NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Brasure M, MacDonald R, Fuchs E, et al. Management of Insomnia Disorder [Internet]. Rockville (MD): Agency for Healthcare Research and Quality (US); 2015 Dec. (Comparative Effectiveness Reviews, No. 159.)

Cover of Management of Insomnia Disorder

Management of Insomnia Disorder [Internet].

Show details

Results

Literature Search and Screening

Key Points

  • Global outcomes were less often measured than sleep outcomes, especially in the drug studies; recent research was more likely to assess global outcomes.
  • Minimum important differences were identified for some instruments used to assess global outcomes, but these were not frequently used nor is it clear whether they are well established. We did not identify established minimum important differences for most sleep outcomes. Remission defined using sleep onset latency and sleep efficiency were the exceptions.
  • A large body of literature tests a wide variety of treatments for insomnia disorder. Strength of evidence suffers because of limited studies with similar comparisons. In addition, sample sizes were typically small and studies often contained multiple arms. Older studies often did not provide data sufficient for analysis.

Our search identified 3572 citations, of which 540 required full text review after title and abstract screening (Figure 2). Of the 540 full text articles screened, we identified 102 eligible references; we identified another 32 eligible references by hand searching for a total of 133 publications of 128 unique RCTs and 3 unique systematic reviews. Systematic reviews included in our analysis synthesized evidence on 41 unique RCTs. Studies excluded after full text review are listed in Appendix C along with exclusion reasons. The most frequent exclusion reasons included a lack of randomization, inadequate study duration, drugs not approved for use in the United States, insomnia not clinically diagnosed, and not available in English. Studies not available in English were often complementary and alternative medicine (CAM) treatments published only in Chinese. We captured results of many of these studies by including systematic reviews (that did not have language restrictions) in lieu of de novo extraction. We also searched for observational studies to supplement our harms discussion, and identified nine observational studies (for medication adverse effects) that met inclusion criteria.

Figure 2 is the literature flow diagram, which is also described in the text. It enumerates articles retrieved through the literature search, articles excluded, and articles retrieved through hand searching.

Figure 2

Literature flow diagram. RCT = randomized controlled trial; SR = systematic review

Evidence tables including study characteristics and outcomes for all included studies are available upon request and will be uploaded to the Systematic Review Data Repository after the final version of this report is posted. AMSTAR ratings, risk of bias assessments, and strength of evidence assessments appear in Appendix D for psychological interventions; Appendix E for pharmacologic interventions; Appendix F for CAM interventions; and Appendix G for combination or comparative effectiveness of interventions across intervention types.

Efficacy and Comparative Effectiveness of Psychological Interventions

Key Points

  • CBT-I across several delivery modes improves global and sleep outcomes compared with passive control in the general adult population (moderate strength evidence). Evidence was insufficient to assess adverse effects of CBT-I.
  • CBT-I across several delivery modes improves global and several sleep outcomes (sleep onset latency, wake time after sleep onset, and sleep efficiency) compared with passive control among older adults with insomnia disorder (low to moderate strength evidence). Sleep outcomes remain improved long term (low strength evidence).
  • CBT-I across several delivery modes improves global and several sleep outcomes (sleep onset latency, total sleep time, wake time after sleep onset, and sleep efficiency) compared with passive control among adults with pain conditions and insomnia disorder (low strength evidence)
  • Multicomponent behavioral therapy and/or BBT improves several sleep outcomes (sleep onset latency, wake time after sleep onset, and sleep efficiency) in older adults with insomnia disorder (low strength evidence).
  • Data on the efficacy of specific cognitive or behavioral interventions alone (stimulus control, sleep restriction, relaxation techniques) were limited and evidence was insufficient to draw conclusions.
  • Evidence was insufficient to assess adverse effects of any psychological treatments.

Efficacy of Cognitive Behavioral Therapy in the General Adult Population

Overview of Studies

We included studies as efficacy of CBT-I if they had an active CBT-I arm and passive control arm (sham treatment/placebo, wait-list control, no treatment, or sleep hygiene/sleep education). We identified 20 RCTs with acceptable risk of bias assessing the efficacy of CBT-I to treat insomnia disorder in the general adult population.51-72 Trials were conducted in the United States,51,54-56,60,67 Sweden,53,61,62,68 Canada,64,66,71 the Netherlands,63,69,70 the United Kingdom,58,59,65 Norway,52 Scotland,57 and China.72 Studies differed in how CBT-I was delivered. Six studied individual in-person CBT-I,54-56,60,62,65 five studied group CBT-I,53,57,59,61,72 one studied phone-delivered CBT-I,51 and ten studied self-help CBT-I using either books or handouts or electronic resources.52,58,63,64,66-71 Comparison groups also varied across trials. Some trials attempted to have a placebo control group that received similar therapy hours (i.e., quasi-desensitization or stress management).54-56,58,60,72 Enrollment criteria varied across trials, most relied on the DSM-IV criteria for enrollment. The mean age was typically in the mid-40s; participants were predominantly female, and most were white (in the trials that reported race). Insomnia duration ranged from an average of 6 months to nearly 2 decades with most trials reporting mean duration of several years. Baseline ISI scores were just over 17 and baseline sleep onset latency was over 45 minutes. Interventions were typically once a week for 1 hour or less and lasted from 4 to 6 weeks. Eighteen of these RCTs (n=1842) provided data sufficient for pooling on one or more outcomes (Table 5). Risk of bias of included trials was predominantly medium.

Table 5. Overview and strength of evidence: efficacy of CBT-I in the general adult population.

Table 5

Overview and strength of evidence: efficacy of CBT-I in the general adult population.

Global Outcomes

Four small trial assessed insomnia remitters (achieving an ISI score ≤7 or PSQI score ≤5 at followup) (Figure 3).51,54,56,62 Two small studies used the “ISI score ≤7” definition of remission. Edinger compared individual CBT-I to placebo.54,56 Jernelov et al. compared self-help CBT-I with therapist support to waitlist controls62 and Arnedt et al. compared CBT-I delivered by phone with a control group who received sleep hygiene information.51 Both Arnedt and Edinger had small sample sizes and failed to reach statistical significance. Pooled results show that CBT-I participants are nearly three times more likely to achieve remission than passive controls; 43 percent more CBT-I participants achieved remission compared with passive controls; and just over two individuals with insomnia would need to be treated to see one achieve remission. Two trials assessed ‘responders’ to treatment according to established thresholds (Figure 4).51,62 The pooled result was not statistically significant.

Figure 3 is a forest plot displaying data presented in Table 5 about the efficacy of CBT-I in the general adult population for the outcome remission from insomnia disorder. When data from four studies were pooled, the risk ratio for remission was 2.95 in the CBT-I group compared with a passive control (significant).

Figure 3

Efficacy of CBT-I in the general adult population: remitters. CBT-I = cognitive behavioral therapy – insomnia; CI = confidence interval; M-H = Mantel-Haenszel

Figure 4 is a forest plot displaying data presented in Table 5 about the efficacy of CBT-I in the general adult population for the outcome response to treatment. When data from two studies were pooled, the risk ratio for response was 3.24 in the CBT-I group compared with a passive control (not significant).

Figure 4

Efficacy of CBT-I in the general adult population: responders. CBT-I = cognitive behavioral therapy – insomnia; CI = confidence interval; M-H = Mantel-Haenszel

Six of the CBT-I efficacy trials across four delivery methods reported mean ISI scores (Figure 5).51,53,62,66,67,71 Only one trial achieved a weighted mean change in ISI scores equal or greater than the minimum important difference of seven.67 The pooled estimate shows that CBT-I across delivery methods achieves a 5.15 point reduction in ISI scores suggesting that many patients will realize important benefits from CBT-I… Studies across four delivery methods reported mean PSQI scores (Figure 6).51,52,56,59,64,66 The pooled estimate showed that CBT-I across delivery methods achieved a 2-point reduction in PSQI scores. We did not identify literature suggesting a minimum important difference, so it is unclear how this change should be interpreted.

Figure 5 is a forest plot displaying data presented in Table 5 about the efficacy of CBT-I in the general adult population for the outcome ISI scores. When data from six studies were pooled, the mean ISI score was 4.48 lower (better) in the CBT-I group than in a passive control (significant).

Figure 5

Efficacy of CBT-I in the general adult population: ISI mean score. CBT-I = cognitive behavioral therapy – insomnia; CI = confidence interval; ISI = Insomnia Severity Index; IV = inverse variance; SD = standard deviation

Figure 6 is a forest plot displaying data presented in Table 5 about the efficacy of CBT-I in the general adult population for the outcome PSQI scores. When data from seven studies were pooled, the mean PSQI score was 2.14 lower (better) in the CBT-I group than a passive control (significant).

Figure 6

Efficacy of CBT-I in the general adult population: PSQI scores. CBT-I = cognitive behavioral therapy – insomnia; CI = confidence interval; PSQI = Pittsburgh Sleep Quality Index; IV = inverse variance; SD = standard deviation

One last global outcome was evaluated in CBT-I efficacy trials, clinical global impression (CGI). Vincent, et al., showed that clinicians reported that participants enrolled in web-based CBT-I were at eight times higher odds of being “much or very much improved” compared with passive controls.71

One trial provided evidence that global outcomes remain improved at 6 month after treatment initiation.

Sleep Outcomes

Most CBT-I efficacy trials reported patient-reported sleep outcomes (Figures 7-9). Improvements in sleep onset latency differed significantly from passive control in only six of the 16 trials that reported poolable data (Figure 7). Pooled data show that the largest improvements in sleep onset latency occurred with CBT-I group compared with self-help CBT-I. However, this was due to a very large effect in two trials reporting mean decrease in sleep onset latency of over 30 minutes.58,72 The pooled estimate across all delivery methods shows that CBT-I participants reduced their sleep onset latency by nearly 12 minutes compared with passive controls. However, trials that had sham treatment placebo controls were less likely to show a significant improvement over placebo. Pooled estimates show that CBT-I participants gained a mean of 15 minutes of total sleep time. Reductions in wake time after sleep onset were demonstrated in five of 12 trials reporting this outcome across four delivery methods. The pooled estimate shows that CBT-I participants decreased their mean awake time after sleep onset by nearly 22 minutes.

Figure 7 is a forest plot displaying data presented in Table 5 about the efficacy of CBT-I in the general adult population for the outcome SOL. When data from 16 studies were pooled, the mean SOL was 11.88 minutes shorter in the CBT-I group than a passive control (significant).

Figure 7

Efficacy of CBT-I in the general adult population: sleep onset latency. CBT-I = cognitive behavioral therapy – insomnia; CI = confidence interval; IV = inverse variance; SD = standard deviation

Figure 8 is a forest plot displaying data presented in Table 5 about the efficacy of CBT-I in the general adult population for the outcome TST. When data from 16 studies were pooled, the mean TST was 15.01 minutes longer in the CBT-I group than a passive control (significant).

Figure 8

Efficacy of CBT-I in the general adult population: total sleep time. CBT-I = cognitive behavioral therapy – insomnia; CI = confidence interval; IV = inverse variance; SD = standard deviation

Figure 9 is a forest plot displaying data presented in Table 5 about the efficacy of CBT-I in the general adult population for the outcome WASO. When data from twelve studies were pooled, the mean WASO was 21.89 minutes shorter in the CBT-I group than a passive control (significant).

Figure 9

Efficacy of CBT-I in the general adult population: wake time after sleep onset. CBT-I = cognitive behavioral therapy – insomnia; CI = confidence interval; IV = inverse variance; SD = standard deviation

Post-intervention sleep efficiency improved with CBT-I in 9 of 16 trials. Mean sleep efficiency at endpoint ranged from 72 to 88 among CBT-I participants and from 64 to 85 among passive controls across the nine trials. The pooled estimate shows that sleep efficiency improved by over six percentage points in CBT-I participants compared with passive controls across six delivery methods. This measure should increase with CBT-I compliance, which often suggests that participants get out of bed if they can't get to sleep, so may reflect compliance as well as efficacy.

