Quantity of Research Available
A total of 503 citations were identified in the literature search. Following screening of titles and abstracts, 465 citations were excluded and 38 potentially relevant reports from the electronic search were retrieved for full-text review. Three potentially relevant publications were retrieved from the grey literature search. Of the 41 potentially relevant articles, 30 publications were excluded for various reasons, while 11 publications met the inclusion criteria and were included in this report. Appendix 1 describes the PRISMA flowchart of the study selection.
Additional references of potential interest are provided in Appendix 5.
Summary of Study Characteristics
Appendix 2 provides further details of individual study characteristics
Study Design
Three systematic reviews,8–10 five randomized controlled trials (RCTs),2, 11–14 and three non-randomized trials.15–17 were identified. Two of the RCTs had cluster designs.2, 12
Country of Origin
The systematic reviews were published by authors originating from the United Kingdom (UK),9 USA,8 and Australia.10 The systematic review from the UK included 16 RCTs from Argentina, Australia, Canada, the UK and seven other European countries. The systematic review from Australia was based on 3 RCTs all from the UK. The systematic review from the USA involved 28 studies of heterogeneous designs (information on primary study designs was not provided) and the countries of origin were not stated. The RCTs included in this review were published by authors originating from Canada,12 Denmark,14 Finland,11 Spain,2 and the USA.13 The non-randomized studies included in this review were published by authors originating from The Netherlands,17 and Spain.15, 16
Patient Population
Participants in all the included studies2, 8–17 were adults with mean ages ranging from 41 to 79 years who received BZD for the treatment of insomnia, anxiety, panic disorders, or psychiatric disorders.
Interventions and Comparators
A combination of BZD dose-tapering with other psychotherapy measures (including cognitive behavioral therapy [CBT]) patient education, written self-help instructions, and pharmacotherapy was the most common intervention in the included studies.
One systematic review9 included a combination of gradual BZD withdrawal (dose-tapering) with either a psychotherapy or prescribing intervention (e.g. medication review, consultation or education), compared with each component alone, or with either treatment as usual, education with/without placebo, or tapering with drug support. Another systematic review8 compared a tapering intervention alone with a combination of tapering plus either CBT or medication substitution. A third systematic review10 compared minimal intervention (e.g. letter from a clinician, self-help information, or short consultation with a general practitioner [GP]) with continuation of usual doses.
One RCT14 compared melatonin with placebo, each in combination with slow tapering of BZD doses. Another RCT11 compared melatonin with placebo, each in combination with psychosocial support. The two cluster RCTs2, 12 compared the combination of patient education and BZD dose-tapering with usual care. In one of the cluster RCTs,2 the intervention also included a fortnightly follow-up visit or written instructions. In another RCT,13 CBT was compared with either BZD tapering alone or the combination of tapering plus relaxation.
One non-randomized study16 compared pregabalin alone with pregabalin plus other drugs (details of the other drugs was not provided). Another non-randomized observational study17 evaluated the effect of GP letters for the discontinuation of BZD. In a third non-randomized study,15 patients undergoing a gradual reduction of BZD dose had the option of pharmacological support with either hydroxyzine or valerian when needed.
Outcomes
The most common reported primary outcomes were complete discontinuation or reduction of BZD use at the end of the study. One of the systematic reviews9 assessed the odds of not using BZD over short (0.5 to 3 months) and long-term (12 months) periods. Adverse events (mainly withdrawal symptoms) were also commonly reported.
Summary of Critical Appraisal
Appendix 3 provides further details of the critical appraisal of individual studies.
All the included studies had well-defined objectives and generally well-described inclusion and exclusion criteria. Two systematic reviews9, 10 were based on comprehensive literature searches However, a more limited literature search was performed in the other systematic review as only one electronic database was searched.8. In each systematic review, multiple reviewers independently screened and selected studies for inclusion, and extracted data. In two systematic reviews,9, 10 multiple reviewers independently evaluated the quality of included studies.
