Although tolerance to midazolam is sometimes described in the palliative care literature, no studies have systemically examined the possibility. To explore the association between midazolam dose for symptom palliation and the administration period, a retrospective study was performed on 62 terminally ill cancer patients who required parenteral midazolam in the final three days of life. The mean maximum dose and administration period of midazolam were 38 +/- 45 mg/day (median = 24) and 10 +/- 19 days (median = 2.5), respectively. Thirteen patients (21%) received midazolam at a dose of 60 mg/day or more, and 13 patients (21%) received it for 14 days or longer. The maximum doses were significantly correlated with patient age (rho = -0.32, P = 0.012) and the administration period (rho = 0.47, P < 0.01); and were significantly higher in patients who received midazolam for 14 days or longer (74 +/- 63 mg/day vs. 28 +/- 34 mg/day, P < 0.01). Multivariate analyses revealed that younger age (< or =70) and longer administration periods (> or =14 days) were independent determinants for a midazolam requirement of 60 mg/day or more (odds ratios [95% C.I.] = 0.091 [0.009 - 0.92], P = 0.042; 11 [2.3 - 54], P < 0.01; respectively). The significant correlation of midazolam doses with administration period suggests that the longer use of midazolam can result in the development of tolerance. This finding suggests that midazolam should be reserved for patients with limited prognoses.