Phase IV study of propofol: validation of the data set

In 1989-1990, Stuart Pharmaceuticals conducted a Phase IV study of propofol on over 26,000 patients, later making the large data base available to a team of epidemiologists and anesthesiologists for analysis. We now describe the process of verifying the data to provide a sound basis for further analyses. Original data were collected by 1819 physicians at 1761 hospitals. In that study, anesthesia was induced by bolus injection of propofol and was maintained by inhaled drug and N2O-O2 (Step 1), or by propofol (either intermittent bolus injection [Step 2] or continuous infusion [Step 3]) and N2O-O2. Forty-six recorded variables described history, physical examination, course and quality of anesthesia and recovery, and adverse events. Data were scrutinized for inaccuracy or bias regarding adverse events, completeness of data, data entry, and violations of the study protocol. The initial data set pertained to 26,841 patients (10,698, Step 1; 8886, Step 2; and 7257, Step 3). Because we excluded data if 25% of the items were missing from the data set, 3.2% of the case reports were eliminated: the final data set used for subsequent analyses contained 25,981 patients (10,184, Step 1; 8672, Step 2; and 7125, Step 3). Inaccuracy of data entry was not excessive, and violations of study protocol were less frequent than in similar studies. The nature and frequency of adverse events were similar to those reported in Phase II and III clinical trials of propofol. Analysis showed that missing data occurred randomly and did not introduce obvious bias. We conclude that the data set was valid and most likely represents perioperative events occurring in similar patients; that Phase IV studies can be valuable because of the range of patients studied and the ability to detect even rare events; and that future Phase IV studies could be improved by more efficient design of data collection forms for both hypotheses to be tested and the entry of data onto forms.