Comparative effectiveness of interventions to facilitate deprescription of benzodiazepines and other sedative hypnotics: systematic review and meta-analysis

Abstract

Objective: To review evidence from randomised trials assessing the effectiveness of strategies to deprescribe benzodiazepines and closely related sedative hypnotics (BSH).

Design: Systematic review and meta-analysis of randomised controlled trials.

Data sources: MEDLINE, Embase, CINAHL, PsycInfo, and CENTRAL, searched from inception to August 2024, and reference lists of included studies and similar systematic reviews.

Eligibility criteria for selecting studies: Eligible studies randomised adults using BSH for insomnia to interventions aimed at deprescribing BSH, strategies to implement these interventions in healthcare settings, or usual care or placebo.

Methods: Reviewers worked independently and in duplicate to screen search results, extract data, and assess risk of bias. Similar interventions were grouped together, frequentist random effects meta-analysis was conducted, and the certainty of evidence was assessed using the GRADE (Grading of Recommendations Assessment, Development, and Evaluation) approach.

Results: The review identified 58 publications reporting on 49 unique trials with more than 39 000 patients. Interventions were classified into the following categories: tapering, patient education, physician education, combined patient and physician education, cognitive behavioural therapy, medication review, mindfulness, motivational interviewing, pharmacist led interventions, and drug assisted tapering and withdrawal. Low certainty evidence suggests that education of patients (144 (95% confidence interval 61 to 246) more per 1000 patients), medication review (104 (34 to 191) more), and a pharmacist led educational intervention (491 (234 to 928) more) may increase the proportion of patients who discontinue BSH compared with usual care. Moderate certainty evidence suggests that education of patients probably has little or no effect on physical function, mental health, and signs and symptoms of insomnia. No evidence was found regarding these other outcomes for medication review or for the pharmacist led educational intervention. No compelling evidence was found that other interventions may help patients to discontinue BSH. Moreover, no high or moderate certainty evidence was found that any of the interventions caused an increase in dropouts. Finally, low certainty evidence suggests that multicomponent interventions may be more effective at facilitating discontinuation of BSH than single component interventions.

Conclusion: The evidence on the effectiveness of interventions to discontinue BSH is of low certainty. Educating patients, doing medication reviews, and a pharmacist led educational intervention may increase the proportion of patients who discontinue BSH.

Fig 1

Study selection

Fig 2

Risk of bias of trials that compared tapering with usual care

Fig 3

Summary of results of interventions excluding drug assisted tapering. *SF-36 physical functioning scale (range 0-100 with higher scores indicating better function). †Hospital Anxiety and Depression Scale anxiety subscale (range 0-21 with lower scores indicating better mental health). ‡Hospital Anxiety and Depression Scale depression subscale (range 0-21 with lower scores indicating better mental health). §SF-36 physical component score (range 0-100 with higher scores indicating better function). ¶SF-12 mental component score (range 0-100 with higher scores indicating better mental health). **Insomnia Severity Index (range 0-28 with lower scores indicating fewer signs and symptoms of insomnia). ††SF-36 mental component score (range 0-100 with higher scores indicating better mental health). CI=confidence interval

Fig 4

Summary of results for drug assisted tapering and withdrawal interventions. *Hamilton Rating Scale for anxiety (range: 0 to 56 with lower scores indicating less anxiety). †Hamilton Rating Scale for depression (range: 0 to 56 with lower scores indicating less depression). CI=confidence interval