Observer reliability in histological grading of astrocytoma stereotactic biopsies

This study provides an objective assessment of the reliability of histological grading of astrocytoma specimens obtained using stereotactic biopsy. Pathological diagnosis of brain tumors provides an index of disease severity and guides clinical practice in their treatment. It also functions as the gold standard in assessing the validity of diagnostic tests such as magnetic resonance imaging. Often diagnoses are made from biopsy material obtained using stereotactic technique. The current study was designed to evaluate this gold standard with regard to interobserver and intraobserver variability. Four certified neuropathologists from academic centers in the United States and Canada were asked to grade 30 brain biopsy specimens obtained stereotactically in patients with astrocytomas. Intraobserver agreement was analyzed in individual observers by comparing their first and second readings, separated by 5 to 14 weeks. Interobserver data were analyzed by comparing initial readings across all observers for individual diagnoses. Kappa analysis was used to measure agreement beyond chance. Intraobserver agreement was 74.73% for glioblastomas multiforme, 51.43% for anaplastic astrocytomas, and 65.22% for low-grade astrocytomas. The most common disagreements were between anaplastic astrocytomas and glioblastomas multiforme, followed by disagreements between anaplastic and low-grade astrocytomas. Interobserver agreement on initial readings was 62.41% (kappa 0.39) for glioblastomas, 36.04% (kappa 0.06) for anaplastic astrocytomas, and 57.14% (kappa 0.48) for low-grade astrocytomas. A significantly greater degree of reliability was seen in histopathological diagnoses of low- or high-grade astrocytomas than in those of intermediate-grade astrocytomas. Therefore, the highest variability occurs at the point of clinical decision making--namely, intermediate-grade tumors that may or may not be selected to receive adjuvant therapy. This considerable variability is an issue that needs to be recognized and further addressed by analysis of current and proposed astrocytoma grading schemes.