Sleep quality improved in 6 of 11 trials reporting sleep quality. In-person CBT-I appears to have larger responses. Several self-help CBT-I trials failed to show efficacy. The pooled estimate of the standardized mean difference suggests that CBT-I creates a moderately sized improvement on sleep quality across delivery methods.

Four trials provided evidence that sleep onset latency and wake after sleep onset remain improved at time periods beyond 6 months from treatment initiation.

Functioning, Mood, and Quality of Life

Several trials mentioned functioning, mood, and quality of life using several different instruments. Most studies were small and few studies used similar instruments.

Adverse Effects

Specific adverse effects were not reported. Most trials reported withdrawals or loss to followup. Two studies did not report withdrawals or loss to followup by treatment group.57,61 No statistically significant differences were found across groups in the rates of withdrawals or loss to followup.

Efficacy of Cognitive Behavioral Therapy in Older Adults

Overview of Studies

We included studies as efficacy of cognitive behavioral therapy (CBT) if they had an active CBT-I arm and a passive control arm (placebo, wait-list control, no treatment, or sleep hygiene/sleep education). We analyzed studies that enrolled older adults separately from those enrolling adults of all ages. We identified six trials that compared CBT-I with passive control in older adults.73-77 Risk of bias for included studies was predominantly moderate. We pooled evidence on common outcomes when possible (Table 6).

Table 6. Overview and strength of evidence: efficacy of CBT-I in older adults.

Table 6

Overview and strength of evidence: efficacy of CBT-I in older adults.

Global Outcomes

Three studies reported global outcomes (Figures 10 and 11).73,76,77 All trials did not report the same instrument scores measured the same way. For instance, all three reported PSQI, but Rybarczyk et al. and Irwin et al. reported total scores and Morgan et al. reported mean changes in scores, so data could not be pooled. All trials showed statistically significant changes in global outcomes. Morgan et al. found statistically significant difference between the mean change on the ISI and the mean change on the PSQI at three time points.73 Global outcomes were improved after CBT-I in older adults and improvements are sustained at 3 and 6 month followup. However mean difference in change between groups or mean change from baseline for the ISI did not achieve the minimum clinical difference of seven points (Figure 10). Similar improvements were demonstrated with the PSQI and the Athens Insomnia Scale; however, clinical significance of the difference in mean change is unclear (Figure 11).

Figure 10 is a forest plot displaying data presented in Table 6 about the efficacy of CBT-I in older adults for the outcome ISI scores. In a single study, mean ISI scores decreased (improved) more in the CBT-I group than comparison group by 3.60 over the mean follow-up time (significant), 2.10 at 3 months (significant), and 3.30 at 6 months (significant).

Figure 10

Efficacy of CBT-I in older adults: ISI. CBT-I = cognitive behavioral therapy – insomnia; CI = confidence interval; ISI = insomnia severity index; IV = inverse variance; SD = standard deviation

Figure 11 is a forest plot displaying data presented in Table 6 about the efficacy of CBT-I in older adults for the outcomes Athens Insomnia Scale and PSQI scores. In a single study, the mean Athens Insomnia Scale was 3.40 lower (better) in the CBT-I group than a passive control at 6 months and 2.20 lower at 16 months (both significant). When data from two studies were pooled, mean PSQI scores were 2.98 lower (better) in the CBT-I group than a passive control at 4 months (significant).

Figure 11

Efficacy of CBT-I in older adults: Athens Insomnia Index and PSQI. CBT-I = cognitive behavioral therapy – insomnia; CI = confidence interval; PSQI = Pittsburgh Sleep Quality Index; IV = inverse variance; SD = standard deviation

Sleep Outcomes

Sleep outcomes were reported in all CBT-I efficacy trials among older adult participants (Figures 12 and 13). One trial attempted to measure the proportion of participants who achieved clinically significant improvement in sleep.78 It defined clinically significant improvement as the attainment of sleep efficacy equal to or greater than the mean in a group of patients without insomnia. More CBT-I participants achieved clinical improvement than passive controls

Figure 12 is a forest plot displaying data presented in Table 6 about the efficacy of CBT-I in older adults for the outcome SOL. When data from three studies were pooled, the mean SOL was 9.98 minutes shorter in the CBT-I group than a passive control at 6 months (significant).

Figure 12

Efficacy of CBT-I in older adults: sleep onset latency. CBT-I = cognitive behavioral therapy – insomnia; CI = confidence interval; IV = inverse variance; SD=standard deviation

Figure 13 is a forest plot displaying data presented in Table 6 about the efficacy of CBT-I in older adults for the outcome WASO. When data from four studies were pooled, the mean WASO was 26.96 minutes shorter in the CBT-I group than a passive control at 6 months (significant).

Figure 13

Efficacy of CBT-I in older adults: wake time after sleep onset. CBT-I = cognitive behavioral therapy – insomnia; CI = confidence interval; IV = inverse variance; SD = standard deviation

Sleep onset latency was reported in three trials.75-77 Differences between groups were significant in only one individual trial and in the pooled analysis. CBT-I led to a decrease of 10 minutes in sleep onset latency. Only one trial reported sleep onset latency at a followup point over 6 months showing no statistical difference between the CBT-I group and passive control group.

No differences were reported in any of the three trials reporting total sleep time at followup.74-77 Pooled analysis was also insignificant post-treatment and at 1- and 2-year followup. Results for two other sleep outcomes were more promising. Wake time after sleep onset was reported in four trials.74-77 Statistically significant reductions were shown in each individual study as well as with the pooled result. CBT-I participants reduced their wake time after sleep onset by nearly 27 minutes. One study showed that this result was maintained at 1-year and not 2-year followup. A similar pattern was demonstrated with sleep efficiency. The pooled analysis demonstrates that the CBT-I group increased their sleep efficiency by about 9 percentage points at followup. In Morin et al. sleep efficiency increased by 18 percentage points at 1-year followup, pooled results showed that sleep efficiency was nearly 8 percentage points higher at 2 years.

Functioning, Mood, and Quality of Life

Morgan et al. reported Fatigue Severity Scores for both groups and found no statistically significant differences post-treatment, or at 3- or 6-month followup.73

Adverse Effects

Most trials reported withdrawals or adverse effects.73,78,79 CBT-I participants were no more likely to withdraw from a study than participants of passive control groups.

Efficacy of Cognitive Behavioral Therapy in Adults With Pain

Overview of Studies

Seven trials reported in eight publications assessed CBT-I delivered to patients with pain and insomnia (Table 7). Four trials studied the general adult population with pain80-84 and three assessed the efficacy of CBT-I in older adults with pain.79,85,86 McCurry is a second publication of an earlier trial analyzing only a portion of the participants with pain.85

Table 7. Overview and strength of evidence: efficacy of CBT-I in the general adult population with pain.

Table 7

Overview and strength of evidence: efficacy of CBT-I in the general adult population with pain.

Global Outcomes

Four trials reported ISI scores. Three of the four trials show a statistical difference between groups (Table 7). Pooled analysis shows that ISI scores were 7.10 points lower with CBT-I (95% CI -12.87 to -1.27) (Figure 14) suggesting that most patients would see important benefits from treatment. Only one study assessed outcomes after 6 months and found ISI scores still better than passive controls, but the magnitude was smaller, with a mean difference of -3.40 (95% CI -6.25 to -0.55).

Figure 14 is a forest plot displaying data presented in Table 7 about the efficacy of CBT-I in the general adult population with pain for the outcome ISI scores. When data from four studies were pooled, mean ISI scores were 7.10 points lower (better) in the CBT-I group than comparison group up to 6 months (significant).

Figure 14

Efficacy of CBT-I in adults with pain: ISI scores. CBT-I = cognitive behavioral therapy – insomnia; CI = confidence interval; ISI = Insomnia Severity Index; IV = inverse variance; SD = standard deviation

Sleep Outcomes

Three efficacy trials enrolling pain patients reported sleep onset latency (Figure 15). In all three trials, the CBT-I groups decreased sleep onset latency compared with passive control. Whether the control was a placebo or sham treatment or wait-list did not matter. Pooled estimate shows that sleep onset latency decreases by over 26 minutes. Four efficacy trials enrolling pain patients reported total sleep time. Total sleep time was similar across groups in all trials as well as the pooled result with CBT-I. The one trial measuring outcomes beyond 6 months found similar total sleep time with CBT-I and placebo. Three trials reported wake time after sleep onset (Figure 16). All trials showed a statistical improvement with CBT-I; pooled results show that CBT-I is associated with a decrease of nearly 40 minutes in wake time after sleep onset.

Figure 15 is a forest plot displaying data presented in Table 7 about the efficacy of CBT-I in the general adult population with pain for the outcome SOL. When data from three studies were pooled, mean SOL was 26.50 minutes shorter in the CBT-I group than comparison group up to 6 months (significant).

Figure 15

Efficacy of CBT-I in adults with pain: sleep onset latency. CBT-I = cognitive behavioral therapy – insomnia; CI = confidence interval; IV = inverse variance; SD = standard deviation

Figure 16 is a forest plot displaying data presented in Table 7 about the efficacy of CBT-I in the general adult population with pain for the outcome WASO. When data from three studies were pooled, mean WASO was 38.18 minutes shorter in the CBT-I group than comparison group up to 6 months (significant).

Figure 16

Efficacy of CBT-I in adults with pain: wake time after sleep onset. CBT-I = cognitive behavioral therapy – insomnia; CI = confidence interval; IV = inverse variance; SD = standard deviation

Functioning, Mood, and Quality of Life

Several trials reported functioning, mood, or quality of life outcomes.

Adverse Effects

Most trials reported withdrawals or adverse effects.

Efficacy of Cognitive Behavioral Therapy in Other Special Populations

Overview of Studies

Two trials studied CBT-I in other special populations (college students and insomnia patients with hearing-impairment).87,88 Because we have only one small trial with moderate study limitations in each of these special populations, evidence is insufficient to draw conclusions about the efficacy of CBT-I.

Efficacy of Multicomponent Behavioral Interventions in the General Adult Population

Overview of Studies

We identified one trial that assessed the efficacy of multicomponent behavioral interventions in the general adult population. Risk of bias was predominantly moderate.89 Evidence from one small trial was insufficient to draw conclusions regarding the efficacy of multicomponent behavioral interventions in treating insomnia disorder in the general adult population.

Efficacy of Multicomponent Behavioral Interventions or Brief Behavioral Therapy in Older Adults

Overview of Studies

We identified three trials reported in four publications comparing multicomponent behavioral therapies (MBT) or BBT with passive control in older adults.90-93 The trials randomized 146 participants, the mean age was around 70, and the majority of participants were female. In the two trials reporting, participants had mean insomnia duration of 15.3 years. All trials were conducted in the United States.90-93 Two trials randomized participants to MBT/BBT or information control.90-92 In the fourth trial, hypnotic-dependent adults with insomnia were randomized to either MBT or placebo.93 We synthesized outcomes from these studies when possible (Table 8).

Table 8. Efficacy of multicomponent behavioral therapy or brief behavioral therapy in older adults.

Table 8

Efficacy of multicomponent behavioral therapy or brief behavioral therapy in older adults.

Global Outcomes

In the one trial reporting PSQI scores, participants receiving BBT saw a statistically significant difference from the passive control group at followup. Multicomponent behavioral interventions or BBT participants scored an average of 3 points lower on PSQI than passive controls.

