PURPOSE: As no curative treatment for advanced pancreatic and biliary cancer with malignant ascites exists, new modalities possibly improving the response to available chemotherapies must be explored. This phase I study assesses the feasibility, tolerability and pharmacokinetics of a regional treatment of gemcitabine administered in escalating doses by the stop-flow approach to patients with advanced abdominal malignancies (adenocarcinoma of the pancreas, n = 8, and cholangiocarcinoma of the liver, n = 1). EXPERIMENTAL DESIGN: Gemcitabine at 500, 750 and 1,125 mg/m(2) was administered to three patients at each dose level by loco-regional chemotherapy, using hypoxic abdominal stop-flow perfusion. This was achieved by an aorto-caval occlusion by balloon catheters connected to an extracorporeal circuit. Gemcitabine and its main metabolite 2',2'-difluorodeoxyuridine (dFdU) concentrations were measured by high performance liquid chromatography with UV detection in the extracorporeal circuit during the 20 min of stop-flow perfusion, and in peripheral plasma for 420 min. Blood gases were monitored during the stop-flow perfusion and hypoxia was considered stringent if two of the following endpoints were met: pH </= 7.2, pO(2) nadir ratio </=0.70 or pCO(2) peak ratio >/=1.35. The tolerability of this procedure was also assessed. RESULTS: Stringent hypoxia was achieved in four patients. Very high levels of gemcitabine were rapidly reached in the extracorporeal circuit during the 20 min of stop-flow perfusion, with C (max) levels in the abdominal circuit of 246 (+/-37%), 2,039 (+/-77%) and 4,780 (+/-7.3%) mug/ml for the three dose levels 500, 750 and 1,125 mg/m(2), respectively. These C (max) were between 13 (+/-51%) and 290 (+/-12%) times higher than those measured in the peripheral plasma. Similarly, the abdominal exposure to gemcitabine, calculated as AUC(t0-20), was between 5.5 (+/-43%) and 200 (+/-66%)-fold higher than the systemic exposure. Loco-regional exposure to gemcitabine was statistically higher in presence of stringent hypoxia (P < 0.01 for C (max) and AUC(t0-20), both normalised to the gemcitabine dose). Toxicities were acceptable considering the complexity of the procedure and were mostly hepatic; it was not possible to differentiate the respective contributions of systemic and regional exposures. A significant correlation (P < 0.05) was found between systemic C (max) of gemcitabine and the nadir of both leucocytes and neutrophils. CONCLUSIONS: Regional exposure to gemcitabine-the current standard drug for advanced adenocarcinoma of the pancreas-can be markedly enhanced using an optimised hypoxic stop-flow perfusion technique, with acceptable toxicities up to a dose of 1,125 mg/m(2). However, the activity of gemcitabine under hypoxic conditions is not as firmly established as that of other drugs such as mitomycin C, melphalan or tirapazamine. Further studies of this investigational modality, but with bioreductive drugs, are therefore warranted first to evaluate the tolerance in a phase I study and later on to assess whether it does improve the response to chemotherapy.