A phase I dose-escalation study was performed to determine whether isolated hepatic perfusion (IHP) with melphalan (L-PAM) allows exposure of the liver to much higher drug concentrations than clinically achievable after systemic administration and leads to higher tumour concentrations of L-PAM. Twenty-four patients with colorectal cancer confined to the liver were treated with L-PAM dosages escalating from 0.5 to 4.0 mg kg(-1). During all IHP procedures, leakage of perfusate was monitored. Duration of IHP was aimed at 60 min, but was shortened in eight cases as a result of leakage from the isolated circuit. From these, three patients developed WHO grade 3-4 leukopenia and two patients died due to sepsis. A reversible elevation of liver enzymes and bilirubin was seen in the majority of patients. Only one patient was treated with 4.0 mg kg(-1) L-PAM, who died 8 days after IHP as a result of multiple-organ failure. A statistically significant correlation was found between the dose of L-PAM, peak L-PAM concentrations in perfusate (R = 0.86, P< or =0.001), perfusate area under the concentration-time curve (AUC; R = 0.82, P<0.001), tumour tissue concentrations of L-PAM (R = 0.83, P = 0.011) and patient survival (R = 0.52, P = 0.02). The peak L-PAM concentration and AUC of L-PAM in perfusate at dose level 3.0 mg kg(-1) (n = 5) were respectively 35- and 13-fold higher than in the systemic circulation, and respectively 30- and 5-fold higher than reported for high dose oral L-PAM (80-157 mg m(-2)) and autologous bone marrow transplantation. Median survival after IHP (n = 21) was 19 months and the overall response rate was 29% (17 assessable patients; one complete and four partial remissions). Thus, the maximally tolerated dose of L-PAM delivered via IHP is approximately 3.0 mg kg(-1), leading to high L-PAM concentrations at the target side. Because of the complexity of this treatment modality, IHP has at present no place in routine clinical practice.