The association between p53 immunohistochemistry and TP53 mutation status has been controversial. The present study aims to re-evaluate the efficacy of p53 immunohistochemistry to predict the mutational status of TP53. A total of 157 diffuse gliomas (World Health Organization grades II–IV) were assessed by exon-by-exon DNA sequencing from exon 4 through 10 of TP53 using frozen tissue samples. Immunohistochemistry with a p53 antibody (DO-7) on paired formalin-fixed paraffin-embedded materials was assessed for the extent and intensity of reactivity in all cases. A total of 72 mutations were detected in 66 samples. They included 60 missense mutations, five nonsense mutations, four deletions and three alterations in the splicing sites. A receiver operating characteristic curve analysis revealed that strong p53 immunoreactivity in more than 10% of cells provided the most accurate prediction of mutation. Using this cutoff value, 52 of 55 immunopositive cases harbored a mutation, whereas only 14 of 102 immunonegative cases showed mutations, sensitivity and specificity being 78.8% and 96.7%. Tumors with frameshift mutations frequently showed negative immunostaining. Staining interpretation by an independent observer yielded comparable accuracy. We thus propose p53 immunohistochemistry as a moderately sensitive and highly specific marker to predict TP53 mutation.