Background and objective Midazolam may occasionally precipitate hostility and violence instead of tranquillity. We characterized these episodes, their rate of occurrence, the potential paradoxical responders and possible predisposing circumstances among patients undergoing lower body surgery under spinal or epidural anaesthesia and midazolam sedation.

Patients and methods Fifty-eight patients who fulfilled the study entry criteria and who underwent surgery within a 3-month period in a large metropolitan, university-affiliated hospital were enrolled. Sedation and restlessness in all patients were controlled by midazolam administered intravenously by the attending anaesthesiologist; these parameters were later objectively confirmed by recorded actigrams. If ‘paradoxical’ events occurred, flumazenil 0.1 mg 10 s−1 was injected until the aberrant behaviour ceased. Patients with paradoxical reactions were later compared with matched control patients selected from the study group to identify epidemiological characteristics.

Results The incidence of paradoxical events was 10.2% (six out of 58 patients, confidence limits 2.3–18.3%) and they occurred 45–210 min after sedation started; the only independent predictor was an age older than that of the entire study group. The mean cumulative and per weight doses of midazolam were similar for both the experimental and the study groups of patients: 7.3 ± 2.8 to 10.1 ± 3.6 mg, and 0.1 ± 0.04 to 0.12 ± 0.05 mg kg−1. Flumazenil 0.2–0.3 mg (range 0.1–0.5 mg) effectively stopped the midazolam-induced paradoxical activity within 30 s and surgery continued uneventfully.

Conclusions Flumazenil completely reverses midazolam-induced paradoxical reactions and they are more frequent in older patients.