Sleep Outcomes

Two of three sleep outcomes improved. Sleep outcomes were reported in all multicomponent behavioral intervention and BBT efficacy trials. Improvements in sleep onset latency favored BBT over passive control in all three trials. The pooled estimate shows that multicomponent behavioral therapies or BBT reduced sleep onset latency by more than 10 minutes over passive control (Figure 17). All three trials reported total sleep time, showing no statistically significant increase when compared with passive control patients. Significant decreases in wake time after sleep onset were demonstrated in two trials (Figure 18).90,92 The pooled estimate shows that multicomponent behavioral therapies or BBT reduced wake time after sleep onset by nearly 15 minutes compared with passive control.

Figure 17 is a forest plot displaying data presented in Table 8 about the efficacy of multicomponent behavioral therapy or brief behavioral therapy in older adults for the outcome SOL. When data from three studies were pooled, mean SOL was 10.43 minutes shorter in the multicomponent or brief behavioral therapy group than comparison group (significant).

Figure 17

Efficacy of multicomponent behavioral or brief behavioral therapy in older adults: sleep onset latency. CI = confidence interval; IV = inverse variance; SD = standard deviation

Figure 18 is a forest plot displaying data presented in Table 8 about the efficacy of multicomponent behavioral therapy or brief behavioral therapy in older adults for the outcome WASO. When data from three studies were pooled, mean WASO was 14.90 minutes shorter in the multicomponent or brief behavioral therapy group than comparison group (significant).

Figure 18

Efficacy of multicomponent behavioral therapy or brief behavioral therapy in older adults: wake time after sleep onset. CI = confidence interval; IV = inverse variance; SD = standard deviation

The pooled estimate shows that older BBT participants increased their sleep efficiency more than 6 percentage points over passive control participants.

Functioning, Mood, and Quality of Life

One trial reported functioning, mood, and quality of life outcomes. Buysse et al. reported the difference in SF-36 scores from baseline to post-treatment (4 weeks). Those in the BBT group reported less disability after 4 weeks (3.85 [SE=1.76]) and those in the passive control group reported more (-2.33 [SE=1.73]).

Adverse Effects

Specific adverse effects were not reported. One trial reported study withdrawals or loss to followup (5%) but neither reported withdrawals or loss to followup by group.90,91

Efficacy of Sleep Restriction in the General Adult Population

Overview of Studies

One trial assessed the efficacy of sleep restriction therapy in the general adult population.94 This study provides insufficient evidence to draw conclusions regarding the efficacy of sleep restriction therapy in the general adult population.

Efficacy of Sleep Restriction in Older Adults

Overview of Studies

We included studies as efficacy of sleep restriction (SR) if they had an active SR arm and passive control arm (wait-list control, no treatment, or sleep hygiene/sleep education). We identified two trials that compared SR to a passive control in older adults (Table 9).95,96 Both trials were conducted in the United States. Studies differed in how the sleep restriction was delivered. The mean age across two studies reporting age was close to 70, the majority of study participants were female, and almost all were white (in the trial that reported race).95 The mean duration of insomnia in the one study which reported it by group was 10.8 years.95 Risk of bias was predominantly moderate.

Table 9. Efficacy of sleep restriction in older adults: overview and strength of evidence.

Table 9

Efficacy of sleep restriction in older adults: overview and strength of evidence.

Global Outcomes

Evidence on global outcomes was insufficient to draw conclusions because only one study reported these outcomes. It found that sleep restriction led to greater proportion of sleep restriction therapy participants to achieve remission (achieving an ISI score ≤7 at followup) and response.95 Sleep restriction participants also achieved better ISI scores compared with passive comparison with a five point improvement in ISI score. This change was lower than the 7 point change associated with “response.”

Sleep Outcomes

Evidence was insufficient to draw conclusions regarding most sleep outcomes. Sleep outcomes were reported in all older adult sleep restriction efficacy trials (Figures 19 and 20). The two trials showed different results resulting in serious inconsistency. Pooled data show a large range in post-intervention sleep onset latency and total sleep time. Due to the large range and heterogeneity, the pooled differences were not statistically significant. Since sleep restriction limits time in bed, it is to be expected that total sleep time would not differ significantly between sleep restriction and comparison groups. Mean sleep efficiency with sleep restriction was not significantly different than those in passive control at followup in one study. Sleep quality was reported in one trial.95 Mean sleep quality of those in the sleep restriction treatment group was significantly higher than those in passive control at followup.

Figure 19 is a forest plot displaying data presented in Table 9 about the efficacy of sleep restriction in older adults for the outcome SOL. When data from two studies were pooled, mean SOL was 11.38 minutes shorter in the sleep restriction group than comparison group (not significant).

Figure 19

Efficacy of sleep restriction among older adults: sleep onset latency. CI = confidence interval; IV = inverse variance; SD = standard deviation

Figure 20 is a forest plot displaying data presented in Table 9 about the efficacy of sleep restriction in older adults for the outcome TST. When data from two studies were pooled, mean TST was 17.57 minutes shorter in the sleep restriction group than comparison group (not significant).

Figure 20

Efficacy of sleep restriction among older adults: total sleep time. CI = confidence interval; IV = inverse variance; SD = standard deviation

Functioning, Mood, and Quality of Life

Epstein et al. reported STAI state anxiety, STAI trait anxiety, and Geriatric Depression Scale scores for both groups and found no statistically significant differences post-treatment on STAI trait anxiety, but found statistically significant improvements in STAI state anxiety and Geriatric Depression Scale scores for the sleep restriction group when compared to passive control.95

Adverse Effects

Specific adverse effects were not reported.

Efficacy of Stimulus Control in the General Adult Population

Overview of Studies

We identified two RCTs that assessed the efficacy of stimulus control to treat insomnia disorder in the general adult population (Table 10).89,97 One was conducted in Australia89 and one in Scotland.97 Studies differed in how the stimulus control was delivered. We pooled data when sufficient data were provided.

Table 10. Efficacy of stimulus control in the general adult population: overview and strength of evidence.

Table 10

Efficacy of stimulus control in the general adult population: overview and strength of evidence.

Global Outcomes

Harris et al. reported mean PSQI scores and showed that scores were higher in the stimulus control group than the placebo group.89

Sleep Outcomes

Poolable data on sleep outcomes were reported in two of the stimulus control efficacy trials (Figures 21 and 22).

Figure 21 is a forest plot displaying data presented in Table 10 about the efficacy of stimulus control in the general adult population for the outcome SOL. When data from two studies were pooled, mean SOL was 31.24 minutes shorter in the stimulus control group than the placebo group (significant).

Figure 21

Efficacy of stimulus control: sleep onset latency. CI = confidence interval; IV = inverse variance; SD = standard deviation

Figure 22 is a forest plot displaying data presented in Table 10 about the efficacy of stimulus control in the general adult population for the outcome TST. When data from two studies were pooled, mean TST was 43.54 minutes longer in the stimulus control group than the placebo group (significant).

Figure 22

Efficacy of stimulus control: total sleep time. CI = confidence interval; IV = inverse variance; SD = standard deviation

Improvements in sleep onset latency and total sleep time were significantly different than passive control in both trials that reported poolable data. Pooled data show that stimulus control decreases sleep onset latency by over 30 minutes and increases total sleep time by nearly 45 minutes when compared with placebo.

Other sleep outcomes were reported in one trial. Harris et al. showed that stimulus control decreases wake time after sleep onset and increases sleep efficiency. However, because these findings are from one small study, evidence is considered insufficient.

Functioning, Mood, and Quality of Life

Neither trial reported functioning, mood, or quality of life outcomes

Adverse Effects

Specific adverse effects were not reported.

Efficacy of Stimulus Control in Older Adults

Overview of Studies

We identified two trials that compared stimulus control with passive control in older adults (Table 11).95,98 One trial was conducted in the United States95 and one was conducted in Canada.98 The mean age across studies reporting age was around 70; most participants were female and 99 percent were white (in the trial that reported race).95 The mean duration of insomnia in the three studies that reported it was 12.7 years. We pooled results when possible (Table 11).

Table 11. Efficacy of stimulus control in older adults: overview and strength of evidence.

Table 11

Efficacy of stimulus control in older adults: overview and strength of evidence.

Global Outcomes

Global outcomes were reported by only one study and are therefore insufficient to draw conclusions regarding efficacy. The same study assessed insomnia remitters (achieving an ISI score ≤8 at followup).95 Stimulus control achieved higher rates of remission compared with passive control. One study reported mean ISI scores.95 Stimulus control resulted in a 5.10 point improvement in ISI score compared with passive control. This difference was less than the 7 points necessary for ‘response’ to treatment.

Sleep Outcomes

Sleep outcomes were reported in all older adult stimulus control efficacy trials. Changes in sleep onset latency were inconsistent across trials (Figure 23). Pooled data show a large range in post-intervention sleep onset latency. Due to the large range and heterogeneity, the pooled difference was not statistically significant. One trial showed a statistically significant difference in total sleep time among stimulus control participants when compared with passive treatment controls, while one did not. Pooled data showed total sleep time improved with stimulus control with a mean increase in total sleep time of over 40 minutes (Figure 24). One trial reporting wake time after sleep onset favored stimulus control. Mean sleep efficiency with stimulus control was significantly different than those in passive control at followup in one study. Sleep quality was reported in one trial. Mean sleep quality of those in the stimulus control treatment group was significantly higher than those in passive control at followup.

Figure 23 is a forest plot displaying data presented in Table 11 about the efficacy of stimulus control in older adults for the outcome SOL. When data from two studies were pooled, mean SOL was 10.36 minutes shorter in the stimulus control group than the comparison group (not significant).

Figure 23

Efficacy of stimulus control among older adults: sleep onset latency. CI = confidence interval; IV = inverse variance; SD = standard deviation

Figure 24 is a forest plot displaying data presented in Table 11 about the efficacy of stimulus control in older adults for the outcome TST. When data from two studies were pooled, mean TST was 40.37 minutes longer in the stimulus control group than the comparison group (significant).

Figure 24

Efficacy of stimulus control among older adults: total sleep time. CI = confidence interval; IV = inverse variance; SD = standard deviation

Functioning, Mood, and Quality of Life

Epstein et al. reported STAI state anxiety, STAI trait anxiety, and Geriatric Depression Scale scores for both groups and found no statistically significant differences post-treatment.95 Morin and Azrin reported that they found no differences between stimulus control and passive control groups on STAI state anxiety, STAI trait anxiety, and Beck Depression Inventory from baseline to followup.98

Adverse Effects

Specific adverse effects were not reported. All trials reported withdrawals or loss to followup; however, not all of them reported by group.

Efficacy of Relaxation Therapy in the General Adult Population

Overview of Studies

We identified two randomized trials comparing relaxation therapy with passive control in the general adult population (Table 12).55,97 Participants had a mean insomnia duration of several years. One trial was conducted in the United States55 and one in the United Kingdom.97 Both trials had a moderate risk of bias. Both trials randomized participants to relaxation therapy or passive control. Espie randomized participants to relaxation therapy, stimulus control, paradoxical intention placebo, or no treatment; only two arms (relaxation therapy and paradoxical intention placebo) are discussed in this section.97 Edinger et al. randomized participants to progressive relaxation or a placebo treatment (quasi-desensitization).55

Table 12. Efficacy of relaxation therapy in the general adult population: overview and strength of evidence.

Table 12

Efficacy of relaxation therapy in the general adult population: overview and strength of evidence.

Global Outcomes

Neither trial reported global outcomes.

Sleep Outcomes

Both trials reported sleep outcomes. The pooled estimate shows that relaxation therapy was similar to placebo in reducing sleep onset latency (Figure 25) and total sleep time.

