The efficacy and safety of zolpidem, a hypnotic of a new chemical class (the imidazopyridines), was compared with a reference benzodiazepine in elderly insomniac patients in a randomized, double-blind, multicenter trial. Hospitalized patients aged 58 to 98 years were randomized to receive zolpidem 5 mg (70 patients), zolpidem 10 mg (74 patients), or triazolam 0.25 mg (77 patients) at bed-time. (Three patients were excluded and 13 patients did not complete the study.) The 3-week active treatment period was preceded by 3 and followed by 7 days of placebo administration. Sleep quality was assessed by the patient via a questionnaire and visual analog scale. A clinician's global impression was also recorded. All measures of sleep quality were improved by both doses of zolpidem and by triazolam. The improvements between the end of the placebo phase and the end of the active treatment phase were significant for all treatments and assessment instruments. Moreover, the significant improvements in all measures were maintained during the week following withdrawal of both doses of zolpidem. The improvement in most sleep assessment parameters was maintained after withdrawal of triazolam. Tolerability of all treatments was excellent. The majority of patients reported no adverse effects. The reported adverse effects in all groups included nightmares, daytime drowsiness, and day- or nighttime agitation. There was no evidence of rebound insomnia; the therapeutic effect of zolpidem outlasted drug treatment. There were no signs of agitation or anxiety following cessation of treatment, which might be indicative of withdrawal phenomena. Confusion was recorded only in the triazolam group. These results suggest that zolpidem is at least as effective as triazolam in geriatric insomniac patients. Zolpidem 5 mg and 10 mg demonstrated a good safety profile. On the basis of these data, zolpidem 5 mg should be given as a starting dose in elderly patients, with a possible increase up to 10 mg in more severe cases of insomnia.