Interobserver variability in the semi-quantitative assessment of 201Tl SPECT in cerebral gliomas

Abstract

201Tl SPECT is used successfully in the diagnosis of recurrent supratentorial glioma and in the evaluation of its response to chemotherapy. However, different methods are used to measure relative tracer uptake in tumour and background. The objective of this study was to assess the interobserver variability of such methods, and, if possible, to provide nomograms for data conversion. Using baseline and follow-up SPECT scans from 20 patients with recurrent glioma treated with chemotherapy, three observers applied manual and semi-automatic ROI techniques to define activity in tumour (manual, semi-automatic) as well as in reference tissue (scalp, mirror, hemisphere). All tumour ROI techniques had intra-class correlation coefficients (ICC) > or = 0.80 indicating almost perfect agreement. The main source of variation with the manual techniques was the tumour intensity; with the semi-automatic method, observer agreement was independent of the level of tumour activity. Agreement for background ROIs was also adequate, but the mirror technique tended to perform poorer at follow-up SPECT scans (ICC 0.68). Measurement of fractional change during treatment revealed no significant differences between observers for any of the investigated ROI methodology variants. Conversion of quantitative methods to measure fractional change was possible using linear regression analysis. 201Tl SPECT in recurrent glioma appears to be a robust method with acceptable interobserver variability. The clinical field in neuro-oncology should consider including 201Tl SPECT parameters in monitoring response to chemotherapy.