Figure 25 is a forest plot displaying data presented in Table 12 about the efficacy of relaxation therapy in the general adult population for the outcome TST. When data from two studies were pooled, mean TST was 10.23 minutes longer in the relaxation therapy group than the placebo group (not significant).

Figure 25

Efficacy of relaxation therapy in the general adult population: total sleep time. CI = confidence interval; IV = inverse variance; SD = standard deviation

Functioning, Mood, and Quality of Life

No trials reported functioning, mood, and quality of life outcomes.

Adverse Effects

Specific adverse effects were not reported. All four trials reported withdrawals or loss to followup. None of the studies reported withdrawals or loss to followup by group.

Comparative Effectiveness of Psychological Treatments

Several trials included other comparisons of active interventions that we have not addressed specifically thus far. However, the lack of similar comparisons yields insufficient evidence to draw conclusions about the comparative effectiveness of different psychological interventions.

Three trials compared delivery modes of CBT-I. Bastien et al.99 compared individual-, group-, and telephone-delivered CBT-I. Mimeault et al. included two arms that compared self-help CBT-I to self-help CBT-I with professional guidance.64 Holmqvist et al. compared web-based versus phone-based CBT-I.100 Lancee et al. compared self-help CBT-I with self-help CBT-I with support.101,102 Rybarczyk et al. compared two types of CBT-I (self-help versus therapist led) in older participants, most of whom had comorbidities.103 Pech et al. compared two multicomponent programs (both contained sleep hygiene education, stimulus control, and progressive relaxation; the two groups additionally got either cognitive therapy or problem solving therapy)104 to stress management programs. Rybarczyk et al. randomized older participants to participants to CBT-I or stress management.105 Two studies assessed the adjunctive efficacy of certain components. Jansson-Frojmark et al. assessed the adjunctive efficacy of a constructive worry program to a multicomponent behavioral treatment.106 Riley et al. studied the adjunctive efficacy of behavioral prompts as adjunctive functions in a computer device provided to all participants.107

Efficacy of Pharmacologic Treatment

Key Points

  • Most RCTs were small and of short duration. Minimally important differences were often not established or used. We found no eligible trials for many insomnia treatments and some insomnia pharmacological treatments are not specifically approved for insomnia disorders.
  • Evidence from RCTs indicates that some pharmacologic interventions improve short-term global and sleep outcomes in selected populations without evidence of serious short-term adverse effects. Effect sizes varied and a large placebo response was observed. Applicability, comparative effectiveness, and long-term efficacy and adverse effects, especially among older adults, are less well known.
  • Nonbenzodiazepine hypnotics have low to moderate strength evidence for efficacy on global and some sleep outcomes in the general adult population. Improvements over placebo in sleep outcomes were higher with eszopiclone and zolpidem than zaleplon. Results for adverse effects were mixed with few differences compared to placebo.
  • Low strength evidence shows that eszopiclone improved one global outcome by a minimum important difference and improved several sleep outcomes, but not sleep onset latency in older adults. Evidence on adverse effects was insufficient. Low strength evidence shows that zolpidem improved sleep onset latency in older adults. Evidence on other outcomes was insufficient.
  • The melatonin agonist, ramelteon did not improve global or sleep outcomes in a clinically meaningful way in the general population. Withdrawals were higher with ramelteon (low strength evidence), but withdrawals for adverse effects and number of patients with more than one adverse effect were similar in both groups (low and moderate strength evidence, respectively).
  • Very few benzodiazepine trials met eligibility criteria. Data were insufficient to assess any global, sleep, or adverse effect outcomes in the general adult or older adult populations.
  • Long-term adverse effects were derived from observational studies and suggest that use of hypnotics may be associated with dementia but not mortality. Zolpidem but not benzodiazepines may be associated with fractures. Withdrawal due to any reason was common especially with ramelteon.
  • The orexin receptor antagonist, suvorexant, improved global and sleep outcomes versus placebo (moderate strength evidence). Adverse effects did not differ between groups.
  • Four small trials compared cognitive behavioral therapy for insomnia (CBT-I) versus nonbenzodiazepine hypnotics or benzodiazepines. Results were mixed and evidence was insufficient.

We identified 37 publications reporting 36 unique RCTs of acceptable risk of bias that evaluated pharmacologic treatments for insomnia disorder in the general adult population and in older adults. We found the most data on the newer FDA-approved drugs. Patients were typically diagnosed with insomnia disorder according to DSM IV criteria (Appendix E). While DSM-IV criteria require symptoms to be present for at least 1 month, the mean duration of symptoms was rarely reported. Additional enrollment criteria were based on thresholds for sleep onset latency (SOL), total sleep time (TST) and/or less frequently wake after sleep onset (WASO) and/or number of awakenings per night during a typical night over the month prior to enrollment. None used global measures for enrollment, though some also required that patients reported some daytime dysfunction associated with insomnia. Only two trials reported severity based on scores of global measures such as the ISI in addition to a total sleep time of ≤6.5 hours and/or SOL of >30 minutes. When WASO was included, the threshold for enrollment ranged from 30 to 120 minutes. Trials rarely assessed treatments longer than 4 weeks. Most enrollees were female, of white race, and less than 50 years of age. Most studies were industry sponsored. Few antidepressant or benzodiazepine trials met eligibility criteria, primarily due to short treatment durations. Global outcomes were less often measured than sleep outcomes. Minimum important differences were identified for some instruments used to assess global outcomes, but these were not frequently used nor is it clear whether they are well established.

Efficacy of Nonbenzodiazepine Hypnotics in the General Adult Population

We identified 14 RCTs that assessed the efficacy of three nonbenzodiazepine hypnotics commonly used to treat insomnia disorder in the United States (eszopiclone [Lunesta], zaleplon [Sonata], and zolpidem [Ambien]) (Table 13).

Table 13. Efficacy of nonbenzodiazepine hypnotics: overview and strength of evidence.

Table 13

Efficacy of nonbenzodiazepine hypnotics: overview and strength of evidence.

Efficacy of Eszopiclone (Brand Name Lunesta)

Overview of Studies

Three moderate risk of bias RCTs (n=1929) analyzed the efficacy of eszopiclone 2-3 mg daily108-110 (Table 13). The mean age was 49 years; 63 percent were female. Most participants were white in the trials that reported race/ethnicity. All trials were conducted in the United States. Participants were randomized to 2 mg110 or 3 mg eszopiclone.108-110 One trial lasted 6 weeks110 and two lasted 6 months.108,109 All trials reported industry sponsorship and had moderate risk of bias.

Global Outcomes

Only Walsh et al. (n=825) reported clinically meaningful improvement in sleep based on ISI scores (Figure 26).109 Eszopiclone more often resulted in remission or no clinically significant insomnia compared with placebo, indicated by an ISI score <7 at endpoint (50% vs. 19%) (low strength evidence). The difference in the mean change of ISI scores from baseline at 12 weeks of was -4.6 points (95% CI, -5.3 to -3.9) but this difference did not reach our minimum important difference of 7 points, indicating ‘responder’ to treatment.

Figure 26 is a forest plot displaying data presented in Table 13 about the efficacy of eszopiclone in the general adult population for the outcome remission from insomnia disorder. In a single study, the risk ratio for remission was 2.66 in the eszopiclone group compared with placebo (significant).

Figure 26

Efficacy of eszopiclone: remitters. CI = confidence interval; M-H = Mantel-Haenszel; SD = standard deviation

Sleep Outcomes

Eszopiclone reduced sleep onset latency by 19 minutes and increased TST by 45 minutes compared with placebo (Figure 27). Mean sleep onset latency remained above 30 minutes in both groups in all three trials. Strength of evidence for both outcomes was moderate. Moderate strength of evidence also showed improved sleep quality with eszopiclone versus placebo. Low-strength evidence showed that eszopiclone decreased wake time after sleep onset more than placebo, but there was substantial heterogeneity between trials (I2 = 70%). Within the two 6-month trials, Walsh et al.109 reported greater improvement in wake time after sleep onset with eszopiclone compared with placebo (mean difference of 18 minutes) and Krystal et al.108 reported eszopiclone was not more effective than placebo.

Figure 27 is a forest plot displaying data presented in Table 13 about the efficacy of eszopiclone in the general adult population for the outcome SOL. When data from three studies were pooled, mean SOL was 19.07 minutes shorter in the eszopiclone group than the placebo group (significant).

Figure 27

Efficacy of eszopiclone: sleep onset latency, minutes. CI = confidence interval; IV = inverse variance; SD = standard deviation

Functioning, Mood, and Quality of Life

Secondary outcomes were rarely reported. Walsh et al. found that eszopiclone led to larger improvements in SF-36 domains of physical functioning, vitality, and social functioning than placebo.109

Adverse Effects

All three trials reported adverse effects. Withdrawal for any reason was higher with placebo than eszopiclone (41% vs. 33%). Withdrawals due to adverse effects did not significantly differ between groups (9% vs. 6%). Strength of evidence was low for both outcomes. A higher percentage of participants reported at least one adverse effect with eszopiclone than placebo (7% vs. 60%) (moderate strength of evidence). Krystal et al. reported a higher rate of serious adverse effects with eszopiclone than with placebo (3% vs. 1%) at 6 months.108 Neither 6-month trial reported evidence of tolerance or withdrawal symptoms following discontinuation.108,109 Specific adverse effects associated with eszopiclone use were somnolence (9% vs. 3% for placebo), unpleasant taste (23% vs. 3%), and myalgia (9% vs. 4%).

Efficacy of Zaleplon (Brand Name Sonata)

Overview of Studies

Two 4-week RCTs (n=973) compared zaleplon with placebo.111,112 The mean age was 42 years; 61 percent were female. Participants were overwhelmingly white. One trial was conducted in the United States112 and one was conducted in Canada and Europe.111 Participants were randomized to 5, 10, or 20 mg doses. Both trials reported industry sponsorship and had moderate risk of bias.

Global Outcomes

Neither zaleplon trial reported global outcomes.

Sleep Outcomes

Fry et al. reported that zaleplon 10 mg but not 5 mg reduced mean sleep onset latency versus placebo (Figure 28).112 Both trials reported that zaleplon did not consistently improve median total sleep time over placebo at 4 weeks. Participants randomized to any zaleplon dose were more likely than placebo participants to report improved sleep quality at week 4 (57% vs. 48%) (moderate strength of evidence) (Figure 29).111,112 Individually, zaleplon doses of 5 and 20 mg, but not 10 mg, were superior to placebo in improving sleep quality at week 4 (57% vs. 48% and 60% vs. 48%, respectively).

Figure 28 is a forest plot displaying data presented in Table 13 about the efficacy of zaleplon in the general adult population for the outcome SOL. In a single study, mean SOL was 2.50 minutes shorter with zaleplon 5 mg (not significant) and 9.90 minutes shorter that zaleplon 10 mg (significant) compared with placebo.

Figure 28

Efficacy of zaleplon: subjective sleep latency, minutes. CI = confidence interval; IV = inverse variance; SD = standard deviation

Figure 29 is a forest plot displaying data presented in Table 13 about the efficacy of zaleplon in the general adult population for improvement in sleep quality. When data from two studies were pooled, the risk ratio for improvement in sleep quality was 1.19 in the zaleplon group compared with placebo (significant).

Figure 29

Efficacy of zaleplon: sleep quality, participants reporting improvement. CI = confidence interval; M-H = Mantel-Haenszel

Functioning, Mood, and Quality of Life

No functioning, mood, and quality of life outcomes were reported in zaleplon trials.

