BACKGROUND Bispectral Index (BIS) has been used to measure sedation depth. Ideally, to guide anesthetic management, range of BIS scores at different sedation levels should not overlap, and BIS should be independent of drug used. This study assessed ability of BIS to predict sedation depth between sevoflurane, propofol, and midazolam. Quality of recovery was also compared. METHODS Patients undergoing surgery with local or regional anesthesia and sedation were randomized to sevoflurane (n = 23), midazolam (n = 21), or propofol (n = 22). Sedation was titrated to Observers's Assessment of Alertness-Sedation score of 3 (responds slowly to voice). BIS and Observers's Assessment of Alertness-Sedation were measured every 5 min. BIS prediction probability (PK) was compared between drugs. Recovery was assessed by BIS and Digit Symbol Substitution and memory tests. RESULTS Bispectral Index of responders to voice was significantly different from nonresponders (86 +/- 10 vs. 74 +/- 14, mean +/- SD; P < 0.001) However, wide variability and overlap in BIS were observed (25th-75th percentile, responders vs. non-responders: 79-96 vs. 65-83). BIS of responders was different for sevoflurane versus propofol and midazolam. BIS was a better predictor of propofol sedation than sevoflurane or midazolam (PK = 0.87 +/- 0.11, 0.76 +/- 0.01, and 0.69 +/- 0.02, respectively; P < 0.05). At 10 min after the procedure, 76, 48, and 24% of sevoflurane, propofol, midazolam patients, respectively, returned to baseline Digit Symbol Substitution scores (P < 0.05). Excitement-disinhibition occurred in 70, 36, and 5% of sevoflurane, propofol, and midazolam patients, respectively (P < 0.05). CONCLUSIONS Individual BIS scores demonstrate significant variability, making it difficult to predict sedation depth. The relation between BIS and sedation depth may not be independent of anesthetic agent. Quality of recovery was similar between drugs, but excitement occurred frequently with sevoflurane.