Intraobserver and interobserver agreement in volumetric assessment of glioblastoma multiforme resection.

Abstract

The role of extent of tumor resection in improving outcome for patients with glioblastoma multiforme (GBM) is still under debate. To analyze intraobserver and interobserver agreement of manual segmentation as a method for volumetric assessment of GBM resection. Three observers performed volumetric assessment of preoperative tumor volume (PreTV) and postoperative tumor volume (PostTV) by manual segmentation on contrast-enhanced T1-weighted MRI data sets of 8 patients. Measurements were repeated after a minimum interval of 2 weeks. Intraobserver and interobserver agreement for PreTV, PostTV, and residual tumor volume (RTV) percentage were expressed in intraclass correlation coefficients (ICCs). Intraobserver agreement is high for PreTV (ICC = 0.99), PostTV (ICC = 0.73-0.94), and RTV (ICC = 0.89-0.94). Interobserver agreement is high for PreTV (ICC = 0.97), but low for PostTV (ICC = 0.54) and RTV (ICC = 0.52). Postoperative assessment of GBM volume seems to offer high intraobserver agreement, but low interobserver agreement. Using absolute RTV values to relate extent of tumor resection with survival may be unreliable. More research is needed before this method can be used as a valid end point for clinical studies. Computer-assisted tumor volume calculation may increase interobserver agreement in the future.