Adverse Effects

Adverse effects were reported in all trials. Low-strength evidence shows that zaleplon at any dose compared with placebo did not increase withdrawals for any reason (12% vs. 8%) or withdrawals due to adverse effects (4% vs. 2%). Moderate-strength evidence shows that the proportion of participants reporting at least one adverse event did not differ between the zaleplon and placebo groups (71% vs. 73%). No individual adverse effect was greater with zaleplon than placebo. Neither trial reported evidence of tolerance or withdrawal symptoms. No RCTs evaluated long-term efficacy or harms (1 year or longer) of zaleplon.

Zolpidem (Brand Name Ambien)

Overview of Studies

Six RCTs compared zolpidem with placebo.60,111-115 Treatment duration was between 4 and 6 weeks for five of the trials. One trial was longer-term, with treatment duration up to 8 months.115, The mean age was 44, and 58 percent were female among the 844 participants randomized. Participants were overwhelmingly white. Five trials were conducted in the United States60,112-115 and one in Europe and Canada.111 Four trials evaluated a 10 mg dose60,111,112,115 and two trials evaluated 10 and 15 mg doses.113,114 One trial administered a 5 mg dose for participants 60 years of age or older.115 Risk of bias was moderate in all trials. Three trials reported industry sponsorship, and two trials were supported by government funding. Sponsorship was unclear in one trial.114

Global Outcomes

No zolpidem trial reported global outcomes.

Sleep Outcomes

Moderate strength evidence showed that zolpidem 5-10 mg reduced sleep onset latency by 15 minutes compared with placebo in four trials lasting 4-6 weeks and reporting poolable data (Figure 30).60,112,114,115 The one longer-term trial by Randall et al. reported that zolpidem was no more effective than placebo in improving sleep onset latency at over 8 months.115 The 15 mg dose in Scharf et al. was better than placebo (28 minutes vs. 48 minutes reduction in sleep onset latency).114 In the trials not pooled due to variations in how they reported outcomes, Lahmeyer et al. reported improvement in sleep onset latency at 4 weeks compared with placebo (reductions from baseline approximately 30 minutes vs. 10 minutes).113 Elie reported that zolpidem was no more effective than placebo in improving sleep onset latency at week 4.111 Moderate strength evidence shows that zolpidem improved sleep quality or the proportion of participants “getting a better night's sleep” more than placebo (69% vs. 49%) (Figure 31). Lahmeyer et al. reported that 10 and 15 mg zolpidem improved clinical global impression of sleep quality over placebo (both 84% vs. 49%).113 Short-term, moderate strength evidence showed that zolpidem 5-10 mg increased total sleep time by 23 minutes compared with placebo in three trials reporting poolable data (Figure 32).60,112,114,115 In the trials not pooled, zolpidem did not consistently improve total sleep time or sleep quality compared with placebo across trials. The one longer-term trial (n=91) reported that zolpidem was no more effective than placebo in increasing total sleep time or improving other subjective sleep outcomes (wake time after sleep onset, sleep quality) over 8 months.115

Figure 30 is a forest plot displaying data presented in Table 13 about the efficacy of zolpidem in the general adult population for the outcome SOL. When data from four studies were pooled, mean SOL was 14.95 minutes shorter in the zolpidem group than the placebo groupover four to six weeks of treatment (significant).

Figure 30

Efficacy of zolpidem: subjective sleep latency, minutes. CI = confidence interval; IV = inverse variance; SD = standard deviation

Figure 31 is a forest plot displaying data presented in Table 13 about the efficacy of zolpidem in the general adult population for improvement in sleep quality. When data from three studies were pooled, the risk ratio for improvement in sleep quality was 1.40 in the zolpidem group compared with placebo (significant).

Figure 31

Efficacy of zolpidem: sleep quality, participants reporting Improvement. CI = confidence interval; M-H = Mantel-Haenszel

Figure 32 is a forest plot displaying data presented in Table 13 about the efficacy of zolpidem in the general adult population for the outcome TST. When data from three studies were pooled, mean TST was 22.95 minutes longer in the zolpidem group than the placebo group over four to six weeks of treatment (significant).

Figure 32

Efficacy of zolpidem: total sleep time. CI = confidence interval; IV = inverse variance; SD = standard deviation

Functioning, Mood, and Quality of Life

No functioning, mood, and quality of life outcomes were reported in zolpidem trials.

Adverse Effects

Study withdrawals for any reason (15% vs. 12%) or reporting of at least one adverse effect (68% vs. 67%) were not greater with zolpidem than with placebo. Strength of evidence was low and moderate, respectively. Moderate-strength evidences suggests that zolpidem resulted in more withdrawals due to adverse effects than placebo (6% vs. 2%). Among adverse effects reported, somnolence was greater with zolpidem than placebo (10% vs. 3%). Frequencies of other adverse effects were comparable to placebo. Two trials reported a higher incidence of withdrawal symptoms and rebound insomnia following discontinuation of zolpidem compared with placebo.111,112 Incidence of withdrawal symptoms and rebound insomnia did not differ between treatment groups in the other two trials.113,114

Zolpidem ‘As Needed’

Overview of Studies

We identified three eligible RCTs that compared zolpidem ‘as needed’ with placebo.116-118 One lasted 12 weeks117 one 8 weeks,118 and one 4 weeks.116 Among the 607 randomized, the mean age was 44, and 73 percent were female. Perlis et al. reported more female in the placebo arm (81%) than the zolpidem arm (69%).117 Most participants in the one trial that reported race/ethnicity were white.117 Two trials were conducted in the United States117,118 and one in France.116 Participants were randomized to 10 mg zolpidem or placebo ‘as needed’ in all trials. Two trials reported industry sponsorship.116,117 Sponsorship was unclear in one trial.118 Risk of bias was moderate for all trials.

Global Outcomes

Only Allain et al. reported a global outcome (Figure 33).116 Low-strength evidence showed that zolpidem “as needed” led to more than a two-fold increase in clinician rated global impression (CGI) “much or very much improvement” versus placebo (54% vs. 24%).

Figure 33 is a forest plot displaying data presented in Table 13 about the efficacy of zolpidem as needed in the general adult population for the outcome clinical global impression of improvement. In a single study, the risk ratio for improvement was 2.22 in the zolpidem as needed group compared with placebo (significant).

Figure 33

Global improvement of zolpidem ‘as needed,’ participants reporting improvement. CI = confidence interval; M-H = Mantel-Haenszel

Sleep Outcomes

In two trials reporting poolable data, moderate-strength evidence showed that zolpidem 10 mg ‘as needed’ reduced sleep onset latency by 15 minutes (Figure 34) and increased total sleep time by 48 minutes compared with placebo (95% CI, 35 to 62) on nights when medication was taken.117,118 There were no significant improvements in SOL between groups when all nights (nights zolpidem was taken and not taken combined) were considered. Allain et al. reported no significant improvements versus placebo in sleep onset latency, total sleep time, wake time after sleep onset, and number of awakenings after sleep onset with zolpidem ‘as needed.’116 Compared with placebo, Perlis et al. reported significant improvements with zolpidem ‘as needed’ in wake time after sleep onset (-22 minutes (95% CI, -37 to -9) and number awakenings after sleep onset on the nights zolpidem was taken.117 There were no significant improvements between groups when data for all nights were pooled for these outcomes.

Figure 34 is a forest plot displaying data presented in Table 13 about the efficacy of zolpidem as needed in the general adult population for the outcome SOL. When data from two studies were pooled, mean SOL was 14.80 minutes shorter in the zolpidem as needed group than the placebo group (significant).

Figure 34

Subjective sleep latency, minutes: zolpidem ‘as needed’ versus placebo. CI = confidence interval; IV= inverse variance; SD = standard deviation

Functioning, Mood, and Quality of Life

Zolpidem 10 mg ‘as needed’ led to greater improvement in the Medical Outcomes Sleep (MOS) questionnaire compared with placebo (SMD 0.48 [95% CI, 0.22 to 0.74]); treatment effects did not differ for any SF-36 domain.116

Adverse Effects

Zolpidem ‘as needed’ and placebo were similar in the number of study withdrawals for any reason (13% vs. 13%) or withdrawals due to adverse effect (4% vs. 1%). The strength of evidence was low and insufficient, respectively. Adverse effects associated with zolpidem ‘as needed’ included anxiety, somnolence, mood alterations, hallucinations, and depression. We identified no RCTs that evaluated the long-term effects (1 year or longer) of zolpidem ‘as needed.’

Efficacy of Zolpidem, Special Formulations: Zolpidem Sublingual

Overview of Studies

One 4-week trial compared low-dose zolpidem sublingual 3.5 mg ‘as needed’ with placebo in participants with difficulty returning to sleep after middle-of-the-night awakenings.119 Among the 295 randomized, the median age was 43; 68 percent were female and 64 percent were white. The trial was industry sponsored and conducted in the United States. Risk of bias was moderate.

Global Outcomes

No global outcomes were reported for zolpidem SL.

Sleep Outcomes

Zolpidem sublingual reduced sleep onset latency after middle-of-the-night awakenings compared with placebo by 18 minutes (low strength evidence) (Figure 35).119 Zolpidem sublingual did not improve total sleep time or wake time after sleep onset following middle of the night awakening over placebo at 4 weeks. The strength of evidence was insufficient for both outcomes. Improvement in sleep quality was reported with zolpidem sublingual during nights when medication was taken.

Figure 35 is a forest plot displaying data presented in Table 13 about the efficacy of zolpidem extended release in the general adult population for the outcomes clinical global impression of improvement, “medications helped me sleep,” and “shortened time to fall asleep.” In a single study, the risk ratio for improvement was 1.77, for “medications helped me sleep” was 1.54, and for “shortened time to fall asleep” was 1.47 with zolpidem extended-release compared with placebo (all significant).

Figure 35

Efficacy of zolpidem extended release: clinical global impression and patient's global impression items at week 24, participants reporting improvement. CI = confidence interval; M-H = Mantel-Haenszel

Functioning, Mood, and Quality of Life

No functioning, mood, and quality of life outcomes were reported for zolpidem sublingual.

Adverse Effects

Withdrawals for any reason (8% vs. 6%) were not different with zolpidem sublingual and placebo. A similar number of participants withdrew due to adverse effects (0% vs. <1%) and reported at least one adverse effect (19% each).119 The strength of evidence was insufficient for both outcomes. Specific adverse effects associated with zolpidem sublingual were headache (3%) and nausea and fatigue (1% each). Nasopharyngitis (3% was the most commonly reported adverse effect with placebo. No deaths occurred during the trial. We identified no trials that evaluated long-term efficacy and harms (1 year or longer) for zolpidem sublingual.

Efficacy of Zolpidem, Special Formulations: Zolpidem Extended Release

Overview of Studies

Krystal et al., compared zolpidem extended-release 12.5 mg taken at least 3 nights per week with placebo over 24 weeks.120 Among the 1018 randomized, the mean age was 46; 61 percent were female and 65 percent were white. The trial was industry sponsored and conducted in the United States. Risk of bias was low.

Global Outcomes

Clinician-rated CGI outcome, “much or very much improvement,” favored zolpidem extended release over placebo (85% vs. 48%) (low strength of evidence).

Sleep Outcomes

Improvements in sleep onset latency, total sleep time, and wake time after sleep onset were greater in the zolpidem extended release group compared with the placebo group. Strength of evidence was low for all outcomes. Zolpidem extended release led to greater improvements in Patient's Global Impression (PGI) items compared with placebo (insufficient evidence).120 More than 90 percent of participants randomized to zolpidem extended release reported “medication helped me sleep” compared with 60 percent of the participants randomized to placebo (insufficient evidence).

Functioning, Mood, and Quality of Life

Krystal et al. reported that the Epworth Sleepiness Scale was significantly lower in the zolpidem extended release group compared with the placebo group during the double-blind treatment phase.120 At month 5, mean change from baseline was -2.5 and -1.8 points in the zolpidem extended release and placebo groups, respectively (p=0.02).

