06
	R132C	UKHC 3	0.13	0.10	0.115	0.0212	0.11
		UKHC 4	0.12	0.14	0.130	0.014	0.07
	R132S	UKHC 5	0.06	0.07	0.065	0.0071	0.07
		UKHC 6	0.06	0.13	0.095	0.0495	0.34
	R132G	UKHC 7	0.11	0.10	0.105	0.0071	0.04
		UKHC 7	0.10	0.13	0.115	0.0212	0.11
	R132L	UKHC 8	0.10	0.09	0.095	0.0071	0.05
		UKHC 9	0.06	0.18	0.125	0.0919	0.48
	R172K	UKHC 10	0.16	0.07	0.115	0.0636	0.36
		UKHC 11	0.12	0.13	0.125	0.0071	0.03
		UKHC 12	0.11	0.04	0.075	0.0495	0.44
	R140Q	UKHC 13	0.53	0.49	0.51	0.0283	0.03
		UKHC 14	0.85	0.93	0.89	0.0566	0.04
Intrarun	R132H	UKHC 1	0.07	0.18	0.125	0.0778	0.41
		UKHC 2	0.08	0.11	0.095	0.0212	0.14
	R132C	UKHC 3	0.09	0.10	0.095	0.0071	0.05
		UKHC 4	0.14	0.09	0.115	0.0354	0.19
	R132S	UKHC 5	0.07	0.07	0.070	0.0000	0.00
		UKHC 6	0.13	0.13	0.130	0.0000	0.00
	R132G	UKHC 7	0.08	0.14	0.110	0.0424	0.24
		UKHC 7	0.11	0.06	0.085	0.0354	0.26
	R132L	UKHC 8	0.10	0.10	0.100	0.1000	0.00
		UKHC 9	0.05	0.06	0.055	0.0071	0.08
	R172K	UKHC 10	0.14	0.18	0.160	0.0283	0.11
		UKHC 15	0.11	0.11	0.100	0.0000	0.00
	R140Q	UKHC 13	0.571	0.486	0.529	0.0601	0.07
		UKHC 16	0.933	0.851	0.892	0.0580	0.04

Performance for LOD testing
Target VAF		Average, %		SD		99% CI
0.5% (for IDH2 R140Q)		0.6250		0.2049		0.473–0.777
0.1% (for other mutations)		0.1136		0.0522		0.090–0.134

ID, identifier; LOD, limit of detection; VAF, variant allele frequency.

Controls and Laboratory Review Process during Clinical Testing

In clinical testing, the authors include controls at 5%, 0.1% (LOD), and wild type for each mutation (0.5% is used for the LOD control of IDH2 R140Q mutation). The authors use the upper limit of the 99% CI obtained from negative specimens for these mutations as the cutoff for acceptable wild-type control value (Table 5). In their validation, the acceptable criteria for wild-type controls are ≤0.068% for all IDH1 mutations and IDH2 R172K mutation (IDH1 R132C has the highest background signal with 99% CI of 0.0405%–0.0681%; therefore, the authors use ≤0.068% as the cutoff). The wild-type criterion is ≤0.374% for IDH2 R140Q (99% CI for this mutation is 0.104%–0.374%).

Similarly, based on 99% CI established during validation, the 0.1% control should fall within 0.090% to 0.134% range, the 0.5% control needs to be between 0.473% and 0.777%, and the 5% controls between 4.91% and 6.78% (Tables 4 and 6). On the basis of these cutoff values, a flowchart describing the clinical review process is depicted in Figure 3.

Figure 3.

Clinical review and decision process. This flowchart depicts the clinical review and decision-making process during digital PCR testing for IDH mutations.

Discussion

Among 60 patients with AML who tested positive for IDH1/IDH2 mutations in our laboratory, IDH2 R140Q was the most frequent (approximately 63.3%) mutation detected. Although this mutation was documented to be the most frequent in other publications, it seems less frequent than our finding (approximately 37%).²¹ Other mutations, including double and triple mutations, were also detected in our cohort, highlighting the necessity of screening all these mutations in clinical testing.

