To systematically review the benefits and risks associated with the use of benzodiazepines to treat insomnia in adults.

MEDLINE was searched from 1966 to December 1998 using the MeSH terms 'bezodiazepine' (exploded), 'benzodiazepine tranqillizers' (exploded), 'clonazepam', 'drug therapy', 'randomized controlled trial', 'random allocation', or 'all random', 'human', and English language'. A similar search was carried out in the Cochrane Controlled Trials Registry. Bibliographies of retrieved articles were checked and manufacturers contacted in an attempt to identify unpublished studies.

Study designs of evaluations included in the review

Only randomised controlled trials (RCTs) that compared a benzodiazipine with either placebo or with another (not benzodiazepine) active treatment. It should be noted that studies in which the active comparator was not available in Canada were excluded from the review.

Specific interventions included in the review

Various benzodiazepines (triazolam, flurazepam, temazepam, midazolam, nitrazepam, estazolam, lorazepam, diazepam, brotizolam, quazepam, loprazolam and flunitrazepam) were compared with either placebo or with another (not benzodiazepine) active treatment. The comparator active treatments included zopiclone, diphenhydramine, glutethimide and promethazine as well as one non-pharmacological intervention (behavioural therapy).

Participants included in the review

Patients suffering from insomnia. Various patient exclusion criteria were applied within individual studies (not all applied in all studies). These exclusion criteria were: those who were receiving or had recently undergone treatment for insomnia; those who had ever been diagnosed with a psychiatric disorder or had a 'serious medical problem', those who had a history of drug or alcohol abuse, those who were pregnant or lactating. The mean age of patients included in the review ranged from 29 to 82 years, although it was only reported in 33 of 45 studies. Fifteen studies included patients over 65 years of age.

Outcomes assessed in the review

The outcomes considered in the review were: sleep latency (1-7 day treatment); total sleep duration (1-7 day treatment); total number of adverse events.

How were decisions on the relevance of primary studies made?

The authors do not state how the papers were selected for the review, or how many of the reviewers performed the selection.

Individual reports were rated for quality on a scale from 0 to 5; for therapeutic efficacy this meant taking into account the quality of randomisation, blinding and follow-up, and for harmful effects it meant examining randomisation, blinding and control for baseline differences between groups. The first 21 articles were classified for quality independently in duplicate. Disagreement was resolved by consensus. Agreement that all validity criteria were met for a study of therapy was 95% (kappa value 0.90) and for an adverse effects study was 76% (kappa value 0.51). Disagreement was resolved by consensus and all subsequent papers were assessed by one reviewer only.

The authors do not state how the data were extracted for the review. The data extraction of the first 21 articles was performed independently in duplicate, but the data from all subsequent papers were extracted by one reviewer only. The categories of data extracted were: study design; condition treated, patient characteristics, setting, duration of the study and the outcome measures. Where appropriate odds ratios (95% confidence intervals (CI) were calculated.

How were the studies combined?

The studies were combined using both narrative and meta-analysis techniques. Studies that had reported data in a comparable way were combined using fixed-effect methods. Mantel-Haenszel common odds ratios (ORs) (95% CI) were obtained for discrete data. Forest plots of the pooled differences or ORs for each of the outcomes measured was presented.

How were differences between studies investigated?

Heterogeneous results were checked using a random-effects model. For discrete data the Breslow-day test for homogeneity was applied and if study results were heterogeneous, the studies were subdivided into predefined groups (type of benzodiazepine, high or low dosage, setting, and quality of methodology) and the common ORs were recalculated.

A total of 89 RCTs were identified and 45 studies were included in the review (n=2672 patients). Of these only 14 studies were included in the meta-analysis (number of patients not clear). The reasons for study exclusion at this stage are listed in the review. Of the studies included in the review, 16 involved triazolam, 14 used flurazepam, 13 temazepam, five midazolam, four nitrazepam and two used estazolam. All other benzodiazepines were used in only one study each.

The methodological quality of the studies included in the review was not uniform. Of the 41 studies that reported on efficacy only 26 (63.4%) met the rudimentary criteria of good follow-up and double-blinding, and for the 45 studies that reported on adverse effects only 25 (55.6%) met all the quality criteria.