Adverse Effects

Withdrawals for any reason were greater with placebo than zolpidem extended release (48% vs. 36%).120 Conversely, withdrawals due to adverse effects were greater with zolpidem extended release than placebo (8% vs. 5%). Reports of at least one adverse effect were also greater with zolpidem extended release than placebo (63% vs. 51%). Strength of evidence was low for all outcomes. No rebound insomnia was reported over the first 3 nights following discontinuation of zolpidem extended release.

Efficacy of Nonbenzodiazepine Hypnotics in Older Adults

Eszopiclone

Overview

A single randomized, double-blind, placebo-controlled trial (n=388) evaluated eszolpiclone in older adults (Table 14).121 The mean age of enrollees was 72 years; 63 percent were female. Most participants were white. Participants randomized to eszopiclone received a 2 mg dose. The duration of the study was 12 weeks. The trial was conducted in the United States. Risk of bias was moderate and the trial reported industry sponsorship.

Table 14. Efficacy of nonbenzodiazepine hypnotics in older adults: overview and strength of evidence.

Table 14

Efficacy of nonbenzodiazepine hypnotics in older adults: overview and strength of evidence.

Global Outcomes

Low-strength evidence shows that compared with placebo, eszopiclone more often resulted in remission or no clinically significant insomnia, indicated by an ISI score <7 at endpoint (37% vs. 24%) (Figure 36). The mean difference in mean change from baseline in ISI scores over 12 weeks of was -2.3 points, but this difference did not reach our minimum important difference of 7 points, indicating ‘responder’ to treatment (Figure 37).

Figure 36 is a forest plot displaying data presented in Table 14 about the efficacy of eszoiclone in older adults for the outcome remission from insomnia disorder. In a single study, the risk ratio for remission was 1.51 in the eszopiclone group compared with placebo (significant).

Figure 36

Efficacy of eszopiclone in older adults: remitters. CI = confidence interval; M-H = Mantel-Haenszel

Figure 37 is a forest plot displaying data about the efficacy of eszoiclone in older adults for the outcome change in ISI score from baseline. In a single study, the mean change from baseline in ISI score was 2.30 lower (better) in the eszopiclone group than the placebo group (significant).

Figure 37

Efficacy of eszopiclone in older adults: ISI scores, mean change from baseline over 12 weeks. CI = confidence interval; IV = inverse variance; SD = standard deviation

Sleep Outcomes

Subjective sleep onset latency was not improved with eszopiclone versus placebo in older adults (insufficient strength of evidence). Compared with placebo, improvements were reported for total sleep time and wake time after sleep onset (Figure 38). Over 12 weeks, differences in the mean changes from baseline were 30 minutes for total sleep time and -22 minutes for wake time after sleep onset. Small significant improvement in sleep quality was also observed. Strength of evidence for was low for these sleep outcomes.

Figure 38 is a forest plot displaying data presented in Table 14 about the efficacy of eszoiclone in older adults for the outcomes SOL, TST, and WASO. In a single study, the mean SOL was 4.70 minutes shorter (not significant), TST was 30.00 minutes longer (significant), and WASO was 21.60 minutes shorter (significant) in the eszopiclone group than the placebo group.

Figure 38

Efficacy of eszopiclone in older adults: patient-reported sleep outcomes, mean changes from baseline. CI = confidence interval; IV = inverse variance; SD = standard deviation

Functioning, Mood, and Quality of Life

Quality of life was evaluated with the 36-Item Short-Form Health Survey (SF-36). Compared with placebo, statistically significant improvements were observed in the vitality and general health scales at week 12.

Adverse Effects

There were no statistically significant differences in study withdrawals, participants reporting at least one adverse effect (insufficient strength of evidence), and study withdrawals due to adverse effects (insufficient strength of evidence), between the eszopiclone and placebo groups. The specific adverse effect associated with eszopiclone use was unpleasant taste (12% vs. 2% in the placebo arm). There were two deaths in the eszopiclone group: one participant committed suicide, and one died of arteriosclerotic heart disease. Based on continued improvements in sleep outcomes in the eszopiclone group during the discontinuation phase, no evidence of rebound effect was reported. However, the percentage of participants with ISI total scores categorized as “no insomnia” and “sub-threshold insomnia” declined in the eszopiclone group from 78 percent at week 12 when treatment was discontinued to 53 percent at week 16. A regression of sleep latency in the eszopiclone group to the level of the placebo group was also observed at day 28 after the drug was withdrawn.

Zolpidem

Overview

We identified one randomized, double-blind, placebo-controlled trial evaluating zolpidem that enrolled older adults.122 The study was a four-arm trial that also included triazolam and temazepam. The trial randomized 166 participants between the ages of 59 and 85 years. Sex and race were not reported. Participants randomized to zolpidem received a 5 mg dose. The duration of the study was 4 weeks. The trial was conducted in the United States. Risk of bias was moderate and the trial reported industry sponsorship.

Global Outcomes

Leppik et al. did not report a global outcome.122

Sleep Outcomes

Subjective sleep onset latency was improved with zolpidem versus placebo in older adults (low-strength evidence) (Figure 39). Mean decreases from baseline were 40 and 21 minutes for the zolpiem and placebo groups, respectively. Total sleep time was not improved with zolpidem (insufficient evidence).

Figure 39 is a forest plot displaying data presented in Table 14 about the efficacy of zolpidem in older adults for the outcomes change from baseline in SOL and TST. In a single study, the mean change from baseline in SOL was 18.30 minutes shorter (significant) and the mean change from baseline in TST was 18.20 minutes longer (not significant) in the zolpidem group than the placebo group.

Figure 39

Efficacy of zolpidem in older adults: patient-reported sleep outcomes, mean changes from baseline. CI = confidence interval; IV = inverse variance; SD = standard deviation

Functioning, Mood, and Quality of Life

Leppik 1997 et al. did not report functioning, mood, and quality of life outcomes.122

Adverse Effects

There were no statistically significant differences in study withdrawals, study withdrawals due to adverse effects, and participants reporting at least one adverse effect between the zolpidem and placebo groups (insufficient evidence). No specific adverse effect was greater with zolpidem compared with placebo. One participant in the placebo group died during the trial.

Efficacy of Nonbenzodiazepine Hypnotics in Patients With Chronic Low Back Pain

Overview of Studies

We identified one randomized, double-blind, placebo-controlled trial evaluating eszopiclone in participants with both chronic insomnia and chronic low back pain (Table 15).123 All participants also received naproxen 500 mg twice a day. The trial randomized 58 participants with a mean age of 43; 63 percent were female. Slightly more participants were African-American (46%) than white (44%). Participants randomized to eszopiclone received a 3 mg dose. The duration of the study was 1 month. The trial was conducted in the United States. Risk of bias was moderate and the trial reported industry sponsorship.

Table 15. Efficacy of nonbenzodiazepine hypnotics in participants with chronic low back pain: overview and strength of evidence.

Table 15

Efficacy of nonbenzodiazepine hypnotics in participants with chronic low back pain: overview and strength of evidence.

Global Outcomes

Remission or no clinically significant insomnia, indicated by an ISI score, was not reported. The difference in mean change of ISI scores from baseline at week 4 of was -6.6 points [95% CI, -9.3 to -3.6]), favoring eszopiclone versus placebo, but this difference did not reach our minimum important difference of 7 points, indicating ‘responder’ to treatment.

Sleep Outcomes

Insufficient strength of evidence shows improvement with eszopiclone versus placebo in sleep outcomes at week 4 in adults with low back pain.

Functioning, Mood, and Quality of Life

Functioning, mood, and quality of life outcomes were not reported.

Adverse Effects

Compared with placebo, overall study withdrawals were lower in the eszopiclone group. The evidence was insufficient for all outcomes.

Efficacy of Melatonin and Ramelteon in the General Adult Population

Melatonin

Overview of Studies

We identified one RCT that compared melatonin 2 mg prolonged release (PR) with placebo reported in two publications (Table 16).124,125 Initially, the 791 randomized participants were randomized to melatonin PR or placebo for a 3-week, double-blind, period. After the 3 weeks, the melatonin group remained on melatonin while those in the placebo group were re-randomized to melatonin PR or placebo for a 26-week extension period (a total of 711 participants [534 melatonin and 177 placebo]). Our review focuses on the outcomes evaluated during the 26-week extension period. Demographic data for the 711 participants entering the extension period were not provided. However, among the 722 participants completing the initial 3-week period, mean age was 62 years, 69 percent were female, and nearly all were white (99%). The trial was conducted in Scotland, reported industry sponsorship, and had a moderate risk of bias.

Table 16. Efficacy and comparative effectiveness of melatonin and melatonin agonists: overview and strength of evidence.

Table 16

Efficacy and comparative effectiveness of melatonin and melatonin agonists: overview and strength of evidence.

Global Outcomes

Evidence was insufficient regarding melatonin PR improving global outcomes. The mean difference in PSQI scores between groups was statistically significant but very small (-0.39 points [95% CI, -0.71 to -0.08]).

Sleep Outcomes

Insufficient-strength evidence found melatonin PR improved subjective sleep onset latency. The mean difference between groups was statistically significant but small (6 minutes [95% CI 2 to 10]). Other sleep outcomes were not reported.

Functioning, Mood, and Quality of Life

Overall, melatonin PR improved WHO-5 quality of life scores compared with placebo. The mean difference between groups was 0.46 points (95% CI, 0.11 to 0.81).

Adverse Effects

Study withdrawals for any reason (21% vs. 24% placebo), withdrawals due to adverse effects (5% vs. 6%), and the proportion of participants reporting at least one adverse effect (74% vs. 77%) were similar with melatonin PR and placebo. Strength of evidence was insufficient for all outcomes. There were 15 serious adverse effects in the melatonin prolonged release group and nine (including one death) in the placebo group. There were no differences in type or frequency of adverse effects.

Ramelteon (Brand Name Rozerem)

Overview of Studies

We identified five RCTs that met our inclusion criteria.126-129 Two of the trials, NCT00237497 and NCT00671567, only had results published in a systematic review. The trials randomized 3124 participants; mean age was 45; 63 percent were female. In the two trials that reported race/ethnicity, most participants were white. Two trials were conducted in the United States,126,129 one in Japan,128 and two were multinational.126,127 Dosing ranged from 4 to 16 mg. All trials were short term (4 to 5 weeks) with the exception of Mayer et al., which lasted 6 months.127 All trials reported industry sponsorship and had moderate risk of bias.

Global Outcomes

None of the ramelteon trials reported global outcomes.

Sleep Outcomes

Patient-reported sleep outcomes from the five trials meeting eligibility criteria are presented in Figure 40. Ramelteon did not reduce sleep onset latency compared with placebo (low-strength evidence). The only study longer than 3 months127 reported an improvement in sleep onset latency of -6.8 minutes (95% CI, -13.5 to -0.1).127

Figure 40 is a forest plot displaying data presented in Table 16 about the efficacy of ramelteon in the general adult population for the outcome SOL. When data from five studies were pooled, mean SOL was 3.12 minutes shorter in the ramelteon group than the placebo group (not significant).

Figure 40

Efficacy of ramelteon: subjective sleep latency, minutes. CI = confidence interval; IV = inverse variance; SD = standard deviation

Low-strength evidence found that ramelteon did not significantly improve total sleep time or wake time after sleep onset compared to placebo. Ramelteon statistically improved sleep quality compared with placebo, but the effect size was less than small (ES 0.08), indicating little difference between groups (low-strength evidence). The 6-month trial by Mayer et al., the only trial lasting more than 3 months, reported no difference between treatment groups on any sleep outcome.127

Functioning, Mood, and Quality of Life

Functioning, mood, and quality of life outcomes were not reported.

