Interlaboratory performance testing on EPIC v2.0 CNS tumor profiling demonstrates high reproducibility of tumor classification but reveals the need for harmonized copy number variation reporting.

Abstract

BeadChip array-based DNA methylation profiling has been recognized by the World Health Organization (WHO) as a key diagnostic tool for brain tumor classification. While its diagnostic utility has been well established, data on technical reproducibility, interlaboratory comparability, and data interpretation under real-world diagnostic conditions remain limited. Bridging this gap, we here report the results of an international proficiency test using the Infinium MethylationEPIC v2.0 platform and the corresponding Brain Tumor Classifier version 12.8. Tissue slides of eight FFPE brain tumor samples, covering a representative range of CNS tumor entities, were distributed among 24 laboratories in 10 different countries. Participants were asked to report methylation classes and copy number variation (CNV) profiles. Pre-array workflows were left to local procedures and results had to be submitted within 15 working days. Technical data reproducibility was high with a median pairwise beta-value correlation of 0.99 (range 0.93-1.0). In general, participating centers generated high-quality data, reflected by consistently low detection p-values (<0.01). Eighteen of the 24 participating centers (75%) successfully passed the test. Of the six centers that failed the test, two laboratories experienced technical issues that led to misclassification of individual cases and contributed to incorrect CNV reporting. Four additional centers showed substantial discrepancies in the interpretation of diagnostically highly relevant CNVs, whereas methylation classification was not impaired. While accurate DNA quantification proved to be an important pre-array step, the use of the DNA restoration kit had only minor influence on overall results. Taken together, our interlaboratory performance testing on EPIC v2.0 CNS tumor profiling confirms high reproducibility of tumor classification but reveals the need for harmonized CNV reporting.