Autor:innen:
P. Ahmad-Nejad (Wuppertal, DE)
S. Neumann (Bonn, DE)
W. Geilenkeuser (Bonn, DE)
L. Mirbach (Mannheim, DE)
A. Duda (Mannheim, DE)
V. Haselmann (Mannheim, DE)
BACKGROUND: Circulating tumor DNA (ctDNA) analysis, commonly referred to as liquid biopsy, has the potential to revolutionize oncologic diagnostics by helping to stratify targeted therapies, serving as a personalized tumor marker for treatment monitoring, and ultimately facilitating the detection of emerging resistance mechanisms. However, several obstacles hamper successful translation into standard care, with the lack of harmonized preanalytical and analytical issues being the most significant concern. This external quality assessment (EQA) scheme aimed to address issues of analytical quality as prerequisite for an urgently needed standardization of laboratory workflows.
METHODS: In each EQA schemes offered, three samples containing 2-3 mL EDTA-plasma spiked with fragmented genomic DNA isolated from tumor cell lines were provided for analysis of sequence variations in KRAS p.G12/p.G13, BRAF p.V600E and EGFR p.T790M. The variant allele frequency (VAF) ranged from 0% to 10%. Laboratories were asked to use their routine procedures and report following: 1) time elapsed for processing of samples, 2) storage temperatures, 3) method used for extraction and quantification, 4) genotyping methods and results.
RESULTS: A total of 184 laboratories from 13 European countries participated in these EQAs. Neither the median shipment time to participants of 2.8 days nor the total time elapsed before analysis of 7 days affected the overall diagnostic performance. 73.5% reported to isolate cfDNA manually, and the most commonly used isolation kit was the QIAamp circulating nucleic acid kit (52.9%). On average, 2.6mL plasma were used for cfDNA isolation, with 83.3% of laboratories using buffer for cfDNA elution. Isolated cfDNA was quantified by Qubit in 57.8% of cases, followed by 9.8% using ddPCR and 9.3% using NanoDrop. Importantly, cfDNA equivalent to 640µL plasma was used on average for ctDNA analysis, representing only 21.3% of the maximum possible input. For analysis, ddPCR (45.2%) was used most frequently, followed by qPCR (16.5%), and MassArray (12.1%). Overall, 95/1152 genotypes were determined inaccurately, resulting in an overall error rate of 8.3%. The false-positive rate was 4.2% and increased with time. The false-negative rate was 8.9%, with the error rate increasing with decreasing VAF, e.g., exceeding 30% at a VAF of 0.1%. Noteworthy, the error rate varied significantly depending on the method used, with the lowest error rate observed for ddPCR and BEAMing.
CONCLUSION: This EQA schemes illustrate the current variability in multiple stages of cfDNA processing and analysis of ctDNA, resulting in an overall error rate of 8.3%. Importantly, the error rate clearly depends on the method used and the VAF of the target gene. Of note, the false-negative rate has improved over time, suggesting that laboratories are using more sensitive techniques, while the false-positive rate is increasing, indicating problems in accurately determining the detection limit.