P05.04.A Disconcordance between different molecular methods to assess homozygous deletion of theCDKN2A/B locus in IDH-mutant astrocytomas

Abstract

Abstract Background The fifth edition of the WHO Classification of Tumors of the Central Nervous System (WHO CNS 5), includes molecular parameters for both diagnosis and grading in addition to histological features. For IDH-mutant astrocytoma, homozygous deletion (HD) of CDKN2A/B now results in WHO grade 4, even in the absence of microvascular proliferation or necrosis. CDKN2A/B deletions can be determined by various techniques including shallow and targeted sequencing, and using genome wide DNA-methylation arrays. Various algorithms to call deletions also exist for each platform. Concordance between the various techniques and algorithms is however unknown. Because of the importance to properly call CDKN2A/B deletions, we compared two techniques to call HD in IDH-mutant astrocytoma patients. Methods Samples from 110 IDH-mutant astrocytoma patients enrolled in the GLASS-NL study, and therefore samples from at least two surgical resections per patient, were available. Overall survival (OS) was measured from date of first surgery. Both DNA-methylation data and shallow whole-genome sequencing (sWGS) was collected from 219 samples from 101 patients. For DNA-methylation analysis, HD of CDKN2A/B was defined by <-0.6 log2 intensity combined with visual assessment on samples with log2 intensity between -0.6 and -0.2 (method 1), or by using a hard cutoff of <-0.415 log2 intensity, as determined by Shirahata et al. (method 2). Absolute copy number profiles were estimated by ACE from sWGS data, HD of CDKN2A/B was defined by a loss of at least 1.2 CDKN2A alleles (method 3). Agreement score and Cohen’s kappa (κ) as an index for interrater agreement was calculated. Results Method 1 and method 2 called an HD of CDKN2A/B in 39 and 33 cases, respectively; method 1 called an HD of CDKN2A/B in 9 cases in which method 2 did not detect an alteration. The agreement was strong with 95% agreement and κ = 0.842. Method 3 called an HD of CDKN2A/B in 45 cases of which 12 and 19 were not detected by method 1 and method 2 respectively. Method 3 showed an almost perfect level of agreement with method 1, with 97% agreement and κ = 0.911, and a moderate agreement with method 2, with 89% agreement and κ = 0.671. Survival analysis showed that there was no significant difference in survival when patients were stratified by CDKN2A/B status of the initial sample of all methods. However, when stratified by CDKN2A/B status of the first recurrence, all methods showed a stronger association with OS (method 1: p = 0.001, HR = 2.61[1.44-4.72]; method 2: p = 0.0024, HR = 2.34[1.33-4.11]; method 3: p = 0.0099, HR = 2.34[1.2-4.55]). Conclusion Different methods and different cutoffs to determine the presence of a HD in CDKN2A/B may result in different test outcomes. Quality of the DNA, tumor cell percentage of the sample and ploidy are likely to influence the call of homozygous CDKN2A/B deletion. Ring tests are recommended to improve reliability.