Adverse Effects

Not all trials reported adverse effects. Ramelteon resulted in more withdrawals than placebo (12% vs. 10%; p=0.007; k=2; low strength evidence). Ramelteon and placebo were similar in withdrawals due to adverse effects (2% vs. 2%) and participants having at least one adverse event (46% vs. 46%) (strength of evidence was low and moderate, respectively). No specific adverse effect was greater with ramelteon than with placebo. Neither trial reported evidence of tolerance or withdrawal symptoms. No randomized studies evaluated long-term effects (1 year or longer) of ramelteon.

Efficacy of Melatonin and Ramelteon in Older Adults

Overview

We identified one randomized, double-blind, placebo-controlled trial evaluating ramelteon that enrolled older adults (Table 17).130 Additional outcomes data for this trial were obtained from the systematic review by Kuriyama et al.126 The three-arm trial randomized 829 participants with a mean age of 72; 59 percent were female. Race was not reported. Participants were randomized to 4 or 8 mg dose. Study duration was 5 weeks. The trial was conducted in the United States. Risk of bias was moderate and the trial reported industry sponsorship.

Table 17. Efficacy of melatonin agonists in older adults: overview and strength of evidence.

Table 17

Efficacy of melatonin agonists in older adults: overview and strength of evidence.

Global Outcomes

A global impression inventory was completed by both participants and clinicians. No statistically significant differences between treatment groups were reported (data were not reported).

Sleep Outcomes

Patient-reported sleep outcomes from all included trials are presented in Figure 41. Ramelteon dosage arms were combined for analyses (Figure 41).126 Ramelteon reduced sleep onset latency by 10 minutes compared with placebo. Ramelteon did not improve total sleep time or sleep quality over the 5 week study duration. Strength of evidence for sleep onset latency was low and insufficient for the other outcomes.

Figure 41 is a forest plot displaying data presented in Table 17 about the efficacy of ramelteon in older adults for the outcomes SOL and TST. In a single study, mean SOL was 10.10 minutes shorter (significant) and mean TST was 5.90 minutes longer (not significant) in the ramelteon group than the placebo group.

Figure 41

Efficacy of ramelteon in older adults: subjective sleep latency and total sleep time, minutes. CI = confidence interval; IV = inverse variance; SD = standard deviation.

Functioning, Mood, and Quality of Life

Roth et al. did not report functioning, mood, and quality of life outcomes.

Adverse Effects

We found no statistically significant differences in study withdrawals, study withdrawals due to adverse effects, or participants reporting at least one adverse effect between the ramelteon and placebo groups. Strength of evidence was insufficient for all outcomes. No specific adverse effect was greater with ramelteon compared with placebo.

Efficacy of Benzodiazepine Hypnotics in the General Adult Population

Overview of Studies

We identified one eligible RCT72 that assessed the efficacy of benzodiazepine, temazepam, versus placebo in the general adult population (Table 18).

Table 18. Efficacy and comparative effectiveness of the benzodiazepine hypnotics in general adult populations: overview and strength of evidence.

Table 18

Efficacy and comparative effectiveness of the benzodiazepine hypnotics in general adult populations: overview and strength of evidence.

Efficacy of Temazepam in the General Adult Population

Overview of Studies

One RCT72 met our inclusion criteria and compared temazepam to placebo in the general adult population. Wu et al. randomized participants to cognitive behavioral therapy alone, temazepam alone, cognitive behavioral therapy with temazepam, or placebo drug alone. For this aspect of the review we examined only the temazepam and placebo arms. Demographic information was not reported for the temazepam and placebo arms separately, but among the four treatment arms, the mean age was 38 years and 53 percent were female. Temazepam recipients initially received 7.5 mg nightly with gradual increases up to 30 mg, and then a decrease to 15 mg in the last treatment week for a total of 8 weeks. The trial was conducted in China and had government funding and was assessed as having a moderate risk of bias.

Global Outcomes

Wu et al.72 did not report global outcomes.

Sleep Outcomes

Sleep outcomes are presented in Figure 42. Temazepam reduced SOL by 31 minutes, increased TST by 94 minutes, and improved sleep efficiency by 14 percentage points compared with placebo. Evidence was insufficient for all outcomes.

Figure 42 is a forest plot displaying data presented in Table 18 about the efficacy of temazepam in the general adult population for the outcomes SOL, TST, and sleep efficiency. In a single study, the mean SOL was 30.90 minutes shorter (significant), TST was 93.50 minutes longer (significant), and sleep efficiency was 14.10 percentage points higher (significant) in the temazepam group than the placebo group.

Figure 42

Efficacy of temazepam: sleep latency minutes, total sleep time minutes, and sleep efficiency (percent). CI = confidence interval; IV = inverse variance; SD = standard deviation

Functioning, Mood, and Quality of Life

Temazepam significantly reduced the daytime dysfunction component of the PSQI compared with placebo.

Adverse Effects

There were no significant differences in overall withdrawals or withdrawals due to adverse effects between temazepam and placebo. Specific adverse effects were not reported. Strength of evidence was insufficient.

Efficacy of Benzodiazepine Hypnotics in Older Adults

We identified one RCT that met our inclusion criteria and assessed the efficacy and adverse effects of the benzodiazepine temazepam in older adults (Table 19).74

Table 19. Efficacy of the benzodiazepine hypnotics in older adults: overview and strength of evidence.

Table 19

Efficacy of the benzodiazepine hypnotics in older adults: overview and strength of evidence.

Efficacy of Temazepam in Older Adults

We identified one RCT74 that met our inclusion criteria and compared temazepam with placebo among older adults. Morin et al.74 et al. randomized participants to cognitive behavioral therapy alone, temazepam alone, cognitive behavioral therapy with temazepam, or placebo drug alone. For this aspect of the review, we examined only the temazepam and placebo arms. Morin et al.74 included only adults at least 55 year old; the 40 participants randomized had a mean age of 65 years and 60 percent were female; Morin et al.74 did not report other baseline characteristics. Morin et al. randomized participants to temazepam 7.5 mg nightly, with increases up to 30 mg nightly possible, depending on response and adverse effects; or to placebo drug. The trial lasted 8 weeks, was conducted in the United States, had government sponsorship, and was assessed as having a moderate risk of bias.

Global Outcomes

Morin et al.74 did not report any global outcomes.

Sleep Outcomes

Morin et al.74 found that wake time after sleep onset and sleep efficiency were significantly better with temazepam than placebo (insufficient evidence), but there was no significant difference in total sleep time (insufficient evidence).

Functioning, Mood, and Quality of Life

Morin et al.74 found no significant difference in the Sleep Impairment Index with temazepam compared with placebo (insufficient evidence).

Adverse Effects

There was no significant difference between temazepam and placebo groups in the proportion of participants withdrawing for any reason or withdrawing due to adverse effects.

Efficacy of Antidepressants in the General Adult Population

Overview of Studies

We identified two RCTs that compared doxepin with placebo in the general adult population131,132 (Table 20). Hajak et al.131 randomized 47 participants to doxepin 25 mg (increasing to 50 mg of doxepin as needed) or placebo. Krystal et al. 132 randomized 229 participants to either doxepin 3 mg, doxepin 6 mg, or placebo. Because different doses of doxepin were used, efficacy outcomes could not be pooled.

Table 20. Efficacy of doxepin in the general adult population.

Table 20

Efficacy of doxepin in the general adult population.

Both trials had active treatment lasting 4 weeks. Overall, the mean age was 45, and 74 percent were female. Only Krystal et al. 2011132 reported ethnicity: in that trial, 48 percent of participants were white. Hajak et al.131 was conducted in Germany and Krystal et al.132 was conducted in the United States. Both RCTs reported industry sponsorship. Both trials had moderate risk of bias.

Global Outcomes

Hajak et al.131 found doxepin significantly enhanced global improvement on the Clinical Global Impression Scale compared with placebo (2.42 vs. 3.00, where lower scores indicate more improvement) (insufficient strength of evidence). Hajak et al. found no significant differences between treatment groups in severity of illness from the Clinical Global Impression Scale.

Sleep Outcomes

Krystal et al.132 found that both doxepin doses significantly improved total sleep onset and wake time after sleep onset compared with placebo (Table 20). Strength of evidence was low for both outcomes. Hajak et al.131 found that doxepin 25 mg significantly improved sleep quality compared with placebo (52 vs. 41 on a 100-point visual-analog scale).

Functioning, Mood, and Quality of Life

Krystal et al.132 found no significant differences between the doxepin dose groups and placebo in the Digit Symbol Substitution Test, the Symbol Copying Test, or daytime sleepiness at 4 weeks. Hajak et al.131 found doxepin 25 mg significantly improved energy and working ability compared with placebo.

Adverse Effects

There were no significant differences in overall study withdrawals, study withdrawals due to adverse effects, participants reporting at least one adverse effect, daytime somnolence, or headache between participants receiving doxepin versus placebo.

Efficacy of Antidepressants in Older Adults

Overview of Studies

We identified two RCTs133,134 that compared doxepin with placebo in older adults (Table 21). Krystal et al.133 randomized 240 participants to either doxepin 1 mg, doxepin 3 mg, or placebo. Lankford et al.134 randomized 255 participants to doxepin 6 mg or placebo. Because different doses of doxepin were used, efficacy outcomes could not be pooled. Krystal et al.133 had an active treatment duration of 12 weeks and Lankford et al.134 was 4 weeks. The mean age was 72, 65 percent were female, and 84 percent were white. Both RCTs were conducted in the United States and reported industry sponsorship. Lankford et al.134 had low risk of bias and Krystal et al.133 had moderate risk of bias.

Table 21. Efficacy of doxepin in older adults.

Table 21

Efficacy of doxepin in older adults.

Global Outcomes

Both trials reported ISI scores. Our analyses found ISI scores were significantly improved with pooled doxepin 1-6 mg doses compared with placebo from 4 to 12 weeks, with a weighted mean difference of -1.9 points [95%CI -2.9 to -1.0] (Figure 43). Mean change in ISI scores at endpoint ranged from -3.4 (1 mg) to -5.4 points (6mg) in the doxepin groups and -2.4 to -3.5 in the placebo groups, respectively. Strength of evidence was moderate.

Figure 43 is a forest plot displaying data presented in Table 21 about the efficacy of doxepin in older adults for the outcome change from baseline in ISI scores. When data from two studies were pooled, mean change from baseline in ISI scores was 1.74 lower (better) (significant) with doxepin than with placebo.

Figure 43

Efficacy of doxepin in older adult population: ISI scores, mean change from baseline. CI = confidence interval; IV = inverse variance; SD = standard deviation

Lankford et al.134 found that doxepin 6 mg significantly improved three of four sleep components of the PGI scale compared with placebo at 4 weeks (Figure 44). Lankford et al.134 found CGI scores were not significantly different with doxepin 6 mg compared with placebo at 4 weeks. Krystal et al. 2010133 found that CGI scores were significantly better with doxepin 1 mg or doxepin 3 mg versus placebo at 12 weeks

Figure 44 is a forest plot displaying data about the efficacy of doxepin in older adults for improvement in components of the patient global impression scale. In a single study, the risk ratio for improvements was 1.46 for “helped sleep” (significant), 1.34 for “shortened onset of sleep” (marginally significant), 1.35 for “increased duration of sleep” (marginally significant), and 1.26 for “got better sleep” (not significant).