Baseline characteristics were generally similar for the study arms of all included RCTs and analysis of outcomes were based on intention-to-treat populations. Four RCTs2, 11, 12, 14 were adequately powered to detect relevant differences in outcomes between treatment groups. However, one RCT13 did not conduct a sample size calculation. Therefore, with a relatively small number (n=47) of participants split among three treatment groups, it was uncertain whether the study was sufficiently powered to detect significant differences in outcomes. One RCT14 involved participants with diagnoses of schizophrenia or bipolar disorders, and another RCT12 restricted participation to patients 65 years of age or older. A third RCT13 involved only patients who were seeking treatment for BZD discontinuation and were therefore likely to be more motivated than the general BZD user population. A fourth RCT2 included only patients who were free from severe medical condition and excluded patients with major depressive or anxiety disorder, or currently receiving psychiatric treatment. Thus, the generalizability of findings from these studies is unknown. One of the RCTs did not use a standardized BZD tapering strategy and the details of the individualized schedules were not provided. Therefore, the comparative effectiveness of the individual withdrawal strategies is indeterminate.
The non-randomized studies15–17 have higher potential for bias due to the absence of randomization to limit differences at baseline to chance, and to permit differences in outcome to the effects of the intervention alone. One of the non-randomized studies16 had a high proportion (47%) of patients with multiple substance abuse disorders which could confound the reported outcomes. Another study17 analyzed long-term (10 years) follow-up data of a discontinuation intervention with limited patient information for the period in-between. Therefore, the possibility of the reported outcomes resulting from influences other than the intervention cannot be ruled out. Although adjunctive pharmacotherapy with hydroxyzine or valerian was permitted on an as needed basis in another non-randomized study15, there were no data or analyses to assess the contribution of adjuvant intervention to the reported outcomes. Thus, it is unknown whether the reported findings were due to dose-tapering, which was the intervention being assessed, or whether the support of the pharmacotherapy had an impact on successful outcomes.
Summary of Findings
A total of 11 studies (three systematic reviews, five RCTs, and three non-randomized studies) met the inclusion criteria of this report and were included.2, 8–17 All the studies assessed interventions for BZD discontinuation among adult long-term users. No evidence-based clinical guidelines on BZD discontinuation were identified. Most of the studies (9 of 11) involved BZD dose-tapering as standalone or as background to other interventions. Two RCTs11, 14 compared the effect of melatonin to placebo when used as adjunct therapy, while one observational study16 evaluated pregabalin as adjunct to BZD dose-tapering. The remaining studies had non-pharmacologic interventions. Further details of findings of individual studies have been provided in Appendix 4.
Rapid Response reports are organized so that the evidence for each research question is presented separately.
What is the clinical evidence regarding strategies to safely and effectively discontinue adult patients from long-term benzodiazepine use?
One systematic review10 and one observational study17 reported that simple interventions such as discontinuation letters from clinicians, self-help information and a single consultation with a GP to discuss risk of long-term BZD use and the benefit of discontinuation are effective intervention strategies to discontinues BZD use in adult long-term users. The systematic review10 found that patients who received such interventions were twice as likely to completely withdraw from BZD use (relative risk [RR] = 2.3; 95% CI: 1.3 to 4.2, P = 0.008) or reduce BZD use (RR = 2.04; 95% CI: 1.5 to 2.8, P<0.001). The observational study17 found that abstinence was sustained among 58.8% patient who discontinued BZD use following discontinuation letters from their GP.
Two systematic reviews8, 9 and one RCT13 assessed the effect of CBT for long-term BZD use. One systematic review9 found that the odds of discontinuing BZD use was highest among patients treated with supervised withdrawal and psychotherapy compared with usual care, or other prescribing interventions (e.g. medication review, consultation or education) (odd ratios [OR] = 5.06; 95% CI: 2.68, 9.57; P <0.00001, number need to treat [NNT] = 3). The other systematic review8 also found that a combination of CBT and BZD dose-tapering resulted in higher BZD discontinuation rates (65% to 85%) compared with dose-tapering alone (25% to 54%). The RCT13 reported that at 6-month follow-up, CBT had a higher BZD discontinuation rate (62.5%) than individualized relaxation therapy (12.5%) and BZD dose-tapering (26.7%) interventions.
