This prospective open-label study was conducted in a 15-bed intensive care unit of the teaching hospital of Saint-Etienne, France. Patients meeting the following criteria were eligible for inclusion in the study: age over 18 years; clinically suspected or proven bacterial infection; isolated or expected causative pathogen susceptible to ceftazidime; and acute renal failure requiring CVVHDF. Exclusion criteria were known allergy to ceftazidime or other β-lactams; use of ceftazidime within the 48 hours before enrollment; pregnancy, as determined by serum β-human chorionic gonadtotrophin testing; and residual glomerular filtration rate, measured by creatinine clearance, over 10 ml/min. The study protocol and the consent document were approved by the institutional review board, and written informed consent was obtained from each patient or a legally designated representative.
Continuous venovenous haemodiafiltration
CVVHDF was performed using the PRISMA machine (Hospal, Meyzieu, France) equipped with a Multiflow 60 AN69HF 0.60 m2 polyacrylonitrile hollow-fibre membrane. Vascular access was obtained by introduction of a 12 French, 16 or 20 cm double-lumen central venous catheter (Arrow, Reading, PA, USA) into a femoral vein. For all patients, operational characteristics of haemodialfiltration were set as follows: blood flow rate 150 ml/min; dialysate flow rate 1 l/hour; and ultrafiltration rate 1.5 l/hour. Substitution fluids were delivered according to a predilutional mode with a flow rate allowing 100–150 ml/min net ultrafiltration. Under this haemodiafiltration regimen, CVVHDF clearance of ceftazidime was predicted to be 32.5 ml/min .
Anticoagulation of the extracorporeal circuit was ordered at the discretion of the attending physician. If a patient was already undergoing CVVHDF, then the haemofilter was changed before the patient's inclusion in the study.
Ceftazidime dosage and administration
After initiation of CVVHDF, all patients received a 2 g intravenous loading dose of cefatzidime (GlaxoSmithKline, Marly-le-Roi, France) infused over three minutes, via a central venous catheter different from that used for CVVHDF, and immediately followed by a 3 g continuous infusion over 24 hours. The ceftazidime dose was expected to provide serum antibiotic concentrations between 30 and 40 mg/l, and was selected according to the equation R0 = CLtot × Css, where R0 is the continuous administration rate, Css is the steady-state serum concentration of ceftazidime, and CL tot is the total clearance of ceftazidime (estimated using the equation CLtot = 0.693 × [volume of distribution/elimination half-life]). We anticipated a volume of distribution close to 0.30 l/kg based on data available in intensive care patients , and hypothesized a half-life of four hours, which is in keeping with the half-life found in a recent study  that examined similar patients and reported a haemofiltration clearance of ceftazidime in the same range as our predicted haemodiafiltration clearance.
Ceftazidime was continuously infused by the means of syringe infusion pump (Ivac Medical System, Basingstoke, UK) for at least 72 hours. Syringes with freshly dissolved antibiotic, were inserted every 16 hours. In case of CVVHDF clotting, the continuous infusion was stopped during change of haemofilter.
Sample collection and assay
Arterial blood samples (7 ml) were collected from a radial catheter before the loading dose then at 3 (immediately after the 2 g bolus infusion), 15 and 30 minutes and at 1, 2, 4, 6, 8, 12, 24, 36, 48 and 72 hours. After collection, blood samples were immediately centrifuged (2000 g for 10 minutes) and stored at -80°C until analysis.
Seven samples (10 mL) were simultaneously drawn from the dialysate/ultrafiltrate outlet before ceftazidime infusion and at 2, 8, 12, 24, 36 and 48 hours. After collection, dialysate/ultrafiltrate samples were frozen and stored at -80°C until analysis. The concentrations of ceftazidime in the serum and dialysate/ultrafiltrate were assayed by high-performance liquid chromatography, employing the technique described by Jehl and coworkers .
Within-day and between-day coefficients of variability were both below 10% and the limit of quantification was 0.05 mg/l for serum and 1.0 mg/l for dialysate.
A noncompartmental model was applied. The following pharmacokinetic parameters were determined for each patient. The steady-state serum concentration of ceftazidime (mg/l) was calculated as the mean of the serum concentrations measured at 24, 36, 48 and 72 hours. The elimination half-life (hours) was calculated as ln2/ke, where ke is the apparent terminal elimination rate constant determined using least-squares regression analysis. The area under the concentration-time curve from time zero to 72 hours (mg/hour per l) was calculated using the linear trapezoidal summation method. The total clearance (ml/min) was calculated as R0/steady-state serum concentration of ceftazidime, where R0 is the continuous administration rate. Finally, the volume of distribution (l) was calculated as D/C0, where D is the loading dose and C0 the serum concentration at the end of the bolus infusion.
In addition, the following CVVHDF parameters were assessed. The sieving coefficient was calculated as CD/UF/CS, where CD/UF is the dialysate/ultrafiltrate concentration of ceftazidime and CS the time-corresponding serum concentration. The haemodiafiltration clearance (ml/min) was calculated as sieving coefficient × combined dialysate/ultrafiltrate flow rate . The contribution (%) of haemodiafiltration clearance to total clearance of ceftazidime was calculated as follows: (haemodiafiltration clearance/total clearance) × 100.
All calculations were made by programming pharmacokinetic and CRRT clearance equations into Microsoft Excel® 97 (Microsoft Corporation, Irvine, CA, USA). The same software program was used to compute mean and standard deviations for the various pharmacokinetic and haemodiafiltration parameters.