Misclassification of Diffuse Gliomas—Letter

We read with interest the article by Iorgulescu and colleagues (1), which queries the nation's largest cancer registry, the National Cancer Database (NCDB), to determine discordance rates between historically classified diffuse gliomas, based solely on histologic characteristics, with their proper molecular diagnosis based on the updated 2016 World Health Organization guidelines. They found that misclassification rates of oligodendrogliomas, astrocytomas, and glioblastomas were approximately 21%–35%, 6%–9%, and 9%, respectively. They validated these findings in a cohort of 450 patients from three tertiary care institutions, yielding similar results. The implication is that results from historical clinical trials and contemporary registry data should be interpreted with caution, as misclassification would have the potential to confound study outcomes.

A limitation of the Iorgulescu study is that 1p/19q status was available for the minority of tumors (45%–47% of oligodendrogliomas, 12%–15% of astrocytomas, and 5% of glioblastomas). This is a potential source of bias, as tumors with more oligodendroglial features may be more likely to be tested for 1p/19q codeletion. In addition, it is unknown whether data from this large U.S. hospital–based registry is generalizable to international cohorts. To determine its applicability to other glioma cohorts, we aggregated cases of primary diffuse gliomas from five high-quality publicly available datasets, spanning approximately 40 international academic institutions: GEO16011 (n = 215; ref. 2), MSK-IMPACT (n = 19; refs. 3, 4), The Cancer Genome Atlas (n = 1,012; refs. 3, 4), UCSF AGS (n = 317; ref. 5), and the Mayo Clinic glioma case–control series (n = 101; ref. 5). 1p/19q Whole-arm codeletion status, determined by FISH (3–5) or DNA microarray (2–5), was available for 1,490 of 1,664 cases, including 93.8% (n = 182) and 91.1% (n = 155) of grade 2 and 3 oligodendrogliomas, 90.8% (n = 79) and 93.2% (n = 164) of grade 2 and 3 astrocytomas, 96.2% (n = 128) and 90.7% (n = 108) of grade 2 and 3 oligoastrocytomas, and 85.9% (n = 674) of glioblastomas, respectively. 1p/19q Codeletion was present in 66.5% and 73.5% of grade 2 and 3 oligodendrogliomas, 6.3% and 3.0% of grade 2 and 3 astrocytomas, 26.6% and 32.4% of grade 2 and 3 oligoastrocytomas, and 2.4% of glioblastomas, respectively.

Here, we expand on the study by Iorgulescu and colleagues, validating their findings in an independent, high-quality international dataset. We found slightly lower rates of 1p/19q codeletion across all histologies (Table 1). The small discrepancy may be due to selection bias in NCDB, and thus, we estimate that misclassification rates would be higher in oligodendrogliomas and lower in astrocytomas and glioblastomas. Still, we believe the authors' conclusions are valid and relevant for contemporary clinical studies and expert consensus guidelines.