Previous reports on osteosarcomas treated with multi-agent chemotherapy have shown that P-glycoprotein expression is a reliable prognostic indicator. The current thinking is that, of the several agents used for the treatment of osteosarcoma, only doxorubicin is involved in drug resistance mediated by P-glycoprotein. This study examines the relationship of P-glycoprotein expression to clinical outcome in osteosarcomas, treated only with doxorubicin in addition to surgery, to determine if the prognostic significance of P-glycoprotein expression reflects the ability of osteosarcoma to respond to this drug. The expression of P-glycoprotein in tumor specimens was assessed by immunohistochemistry in 37 nonmetastatic, operable osteosarcomas treated at a single institution with doxorubicin as a single adjuvant drug. The P-glycoprotein status was analysed in relation to the length of event-free survival. A widespread pattern of P-glycoprotein expression in tumor cells at diagnosis was significantly associated with a higher rate of systemic relapse (p < 0.001). On comparison of this group of patients with a similar series of 92 patients, all treated with multi-agent chemotherapy plus surgery of the primary lesion and previously analysed for P-glycoprotein status, only P-glycoprotein-positive, doxorubicin-resistant tumors consistently benefited from the addition of drugs other than doxorubicin (p < 0.001). Osteosarcomas with different abilities to respond to adjuvant chemotherapy can be: identified by the expression of P-glycoprotein in tumor cells at the clinical onset. P-glycoprotein status may serve as a basis for risk-adapted, individualized therapeutic regimens. Standard programs are sufficient for P-glycoprotein-negative osteosarcomas, whereas P-glycoprotein-positive tumors may benefit from the use of mure intensive therapeutic approaches.