Comparative effectiveness and safety of pharmacological and non-pharmacological interventions for insomnia: an overview of reviews

Insomnia is a common disorder in the general population. While precise estimates vary, multiple population-based studies in different countries have consistently found that approximately one third of adults (> 18 years of age) reported dissatisfaction with their sleep and at least one symptom of insomnia [1, 2] and 6–10% of the adult population met stricter criteria for a diagnosis of insomnia such as the Diagnostic and Statistical Manual of Mental Disorders (DSM-5) [3] or International Classification of Sleep Disorders (ICSD) [4]. Insomnia can contribute to significant functional impairments at work or at home and is linked to reduced quality of life, problems with attention and memory, mood disturbances, and reduced ability to carry out normal daily activities [5]. Furthermore, studies have indicated that insomnia may be an important risk factor for the onset of mental health disorders such as depression, anxiety, and substance abuse [5].

Clinical practice guidelines published in the USA, Canada, and Europe unanimously recommend that non-pharmacological approaches, especially cognitive behavioral therapies, should be the first-line treatment for chronic insomnia (symptoms for > 3 months) and that pharmacological treatment should only be used in acute cases (< 3 months) or as a short-term supplement to non-pharmacological approaches [6‐8]. Evidence for over-the-counter (e.g., diphenhydramine) or natural remedies (melatonin, valerian) is considered weak or inconclusive, and these approaches are not recommended for acute or chronic insomnia [6‐8]. Despite this, the rate of prescription sleep aid use, particularly non-benzodiazepines and off-label use of antidepressants, has risen significantly over the last 20 years [9‐11], in some cases outpacing the diagnosis of sleep disorders among the general population [10]. Furthermore, a large prospective study of former and current insomnia sufferers found that 70% of patients using a prescription sleep aid continued to do so at 1-year follow-up but did not demonstrate significant improvements in sleep compared to non-users [12]. The use of non-prescription sleep aids is also common alongside prescription drugs; up to 60% of sleep aids used by adults with insomnia are non-prescription [12, 13].

Evidence is needed to support the development of guidelines that encourage the appropriate use of pharmacological interventions to treat insomnia and increase access to and uptake of non-pharmacological approaches. The objective of this overview of systematic reviews was to assess what has been established regarding the clinical effectiveness and safety of pharmacological and non-pharmacological interventions in adults with insomnia and identify areas where further research or policy development is needed.

This comprehensive overview of reviews included 64 systematic reviews representing 358 unique primary studies and found consistent evidence of effectiveness for both pharmacological and non-pharmacological interventions based on data from moderate to high quality SR + MAs. There was evidence of effectiveness across multiple outcomes reported in more than one high- or moderate quality SR + MA for zolpidem, suvorexant, doxepin, and melatonin, and evidence of effectiveness across multiple outcomes reported in one high-quality SR + MA for temazepam, triazolam, zopiclone, and trazodone. Additionally, the evidence for these interventions included reviews rated as having a high (melatonin) or medium (temazepam, triazolam, zolpidem, zopiclone, suvorexant, doxepin, and trazodone) strength of evidence based on GRADE. However, there was very little harms data available for these interventions. There was little to no evidence of effectiveness or no high- or moderate quality evidence available for flurazepam, quetiapine, or diphenhydramine. Moreover, most interventions were studied in the short term (< 12 weeks) and the primary studies included in the reviews tended to have small sample sizes. The lack of harms data and small study sizes are concerning given that a large proportion of the general population are on these medications. Likewise, there was evidence of effectiveness across multiple outcomes reported in multiple high- or moderate quality SR + MAs for CBT and reported in one high-quality SR + MA for BT; there were no high-quality SR + MAs that examined mindfulness-based or combination therapies. The evidence for these interventions also included reviews rated as a high (CBT) or medium (CBT and behavioral therapy) strength of evidence based on GRADE. The studies that examined CBT and BT were often conducted in the short term, and only one SR + MA examined the effect of online versus in-person CBT, which is an important question for future research given the cost of and difficulties accessing in-person CBT [88].

This overview of reviews identified several evidence gaps in the field of insomnia research, particularly the lack of harms data for pharmacological interventions, the effects of different doses, the effectiveness of sequencing or combining drug and non-drug interventions, and a dearth of head-to-head studies directly comparing pharmacological or non-pharmacological interventions. Additionally, the clinical significance of symptomatic changes in insomnia is poorly understood and standards that allow researchers to interpret whether a statistically significant change translates to a clinically significant one are needed (e.g., the minimal clinically important difference).

There are limitations of the included systematic reviews worth noting, particularly the low quality of the included evidence with more than 50% of the included reviews receiving a low- or critically low-quality score on the AMSTAR2 tool. This suggests that substantial improvements in the methods used to synthesize knowledge in this field are needed and that current results should be interpreted with caution. Systematic reviews in this field could be improved by increasing the use of a priori protocols, providing a rationale for including or excluding certain study designs, providing a list of excluded studies with reasons for exclusion, and transparently reporting the funding sources of primary studies included in the review.

There are also some limitations to the conduct of this overview that should be taken into consideration. Due to time and resource constraints, targeted searches for primary studies reporting harms outcomes could not be conducted, which is a deviation from our original protocol [14]. Additionally, although the literature search attempted to find unpublished research and reviews in multiple languages, only one unpublished review and 2 reviews in languages other than English were identified, suggesting that these results are not generalizable beyond systematic reviews published in English. Additionally, the definition of inactive controls used in this overview included standard care interventions such as sleep hygiene and patient education, which may have resulted in underestimation of the effectiveness of some of the non-drug interventions as they were largely compared with these types of controls rather than true control conditions such as placebo or sham interventions. Also, the behavioral, mindfulness and cognitive behavioral interventions included in this review were categorized as reported by review authors. In the interest of capturing a comprehensive evidence base, we did not put any limitations on the eligibility of these interventions leading to a high degree of variability across the reviews. Finally, as stated previously, due to a lack of clinical standards for interpretation, none of the changes in outcomes reported here could be evaluated in terms of their clinical or symptomatic relevance.

There are several strengths of this overview that are worth noting, particularly the use of the Cochrane handbook [18] and an a priori protocol to guide the conduct of the overview, as well as the use of the AMSTAR2 [22] tool for quality appraisal. The literature search was comprehensive and included both published and unpublished sources of information and had no restrictions on publication date or language of publication. The final list of eligible interventions and outcomes was developed in consultation with project stakeholders and clinical experts who were consulted throughout the overview process. Finally, the 64 included systematic reviews were closely examined for overlaps in the primary evidence which was found to be extensive and which we clearly highlighted throughout the “Results” section.

Based on the results of this overview, clinicians and patients with insomnia can consider CBT as a first-line intervention due to its consistent evidence of effectiveness and a high strength of evidence across multiple outcomes and because it is likely associated with few or no serious harms though there is insufficient evidence to properly evaluate the benefit to harm ratio for this intervention. If CBT is not effective, then other behavioral interventions can be considered or short courses of melatonin, zolpidem, suvorexant, or doxepin can be added to non-pharmacological therapy. However, these agents have only been tested in short-term studies and there is little evidence for their effectiveness or safety beyond 16 weeks of treatment.