For the effects on sleep latency, eight comparisons between a benzodiazepine and placebo in four studies involving 159 patients were included in a meta-analysis. The pooled difference indicated that the latency to sleep for patients receiving a benzodiazepine was 4.2 minutes (95% CI:-0.7 to 9.2) shorter than those receiving placebo. Patient's estimate of sleep latency was analysed from eight studies (n=539) and the pooled difference was 14.3 minutes (95% CI: 10.6 to 18.0) in favour of benzodiazepines although this analysis showed statistically significant heterogeneity. When only the better quality studies were included in the analysis, the heterogeneity ceased to reach statistical significance and although the treatment difference in favour of benzodiazepines was reduced, it was still statistically significant.

Effects on sleep duration were analysed from two studies (n=35 patients). The pooled difference indicated that the duration of sleep for patients receiving a benzodiazepine was 61.8 minutes (95% CI: 37.4, 86.2) longer than those receiving placebo. The patient's estimate of sleep duration assessed from eight studies (n=566) found a treatment difference of 48.4 minutes (95% CI: 39.6, 57.1) in favour of benzodiazepines. The possibility of tolerance developing to any sleep promoting effect of benzodizepines could not be investigated due to the fact that all studies eligible for the meta-analysis were of short duration.

Seven studies (n=821) were pooled to analyse the proportion of patients reporting adverse events. Patients receiving a benzodiazepine were more likely to report an adverse event over the next three to seven days of therapy(OR 1.8, 95% CI: 1.4, 2.4).

Six good quality studies that could not be included in the meta-analysis were reported to support the findings of the meta-analysis. These studies are not included in the printed version of the review but are apparently available from the journal's website.

Benzodiazepines compared with zopiclone was assessed for efficacy in three studies (n=96). There was no statistically significant difference between treatments for sleep latency but the pooled difference for sleep duration favoured benzodiazepines (23.1 minutes, 95% CI: 5.6, 40.6). These findings were supported by data from six studies that could not be included in the meta-analysis and are apparently available from the journal's website. Data from four studies (n=252) gave a pooled OR for adverse effects of 1.5 (95% CI: 0.8, 2.9) indicating adverse events were more common with benzodiazepines than with zopiclone.

Comparisons of benzodiazipines with other agents did not detect any statistically significant differences. Other trials selected for the review but not included in the meta-analysis are presented in additional tables (3 to 5 inclusive) at the web site previously noted.

The use of benzodiazepines in the treatment of insomnia is associated with an increase in sleep duration, but this is countered by a number of adverse effects. Benzodiazepines do not provide a major advantage over placebo. There does not appear to be a clearly efficacious and safe pharmaceutical alternative to benzodiazepines.

The review addressed an appropriate question with adequately defined inclusion and exclusion criteria. The search for relevant studies was limited to two databases and therefore, although the electronic searches were supplemented with checking of bibliographic lists, some important studies could have been missed. The authors did contact the manufacturers of benzodiazepines in an attempt to include unpublished studies, but it is not clear to what extent this was successful, or what impact was made upon publication bias. Although studies were assessed for validity, the exact nature of the rating scales used was not described in the review, making interpretation difficult. Details of the individual studies included in the review are not tabulated or described in the review. The review does include a web-site address where details of (some or all) studies not included in the meta-analysis (i.e. quantitative synthesis) are tabulated. However it should be noted that the web page given in the article may not be accessible. Although the review included a total of 45 studies, only 14 of these were included in the meta-analysis and the authors do not give explicit criteria for inclusion/exclusion in the meta-analysis. The narrative synthesis of the remaining studies is very brief. The authors conclusions are supported by the data presented in the review, except that the association of benzodiazepines with an increased sleep duration is supported by data from only 35 patients relating to an objective measure of this outcome.

Practice: The authors state 'In the light of the strength of the placebo effect in patients with insomnia, physicians and patients who believe that benzodiazepines are highly effective may wish to reconsider their treatment choice'.

Research: The authors suggest that two major clinical questions, relating to the impact of benzodiazepines on the quality of patients' sleep, quality of life and functional status, and the risk/benefit ratio of benzodiazepine use, remain unanswered. They also stated that 'Additional studies evaluating the efficacy of non-pharmacological interventions would be valuable'.

Canadian Pharmaceutical Association; Canadian Medical Association.

1. Lord RW. What are the benefits and risks associated with the use of benzodiazepines to treat insomnia? J Fam Pract 2000;49(5):468-4692 This additional published commentary may also be of interest: Endeshaw Y. The role of benzodiazepines in the treatment of insomnia. J Am Geriatr Soc 2001 49:824-826.

This is a critical abstract of a systematic review that meets the criteria for inclusion on DARE. Each critical abstract contains a brief summary of the review methods, results and conclusions followed by a detailed critical assessment on the reliability of the review and the conclusions drawn.