Multicentre audit of the use of drotrecogin alfa (activated) in UK critical care units

Background

Following positive results from PROWESS, drotrecogin alfa (activated) (DrotAA) was approved for use in Europe in August 2002. At this time, ICNARC commenced an audit to monitor the diffusion of the drug into routine UK practice and to undertake a nonrandomised evaluation of its effectiveness.

Methods

A data collection form was developed and tested to mirror the information collected in PROWESS. This form was completed for every admission that received DrotAA and a senior clinician confirmed completeness. Data were entered centrally and validated.

Analysis

Admissions receiving DrotAA and with severe sepsis and two or more organ dysfunctions in the first 24 hours following admission to the unit were matched to controls on: source of admission; organ dysfunctions; ICNARC physiology score; and age. Four pools of control patients were used for matching: (a) historic admissions (January 2000–August 2002) from the same unit; (b) contemporaneous admissions from the same unit; (c) contemporaneous admissions from units that never used DrotAA; and (d) contemporaneous admissions from units prior to their first use of DrotAA. Analyses were undertaken using conditional, fixed-effects, Poisson regression.

Results

One hundred and twelve units participated in the audit; 1,079 admissions (one in 16) with severe sepsis and two or more organ dysfunctions in the first 24 hours following admission to the unit received DrotAA. For the four control pools, matching was successful for: (a) 657 (61%); (b) 820 (76%); (c) 702 (65%); and (d) 965 (89%). Matched cases were older, more acutely ill and had higher hospital mortality than unmatched cases. The relative risks (95% confidence interval) associated with DrotAA were: (a) 0.84 (0.77–0.92); (b) 0.85 (0.78–0.93); (c) 0.75 (0.68–0.83); and (d) 0.80 (0.73–0.86). A priori subgroup analyses indicated greater effect for patients with three or more organ dysfunctions.

Interpretation

All results were consistent with PROWESS, but need to be interpreted with caution due to their nonrandomised nature and the potential existence of important unknown confounders. In addition, the fact that only one in 16 potentially suitable admissions received DrotAA suggests a strong possibility for treatment bias.