Abstract
The natural history of suspected diffuse low-grade gliomas (DLGG) depends heavily upon the molecular status. To fully comprehend this integrated information preoperatively, a clinical phenotype incorporating both clinical and radiological information may be of value. We aimed to analyse this systematically in a large multicentre study to identify clinical/radiological phenotypes of DLGG molecular subgroups at the onset. Patients from nine Scandinavian centres, with confirmed World Health Organization (WHO) grade 2 at the time of diagnosis (according to the WHO 2016/2007 classification), known molecular status [isocitrate dehydrogenase (IDH) status and 1p19q codeletion status] and preoperative images of adequate quality, were analysed. MRI-based tumour volume segmentation was used to create a frequency map of their locations in the Montreal Neurological Institute space. A Brain-Grid (BG) system was also used for tumour invasiveness analysis. Variables were analysed for each subgroup of DLGG with regression analyses. A total of 235 patients were included in the study. The three molecular subgroups differed in age, tumour location, epileptic onset and cognitive status. Seizure onset was linked to the number of BG voxels and the A3C2S2 location in all three molecular subgroups. Cognitive deficits were related to increasing age (IDH-mutated oligodendrogliomas), female gender (IDH-wildtype) and tumour volume (oligodendrogliomas). Patients with IDH-mutated astrocytomas (n = 65) displayed younger age, left-sided fronto-insular preferential location, infiltration of anterior ventral inferior fronto-occipital fasciculus (IFOF) and external capsule, and seizure as the onset symptoms. Oligodendrogliomas (n = 116) were more often found in patients >40 years old, with frontal location, dorsal IFOF, frontal aslant tract and superior longitudinal fasciculus invasion, and seizures as the onset symptoms. IDH-wildtype astrocytomas (n = 54) displayed: age >40 years old, left-sided temporo-insular preferential location, invasion of posterior IFOF and cortico-spinal tract, cognitive deficits at onset and the infiltration of posterior left peri-insular voxel (A3C2S3) as a strong predictor of IDH-wildtype. Using an integrated clinico-radiological approach, we identified differences in age, clinical presentation, preferential location and white matter infiltration among specific molecular subgroups of suspected DLGG. The systematic combination of patient-specific variables (age/clinical onset) and tumour-specific features (sub-lobar preferential location) may be relevant to create future prediction models and to better understand the onco-functional trajectory already at the preoperative stage. Prediction models may benefit from combining information rather than, for instance, analysing images only.