High Interobserver Agreement in the Subjective Classification of 5‐Aminolevulinic Acid Fluorescence Levels in Newly Diagnosed Glioblastomas

BACKGROUND AND OBJECTIVES Fluorescence-guided resection of glioblastomas (GBM) using 5-aminolevulinic acid (5-ALA) improves intraoperative tumor visualization and is thus widely used nowadays. During resection, different fluorescence levels can usually be distinguished within the same tumor. Recently, we demonstrated that strong, vague, and no fluorescence correspond to distinct histopathological characteristics in newly diagnosed GBM. However, the qualitative fluorescence classification by the neurosurgeon is subjective and currently no comprehensive data on interobserver variability is available. The aim of this study was thus to investigate the interobserver variability in the classification of 5-ALA fluorescence levels in newly diagnosed GBM. STUDY DESIGN/MATERIALS AND METHODS A questionnaire investigating the interobserver variability in 5-ALA fluorescence quantification was performed at a nation-wide neurosurgical oncology meeting. The participants involved in the neurosurgical/neurooncological field were asked to categorize 30 cases of 5-ALA fluorescence images derived from GBM resection on a lecture hall screen according to the widely used three-tier fluorescence classification scheme (negative, vague, or strong fluorescence). Additionally, participants were asked for information on their medical background such as specialty, level of training, and experience with 5-ALA fluorescence-guided procedures. Interobserver agreement was defined as the calculated mean κ values for each observer. RESULTS A total of 36 questionnaires were included in the final analysis. The mean average κ value in fluorescence classification within the entire cohort was 0.71 ± 0.12 and 29 (81%) participants had a substantial or almost perfect interobserver agreement (κ values 0.6-1.0). Interobserver agreement was significantly higher in neurosurgeons (mean κ: 0.83) as compared with non-neurosurgeons involved in the neurooncological field (mean κ: 0.52; P