P12.55.B RETROSPECTIVE ANALYSIS OF RADIOLOGICAL VERSUS HISTOPATHOLOGICAL DIAGNOSIS OF GLIOBLASTOMA

Abstract BACKGROUND Window-of-opportunity (WOO) and Phase 0 trials, where drugs are administered shortly before surgery to assess pharmacokinetics and biological effects in the tumor without therapeutic intent, are rapidly gaining traction in glioblastoma drug development. A critical matter of debate in the design of these trials is whether biopsy is required for definitive histopathological diagnosis before patient inclusion. This study explores whether MRI-based radiological diagnosis of glioblastoma is sufficiently accurate to justify patient inclusion in WOO/Phase 0 trials without biopsy confirmation. MATERIAL AND METHODS MRI scans (T1w with and without gadolinium, T2w, FLAIR, DWI and PWI) of patients referred to the neurosurgery outpatient clinic at Erasmus MC, Rotterdam, the Netherlands, because of suspected adult-type glioma, were prospectively evaluated by an expert neuroradiologist for the likelihood of glioblastoma. Patients with a history of cancer were excluded if they had multiple or infratentorial enhancing lesions, but not if they had a solitary supratentorial enhancing lesion. The neuroradiologist’s assessment was based on a 5-point Likert scale (1: definitely glioma to 5: definitely not glioma) and an estimation of the glioma’s WHO grade (1-4). The neuroradiologist’s predictions were compared to the final integrated WHO CNS5 tissue diagnosis, as concluded by a neuropathologist, and sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) were calculated. RESULTS A total of 90 patients were included (median age: 62 years, range 21-85 years). 52 patients were radiologically diagnosed with glioblastoma (Likert scale 1-3 and estimation of WHO grade 4). MRI-based imaging demonstrated a sensitivity of 80.4% (95% CI: 67.5-89.0%), specificity of 72.0% (95% CI: 56.2-83.5%), PPV of 78.8% (95% CI: 66.0-87.8%) and NPV of 73.6% (95% CI: 58.0-85.0%). The 11 false positives (21.2% (11/52)) included astrocytoma, IDH-mutant grade 4 (n=4), metastasis (n=5), H3K27M-mutant diffuse midline glioma (n=1) and giant cell glioma (n=1)). Age (OR 1.00, 95% CI 0.96-1.03, p = 0.84) and gender (OR 0.73, 95% CI 0.25-2.01, p = 0.56) were no predictive factors of diagnostic accuracy of radiological imaging. CONCLUSION We consider a PPV of 78.8% for MRI-based diagnosis of glioblastoma sufficiently accurate to justify patient inclusion before definitive tissue diagnosis in WOO/phase 0 studies without the need for biopsy confirmation. False positives could potentially be reduced by incorporating CT scans of the thorax/abdomen. Further research using advanced MRI techniques or nuclear imaging may enhance diagnostic accuracy. We recommend including replacement of false positive patients in the study protocol of future WOO/phase 0 studies.