The Role of Benzodiazepines in the Treatment of Insomnia

PURPOSE: To obtain a precise estimate of the efficacy and common adverse effects of benzodiazepines for the treatment of insomnia compared with those of placebo and other treatments. BACKGROUND: Insomnia, also referred to as disorder of initiating or maintaining sleep, is a common problem and its prevalence among older people is estimated to be 23% to 34%. 1 The total direct cost in the United States for insomnia in 1995 was estimated to be $13.9 billion. 2 The complaint of insomnia in older people is associated with chronic medical conditions; psychiatric problems, mainly depression, chronic pain, and poor perceived general condition; 1,3,4 and use of sleep medications. 5 Thus in most cases, insomnia is due to some other underlying problem and is not just a consequence of aging. 6 Accordingly, the management of insomnia should focus on addressing the primary problem and not just short‐term treatment of the insomnia. Benzodiazepines belong to the drug class of choice for the symptomatic treatment of primary insomnia. 7 This abstract will appraise a meta‐analysis that compared the effect of benzodiazepines for short‐term treatment of primary insomnia with placebo or other treatment. DATA SOURCES: Data sources included articles listed in Medline from 1966 to December 1998 and the Cochrane Controlled Trials Registry. The medical subject heading (MeSH) search terms used were “benzodiazepine” (exploded) or “benzodiazepine tranquillizers” (exploded) or “clonazepam,”“drug therapy,”“randomized controlled trial” or “random allocation” or “all random,”“human,” and “English language.” In addition, bibliographies of retrieved articles were scanned for additional articles and manufacturers of brand‐name benzodiazepines were asked for reports of early trials not published in the literature. STUDY SELECTION CRITERIA: Reports of randomized controlled trials of benzodiazepine therapy for primary insomnia were considered for the meta‐analysis if they compared a benzodiazepine with a placebo or an alternative active drug. DATA EXTRACTION: Data were abstracted from 45 randomized controlled trials representing 2,672 patients, 47% of whom were women. Fifteen studies included patients age 65 and older and four studies involved exclusively older patients. Twenty‐five studies were based in the community and nine involved inpatients. The duration of the studies ranged from 1 day to 6 weeks, with a mean of 12.2 days and median of 7.5 days. The primary outcome measures analyzed were sleep latency and total sleep duration after a sleep study, subjects' estimates of sleep latency and sleep duration, and subjects' report of adverse effects. Interrater reliability was checked through duplicate, independent abstraction of the first 21 articles. Overall agreement was between 95% and 98% (kappa value of 0.90 and 0.95 accordingly) for classification of the studies and validity of therapy, and 76% (kappa value of 0.51) for study of harmful effects. A scale of 0 to 5 was used to rate the individual reports, taking into account the quality of randomization, blinding, follow‐up, and control for baseline differences between groups. Tests for homogeneity were applied across the individual studies and, when studies were found to be heterogeneous, subgroup analysis according to a predefined group was performed. MAIN RESULTS: The drugs used in the meta‐analysis included triazolam in 16 studies; flurazepam in 14 studies; temazepam in 13 studies; midazolam in five studies; nitrazepam in four studies; and estazolam, lorazepam, and diazepam in two studies each. Alternative drug therapies included zopiclone in 13 studies and diphenhydramine, glutethimide, and promethazine in one study each. Only one article reported on a nonpharmacological treatment (behavioral therapy). The mean age of patients was reported in 33 of the 45 studies and ranged between 29 and 82. SLEEP LATENCY: In four studies involving 159 subjects, there was sleep‐record latency (time to fall asleep) data for analysis. The pooled difference indicated that the latency to sleep for patients receiving a benzodiazepine was 4.2 minutes (95% CI = (−0.7) (−9.2)) shorter than for those receiving placebo. Patient's estimates of sleep latency examined in eight studies showed a difference of 14.3 minutes (95% CI = 10.6–18.0) in favor of benzodiazepines over placebo. TOTAL SLEEP DURATION: Analysis of two studies involving 35 patients in which total sleep duration using sleep‐record results was compared indicated that patients in the benzodiazepine groups slept for an average of 61.8 minutes (95% CI = 37.4–86.2) longer than those in the placebo groups. Patient's estimates of sleep duration from eight studies (566 points) showed total sleep duration to be 48.4 minutes (95% CI = 39.6–57.1) longer for patients taking benzodiazepines than for those on placebo. ADVERSE EFFECTS: Analysis of eight studies (889 subjects) showed that those in the benzodiazepine groups were more likely than those in the placebo groups to complain of daytime drowsiness (odds ratio (OR) 2.4, 95% confidence interval (CI) = 1.8–3.4). Analysis of four studies (326 subjects) also showed that subjects in the benzodiazepine groups were more likely to complain of dizziness or lightheadedness than the placebo groups. (OR 2.6, 95% CI = 0.7–10.3). Despite the increased reported side effects in the benzodiazepine groups, drop‐out rates were similar in the benzodiazepine and placebo groups. For patient reported outcome, there was no strong correlation found for sleep latency data, ( r = 0.4, 95% CI = (−0.3) (−0.9)) or for sleep duration ( r = 0.2, 95% CI = −0.8–0.4) between benzodiazepine dose and outcome. COMPARISON WITH OTHER DRUGS AND TREATMENTS: In three trials with 96 subjects, meta‐analysis of the results comparing benzodiazepines with zopiclone, did not show significant difference in sleep latency in the benzodiazepine and placebo groups, but the benzodiazepine groups had increased total sleep duration (23.1 min. 95% CI = 5.6–40.6). In four trials with 252 subjects, the side effect profile did not show a statistically significant difference (OR 1.5, CI 0.8–2.9). There was only one study comparing the effect of behavioral therapy with triazolam. The result showed that triazolam was more effective than behavioral therapy in decreasing sleep latency, but its efficacy declined by the second week of treatment. Behavioral therapy remained effective throughout the 9‐week follow‐up period. There were four small trials that involved older patients exclusively, with three of the studies having less than 2 weeks of follow‐up. The results were mixed regarding benefits and adverse effects were poorly reported. CONCLUSION: The result of the meta‐analysis shows that the use of benzodiazepines results in a decrease in sleep latency and a significant increase in total sleep time as compared with placebo. There was also a report of significantly increased side effects, but this did not result in increased discontinuation rate. There was no dose‐response relationship for beneficial effect seen with the use of benzodiazepines, although the data are scant. Zopiclone was the only alternative pharmacological therapy that could be studied with any precision. There was no significant difference in the outcome when benzodiazepines were compared with zopiclone. There was only one study that compared the effect of benzodiazepines with nonpharmacological therapy; thus available data are insufficient to comment.