MRD is critical information in guiding treatment decisions, particularly in acute lymphoblastic leukemia, where it was first used to evaluate the early response to treatment.²² The development of drug regimens that achieve complete clinical remissions has sparked an increased interest in MRD evaluation for other hematologic malignancies, such as chronic lymphocytic leukemia, multiple myeloma, and mantle cell lymphoma.^23,24 The advantages of digital PCR platforms make them amenable to MRD monitoring in the clinical arena. Some early studies have shown that dPCR for IDH1/IDH2 mutations in AML is as sensitive, if not more, than NGS.²⁵ Specifically, Grassi et al²⁶ described the lowest VAF detected in their patient's specimen as 0.39%, and the LOD can reach 0.1% in diluted specimens with a digital droplet PCR system.

The significance of MRD could be affected by different clinical scenarios and the specific location of the mutation. A recent study investigated the prognostic impact of IDH mutations in allogeneic transplant recipients.¹⁰ The study showed that, although the diagnostic presence of IDH mutations in AML did not have a significant prognostic impact when consolidated with hematopoietic stem cell transplantation, IDH1 R132 and IDH2 R172 MRD positivity before the stem cell transplant was associated with an increased risk of relapse, whereas IDH2 R140 mutations were not. These findings highlighted that the prognostic impact of IDH can be mutation specific.¹

Currently, there are no established protocols for analyzing and interpreting MRD using dPCR, and its capabilities are still being explored. Nonetheless, the Europe MRD consortium is actively working on standardizing this technique for its eventual integration into routine clinical practice (http://www.euromrd.org, last accessed July 20, 2024).¹² The validation of the QuantStudio 3D digital PCR assay for the detection of IDH1/IDH2 common mutations in the current study represents a significant advancement in the use of digital PCR in clinical MRD monitoring. Our results demonstrate that this assay provides highly accurate and specific detection of the common IDH1/IDH2 mutations that can be potentially used for MRD monitoring in patients with AML with IDH mutations.

We demonstrated a high concordance of VAF detected by dPCR and NGS. This level of agreement is crucial for the clinical adoption of dPCR as a follow-up test, as NGS is often used to measure VAF at the time of diagnosis.^27,28 It also indicates that our assay can be used for dynamic monitoring during disease progression/remission, as it will significantly reduce the cost of follow-up NGS testing, especially when mutations are at an extremely low level. The absence of cross-reactivity among these mutations highlights the robustness of the digital PCR assays.

The MRD level of 0.1% is often used clinically to stratify patients' risk and guide treatment decisions. For instance, the cutoff at 0.1% on multiparameter flow cytometry has been shown to correlate with patient outcomes.^29,30 Although more sensitive methods can achieve lower LODs, the impact of detecting very low MRD is still a matter of investigation. For example, in the setting of patients with chronic phase chronic myeloid leukemia receiving treatment with tyrosine kinase inhibitors, an MRD threshold of 0.1% by RT-PCR is accepted, although PCR could detect much lower levels.³¹ It also needs to be noticed that the value of MRD on clinical management stratification could be gene/mutation specific. In a recent study containing a cohort of 26 IDH mutation-positive patients, the persistence of IDH mutations at the MRD level may not necessarily be associated with an increased risk of relapse or poor survival in adults with AML under certain treatment regimens.¹⁰

Although the assay's performance is robust, this assay can only be applied to patients harboring known common mutations in IDH1 and IDH2 genes, which only accounts for less than approximately 20% of patients with AML. MRD monitoring for patients without IDH1 or IDH2 mutation or rare IDH1/IDH2 mutations is not possible by this assay.

Larger prospective studies are necessary to validate the clinical utility of digital PCR platforms in general in guiding therapeutic decisions and improving patient outcomes. The adoption of this assay in routine clinical practice would benefit from its fast turnaround time and cost-effectiveness. The potential integration of digital PCR and other technologies should be explored to develop a comprehensive MRD assessment strategy.

In summary, the QuantStudio 3D digital PCR assay for IDH1/IDH2 mutations exhibits high accuracy, sensitivity, and specificity, with an LOD conducive to MRD monitoring in patients with AML. It may provide useful information for patient management in AML.

Disclosure Statement

The authors are previous or current employees of University of Kentucky HealthCare genomics laboratory. This laboratory is a fee-for-service diagnostics laboratory using the platform in this study.