Interobserver variability associated with the MIB-1 labeling index - High levels suggest limited prognostic usefulness for patients with primary brain tumors

BACKGROUND. The use of the MIB-1 labeling index (LI) as a potential prognos tic marker for patients with primary brain tumors is controversial. Many st udies advocating its prognostic usefulness have suggested discrete MIB-1 LI cut-off values, above which patients have significantly worse outcomes. Ho wever, interobserver variability associated previously with MIB-1 LI calcul ation has not been reported despite the fact that the degree of interobserv er variability impacts the clinical usefulness of such cut-off values. METHODS. MIB-1 LIs were calculated independently using a standardized proto col by six pathologist observers for 50 astrocytic gliomas of varying grade s. The level of interobserver agreement was determined by calculating kappa statistics for pairwise pathologist comparisons using MIB-1 LI cut-off val ues of 2.5%, 5.0%, 8.0%, 11.0%, and 15.0%. Spearman rank correlation coeffi cients were used to assess the pairwise associations between observer MIB-1 LIs. RESULTS. Although there was general agreement among pathologists regarding whether an MIB-1 LI for a given astroglial tumor was low, moderate, or high based on the analysis of correlation, a high level of interobserver variab ility was associated with the determination of specific MIB-1 LIs. The high est level of agreement occurred using a cut-off value of 5.0%, with pairwis e kappa statistics for this value ranging from 0.52 to 0.80. CONCLUSIONS. The high level of interobserver variability suggests that prop osed discrete MIB-1 LI prognostic cut-off values most likely are not useful clinically for predicting outcome for individual patients with primary bra in tumors. Further prospective studies are needed investigating the prognos tic usefulness of MIB-1 LI ranges that optimize interobserver agreement. (C ) 2001 American Cancer Society.