Two cluster RCTs2, 12 assessed adjunctive educational interventions. One RCT12 reporting that the likelihood of achieving discontinuation of long-term BZD was significantly increased using patient empowerment education in combination with dose-tapering compared with usual care (OR 8.3 [95% CI: 3.3, 20.9]). In the other RCT,2 a higher discontinuation rate was reported for either structured individualized education with dose-tapering and follow-up visits (45%), or structured individualized education with dose-tapering and written instructions every two weeks (45%),when compared with usual care (15%) among long-term BZD users.
Three included RCTs11, 14, 16 assessed the effectiveness of pharmacotherapy for discontinuation of long-term BZD use. One RCT14 found no significant difference in BZD discontinuation rates between prolonged-release melatonin and placebo among long-term BZD users undergoing slow dose-tapering after 24 weeks of treatment (38.1% versus 47.7%, respectively; OR = 0.64; 95% CI: 0.26, 1.56). Another RCT11 reported that controlled-release melatonin (CRM) with dose tapering, resulted in a higher discontinuation rate than placebo with dose tapering (85% versus 67%) among long-term BZD users after one month. After 6 months, 30.4% of participants in the CRM group and 43.5% in the placebo group remained non-users of BZD. The difference was not statistically significant in either analysis. A prospective, uncontrolled, observational study16 reported that 52% (95% CI: 46%, 58%) of patients who used pregabalin as adjunctive treatment to BZD dose-tapering achieved a BZD-free status after 12 weeks.
One before-after pseudo-experimental study15 assessing gradual reduction of BZD dose with the option of pharmacological support with hydroxyzine (25 mg/day) or valerian when needed, reported that 80.4% of the patients had discontinued BZD by the end of the 6-month intervention, and 64% maintained abstinence at one year.
What are the evidence-based guidelines regarding discontinuation of long-term benzodiazepine use?
The literature search for this review did not find any literature on evidence-based guidelines regarding discontinuation of long-term BZD use.
Limitations
Benzodiazepine dose-tapering was the most common intervention in the included studies. However, in many cases the details about the number of dose reductions and time schedules were not provided. Further details of the limitations of the individual included studies have been provided in Appendix 3.
A major limitation of two of the systematic reviews9, 10 is the small number of included primary studies. A total of 16 RCTs were included in one systematic review,9 however a variety of interventions were considered resulting in a relatively small number of studies per intervention (ranging from 1 to 4 studies). Another systematic review8 had 28 primary studies of heterogeneous designs, and the methodological quality of each were not assessed.
In one included RCT,14 participants were patients diagnosed with schizophrenia or bipolar disorder, which make the generalizability of its findings in other patient populations uncertain. Another RCT11 limited participation to patients who used one particular BZD (temazepam) or two other drugs (zolpidem or zopiclone) which are technically not BZDs but have similar short-term effects. Therefore, it is uncertain whether the reported results from this RCT11 will be reproducible using other BZDs. In one RCT,13 there was no indication that a sample size calculation was performed, and included a relatively small sample size (n=47) across three treatment arms. Thus, it is uncertain if it was sufficiently powered to detect statistically significant differences in clinical outcomes. One cluster RCT12 had only 30 (18%) eligible community pharmacies and 303 (11%) patients agreeing to participate. Reasons for declining participation included lack of interest, competing priorities, inability to obtain consent from owners, and inadequate staff. Although there does not seem to be selection bias, the low level of participation raises concern about how adequately representative the findings are of the communities under study. In the other cluster RCT,2 only 34% of eligible patients participated. Therefore, it is unknown whether the study participants and, by extension, the reported outcomes in these studies2, 12 were sufficiently representative populations.
For the non-randomized studies,15–17 the absence of randomization increased the potential for bias due to differences in potential confounders in study participants; and the possibility of variable influences besides the studied interventions to contribute significantly to the reported outcomes cannot be ruled out. One study16 included an intervention consisting of a combination of pregabalin and other drugs, but failed to provide further details regarding the dose or type of drugs that were included. Thus, the interpretation of comparative effectiveness from this study is limited.
