Large cell/anaplastic (LC/A) medulloblastoma (MB) is a recently recognized variant of medulloblastoma known to be associated with an advanced stage and a poor prognosis. Although Eberhart et al. suggested histopathologic grading of medulloblastoma in 2002, no consensus has been reached in terms of determining the criteria of an LC/A variant, and its biological behavior continues to be the subject of debate. We retrospectively analyzed 74 cases (range 0.25-15 years) of MB clinicopathologically using the criteria established by Eberhart et al. The LC/A variant was identified in 16 cases (22% of MB cases), five of which showed a poor outcome. Most LC/A variant cases revealed synaptophysin immunoexpression (75%), but no epidermal growth factor receptor (EGFR) expression. Expression of synaptophysin, NeuN, GFAP, p53, c-erbB2, and EGFR did not differ in LC/A and non-LC/A variants. Seven of the 74 cases of medulloblastoma showed erbB2 amplification by FISH, four of which were LC/A variants. N-myc amplification was observed in only one LC/A variant, but no c-myc amplification was found. In patients younger than 10 years, the LC/A variant showed a significantly poorer outcome than the non-LC/A variant (P = 0.02), while no difference was found in older patients. Multivariate analysis revealed only metastasis on MRI and p53 expression, but not anaplasia as unfavorable prognostic factors. Our study suggests that prognostic implications of anaplasia in medulloblastoma are uncertain, and that the reproducibility of the histopathologic criteria of the LC/A variant should be reassessed before they can be applied in practical use.