Performance of computer-assisted continuous infusion at low concentrations of intravenous sedatives

We studied the performance of a target-controlled drug infusion device, computer-assisted continuous infusion (CACI). Forty-one volunteers received one of midazolam (n=11), propofol (n=10), thiopental (n=10), or fentanyl (n=10) in sedative concentrations. Concentrations were kept constant for 45-70 min at five sequential target concentrations in each subject. Twenty-six subjects had arterial sampling and 15 had venous sampling to determine drug concentrations. Median performance errors, median absolute performance error (MDAPE), wobble, divergence, and median absolute constancy error (MDACE), defined as error around mean actual concentration at each target, were calculated. CACI demonstrated significant performance errors, which were different among drugs. MDAPE (5%-95% confidence interval) ranged from 22.9% (12.1%-39.6%) for propofol to 82.2% (36.0%-183.0%) for midazolam. Although performance errors could be large, CACI was able to maintain a constant serum concentration over time very successfully. The MDACE ranged from 5.6% (3.9%-17.3%) for fentanyl to 11.2% (8.9%-20.4%) for propofol. Few differences occurred between arterial and venous sampling, although when they occurred, arterial samples indicated larger errors. It is concluded that CACI is very successful at maintaining constant serum concentrations of these drugs at sedative concentrations. Arterial sampling should be used when the performance characteristics of an infusion device are being tested. However, venous sampling may be adequate to determine serum concentrations when a pseudo-steady state has been achieved.