Article
Challenge and clinical relevance of a non-matching classifier output in genome-wide DNA methylation analysis for CNS neoplasms
Klinische Herausforderung eines nicht übereinstimmenden Classifier-Ergebnisses bei der DNA-Methylierungsanalyse für ZNS-Neoplasien
Search Medline for
Authors
-
Richard Drexler
- Universitätsklinikum Hamburg-Eppendorf, Klinik und Poliklinik für Neurochirurgie, Hamburg, Deutschland
-
Manfred Westphal
- Universitätsklinikum Hamburg-Eppendorf, Klinik und Poliklinik für Neurochirurgie, Hamburg, Deutschland
-
Ulrich Schüller
- Universitätsklinikum Hamburg-Eppendorf, Institut für Neuropathologie, Hamburg, Deutschland
-
Malte Mohme
- Universitätsklinikum Hamburg-Eppendorf, Klinik und Poliklinik für Neurochirurgie, Hamburg, Deutschland
| Published: | May 25, 2022 |
|---|
Outline
Text
Objective: The molecular classification of CNS tumors has revolutionized our understanding of the biological heterogeneity and diversity of tumor subtypes. DNA methylation-based classification of CNS tumors allows to discriminate subtypes using a chip-based methylation array. Although this standardized approach can diagnose tumors with unmatched specificity, there are still tumors which cannot be classified accordingly, thus complicating the definition of the therapeutic regimen. The aim of our study was to gain further insight into these challenging cases.
Methods: 81 patients, who underwent surgery for a tumor of the central nervous system (CNS) were unclassifiable, or had a low methylation score (<0.9) using the DNA methylation-based CNS tumor classifier, were included. Genomic DNA was analyzed for genome-wide DNA methylation patterns using the Illumina EPIC (850k) array.
Results: Overall, 47 patients had a different output using the classifier when compared to their histological diagnosis. Of these, 41 patients (87.2%) did not have any diagnosis from the methylation classifier (“no matching methylation class”). Surgical and clinicopathological features were balanced in our defined groups and had no impact on the confidence score. Cases with non-classifiable tumors had a significantly longer time until a decision for adjuvant therapy and these cases were presented more often in tumor boards (p<0.01). Further analyses in 23 glioblastoma patients revealed comparable results for the OS, but a significantly shorter PS in cases with a low confidence score. The application of the newest classifier version enabled the classification in 67.9% and resulted in re-classification of the diagnosis in 46.9% of cases.
Conclusion: Our study presents unclassifiable cases and the clinical impact on tumor board decision making and time to initiation of treatment. Even though DNA methylation profiling significantly contributes to advanced CNS tumor diagnostics, clinicians should be aware of a prolonged interval to treatment initiation, especially for highly malignant brain tumors. We would recommend to schedule adjuvant treatment as early as possible if surgical and histological results are suspicious of malignant/anaplastic disease and adjuvant options are limited in choice anyway. Furthermore, recurring tumors in which a previous DNA methylation analysis with a non-matching output score is available, should be re-analyzed using the newest classifier version to update the diagnosis for treatment decisions.
