Twenty-nine children treated for medulloblastoma between 1987 and 1991 were reviewed. Thirteen patients with high-risk medulloblastoma characterized by incomplete resection, diploid tumor or subarachnoid dissemination received chemotherapy following radiation therapy. Three received postoperative chemotherapy. Eight patients who had been treated with postoperative radiation therapy also received chemotherapy for recurrent tumors. After a minimum 3-year follow-up period, 16 were alive but 13 had died from recurrent tumors. In order to evaluate the possible participation of P-glycoprotein (Pgp)-mediated multidrug resistance (MDR) in medulloblastoma therapy and its correlation with prognosis, archival specimens were examined by immuno-histochemistry utilizing 3 monoclonal antibodies against Pgp and 6 cases by reverse-transcriptase polymerase chain reaction (RT-PCR) using MDR1-specific primers. Sixteen patients (55%) had MDR expression detected either by 1 of the 3 antibodies or by RT-PCR. DNA ploidy study was also performed on 18 specimens. We correlated patients' outcome with variable factors (extent of surgical resection, chemotherapy, DNA ploidy) and MDR expression. Patients who were treated with radiation therapy and adjuvant chemotherapy had a significantly better (p = 0.036) survival than those with radiation therapy alone, despite the fact that the former group of patients was considered to be high-risk. The extent of surgical resection and DNA ploidy did not correlate with prognosis. However, a statistically significant association was found between MDR expression and outcome (p = 0.007). Among the patients who received chemotherapy, positive MDR expression significantly correlated with poor outcome (p = 0.036). Our results showed that Pgp-mediated intrinsic MDR in medulloblastomas seems to correlate with an adverse outcome. This information may be used in designing new therapeutic protocols for medulloblastoma.