Figure 44

Efficacy of doxepin in older adults: patient global impression of sleep quality at final visit, participants reporting improvement. CI = confidence interval; SD = standard deviation; M-H Mantel-Haenszel

Sleep Outcomes

Krystal et al.133 reported significant improvement in sleep onset latency with doxepin compared with placebo (low strength of evidence) (Figure 45). Moderate strength evidence found improvements in total sleep time by 24 minutes (Figure 46) compared with placebo. Lankford et al.134 also reported that doxepin 6 mg improved WASO (Figure 47) an all studies reported improvement in sleep quality compared with placebo. At 12 weeks, Krystal et al. found all five sleep quality components of the PGI scale were significantly better with doxepin 1 mg and doxepin 3 mg compared with placebo.

Figure 45 is a forest plot displaying data presented in Table 21 about the efficacy of doxepin in older adults for the outcome SOL. In a single study, mean SOL was 14.70 minutes shorter (significant) with doxepin than with placebo.

Figure 45

Efficacy of doxepin in older adult population: sleep onset latency, mean change from baseline. CI = confidence interval; IV = inverse variance; SD = standard deviation

Figure 46 is a forest plot displaying data presented in Table 21 about the efficacy of doxepin in older adults for the outcome change from baseline in TST. When data from two studies were pooled, mean change from baseline in TST was 23.85 minutes longer (significant) with doxepin than with placebo.

Figure 46

Efficacy of doxepin in older adult population: total sleep time, mean change from baseline. CI = confidence interval; IV = inverse variance; SD = standard deviation

Figure 47 is a forest plot displaying data about the efficacy of doxepin in older adults for the outcome WASO. In a single study, mean WASO was 17.00 minutes shorter (significant) in the doxepin group than the placebo group.

Figure 47

Efficacy of doxepin in older adult population: wake time after sleep onset, mean change from baseline. CI = confidence interval; IV = inverse variance; SD = standard deviation

Krystal et al. 2010133 found no significant differences in next-day residual function and effects between both doxepin doses and placebo in the Digit Symbol Substitution Test, the Symbol Copying Test, or daytime sleepiness at 12 weeks.

Adverse Effects

There were no significant differences in overall study withdrawals, study withdrawals due to adverse effects, participants reporting at least one adverse event, or daytime somnolence, between participants receiving doxepin versus placebo. However, there were significantly fewer headaches (RR 0.29 [95% CI 0.29 to 0.70]) among participants receiving doxepin versus placebo.

Efficacy of Suvorexant in the General Population and Older Adults

Overview of Studies

We identified three RCTs that compared the orexin receptor antagonist, suvorexant, to placebo in mixed general and older populations (Table 22).135,136 Two of the trials were reported in one publication.137 The three trials randomized 2811 participants; mean age was 58; 62 percent were female. Most participants were white. The majority of the participants in the trial by Michelson et al. were aged 65 years of age or older (59%).135 Based on ISI scores at baseline, the participants typically had clinical insomnia of moderate severity. The two trials reported by Herring et al. evaluated two dose groups, a 20 mg (for participants <65 years of age)/15 mg (for participants ≥65 years) dose group and a 40 mg (<65 years)/30 mg (≥65 years) dose group.136 Michelson et al. evaluated the combined doses of 40 mg (<65 years) and 30 mg (≥65 years).136 The dose recommended is 10 mg but should not exceed 20 mg daily. Strength of evidence was determined for the lower 20/15 mg dose group (n=1260 participants). The trials included in Herring et al were short term, lasting 3 months. Michelson et al. was longer-term, with a double-blinded, treatment phase of one year. All trials reported industry sponsorship and had an overall moderate risk of bias.

Table 22. Efficacy of orexin receptor antagonists in the general population and older adults: overview and strength of evidence.

Table 22

Efficacy of orexin receptor antagonists in the general population and older adults: overview and strength of evidence.

Global Outcomes

Short-term, both trials evaluating 20/15 mg doses of suvorexant reported clinically meaningful improvement in sleep based on ISI scores (Figure 48).136 Moderate-strength evidence shows that compared with placebo, suvorexant more often resulted in response to therapy, indicated by a in the ISI score (55% vs. 42%). The mean difference in ISI scores at 3 months of was -1.2 points (95% CI, -1.8 to -0.6) but this difference was below the minimum important difference of 7 points, indicating ‘responder’ to treatment.

Figure 48 is a forest plot displaying data presented in Table 22 about the efficacy of suvorexant in the general population and older adults for the outcome response to therapy based on ≥ 6 point improvement in ISI scores. When data from two studies were pooled, the risk ratio for improvement was 1.32 (significant) in the suvorexant group compared with the placebo group, and the risk difference for improvement was 0.13 higher (significant) in the suvorexant group than the placebo group.

Figure 48

Efficacy of suvorexant 20/15 mg, participants responding to therapy. CI = confidence interval; SD = standard deviation; M-H = Mantel-Haenszel

Higher Dose (40/30 mg) Findings

Comparable to the 20/15 mg group, suvorexant 40/30 mg more often resulted in response to therapy compared with placebo short-term (55% vs. 42%). Pooled results from all three trials found the mean difference in ISI scores at 3 months of was -1.7 points (95% CI, -2.3 to -1.0) but this difference was also below the minimum important difference of 7 points. At one year, Michelson et al. reported marginally significant improvement versus placebo. The mean difference between groups was -0.9 (95% CI, -1.8 to -0.0).135

Sleep Outcomes

Moderate strength evidence shows suvorexant 15 or 20 mg treatment reduced sleep onset latency by 6 minutes compared with placebo (Figure 49).136 However, mean sleep onset latency remained above the 30 minute threshold indicating ‘no insomnia’ in both groups in all three trials. Compared with placebo, short-term suvorexant 20/15 mg therapy also improved TST by 16 minutes (Figure 50) (moderate strength evidence). WASO and sleep quality were improved with suvorexant versus placebo but the magnitude of the improvements were small (moderate strength of evidence).

Figure 49 is a forest plot displaying data presented in Table 22 about the efficacy of suvorexant in the general population and older adults for the outcome change from baseline in SOL. When data from two studies were pooled, the mean change from baseline in SOL was 5.97 minutes shorter (significant) in the suvorexant group than the placebo group.

Figure 49

Efficacy of suvorexant 20/15 mg: subjective sleep latency, mean change from baseline in minutes. CI = confidence interval; IV = inverse variance; SD = standard deviation

Figure 50 is a forest plot displaying data presented in Table 22 about the efficacy of suvorexant in the general population and older adults for the outcome change from baseline in TST. When data from two studies were pooled, the mean change from baseline in TST was 15.97 minutes longer (significant) in the suvorexant group than the placebo group.

Figure 50

Efficacy of suvorexant 20/15 mg: subjective total sleep time, mean change from baseline in minutes. CI = confidence interval; IV = inverse variance; SD = standard deviation

Higher Dose (40/30 mg) Findings

Short-term suvorexant 40/30 mg treatment reduced sleep onset latency by 10 minutes and increased TST by 23 minutes compared with placebo.136 Similar to findings in the lower dose groups, mean sleep onset latency remained above the 30 minute threshold indicating ‘no insomnia’ in both groups in all three trials. At one year, improvements were comparable to short-term.135

Functioning, Mood, and Quality of Life

Functioning, mood, and quality of life outcomes of interest were rarely reported. Over one year, suvorexant 40/30 mg had no effect on mood compared with placebo, based on assessment with the Quick Inventory of Depressive Symptomatology-Self Report.135

Adverse Effects

Low strength evidence found withdrawal for any reason and withdrawals due to adverse effects did not significantly differ between the suvorexant 20/15 mg and placebo groups short-term.136 Moderate strength evidence found no difference between groups in the proportions of participants reporting at least one adverse effect. The specific adverse effect most associated with short-term suvorexant 20/15 mg use was somnolence (7% vs. 3% for placebo; RR 2.5 [1.4 to 4.4]). One death was reported in the placebo group due to cerebrovascular accident. Suicidal ideation was reported in one suvorexant and placebo participant each (<1%). Incidence of excessive daytime sleepiness was reported in three participants in the suvorexant group and one in the placebo group. One incidence each of hypnagogic hallucination, hypnopompic hallucination, and sleep paralysis were reported in the suvorexant group, none in the placebo group. There were no differences in incidence of falls (4 vs. 7 for placebo) and motor vehicle accidents or violations (10 vs. 12) between the suvorexant and placebo groups.

Higher Dose (40/30 mg) Findings: Short Term (3 Months)

Similar to the findings of the suvorexant 20/15 mg group, withdrawal for any reason, withdrawals due to adverse effects and participants reporting at least one adverse effect did not significantly differ between the placebo groups.136 Somnolence was the most common adverse effect associated with suvorexant use, 11 percent versus 3 percent for placebo (RR 3.97 [2.58 to 6.09]).135,136 Two participants died during the trials, one in suvorexant group due to hypoxic-ischemic encephalopathy and one in the placebo group due to cerebrovascular accident. Suicidal ideation was reported in two suvorexant participants and one placebo participant.

Higher Dose (40/30 mg) Findings: Long Term (1 Year)

Withdrawal for any reason (suvorexant 38% vs. 37% for placebo), withdrawals due to adverse effects (12% vs. 9%), and participants reporting at least one adverse effect (70% vs. 64%) did not significantly differ between the suvorexant 40/30 mg and placebo groups.135 Specific adverse effects associated with suvorexant 40/30 mg use were somnolence (13% vs. 3% for placebo), fatigue (7% vs. 2%), and dry mouth (5% vs. 2%). Suicidal ideation was reported for four suvorexant participants (<1%), leading to study withdrawal for two of the participants. Excessive daytime sleepiness was more common in the suvorexant group (2.5% vs. 0.8% for placebo). Three incidences of hypnagogic hallucination and one each of somnambulism and hypnopompic hallucination were reported in the suvorexant group and none in the placebo group. There were no differences in incidence of falls between the suvorexant and placebo groups (2% vs. 3%).

Comparative Effectiveness of Pharmacologic Interventions for Insomnia Disorder

Zolpidem Versus Temazepam

Overview of Study

We identified one RCT that compared the nonbenzodiazepine zolpidem 10 mg to the benzodiazepine temazepam 20 mg over a 4 week treatment period (Table 23).138 Among the 223 randomized, baseline characteristics were available for 159 participants; mean age was 46 years; 67 percent were female. The trial was conducted in the Netherlands, reported industry sponsorship, and had a moderate risk of bias.

Table 23. Comparative effectiveness of nonbenzodiazepines versus benzodiazepines: overview and strength of evidence.

Table 23

Comparative effectiveness of nonbenzodiazepines versus benzodiazepines: overview and strength of evidence.

Global Outcomes

Evidence was insufficient to assess differences between groups in global outcomes. Following 4 weeks of treatment (Figure 51), Voshaar et al. found that 22 percent in the zolpidem group and 33 percent in the temazepam group reported that symptoms were “much-very much” improved on the CGI.

Figure 51 is a forest plot displaying data presented in Table 23 comparing the efficacy of zolpidem and temazepam for the outcome global improvement in sleep quality in the general adult population. Global improvement was defined as clinician's rating of “much or very much improved” on the CGI scale. In a single study, the risk ratio for improvement was 0.66 with zolpidem compared with temazepam (not significant).

Figure 51

Comparative effectiveness of zolpidem versus temazepam: global improvement, participants reporting improvement. CI = confidence interval; IV = inverse variance; SD = standard deviation

Sleep Outcomes

Sleep outcomes are presented in Figure 52. Evidence was insufficient to assess sleep outcomes. Voshaar et al. found that total sleep time improved with zolpidem compared with temazepam. There were no differences between groups for sleep onset latency and wake time after sleep onset.

Figure 52 is a forest plot displaying data presented in Table 23 comparing the efficacy of zolpidem and temazepam for the outcomes SOL, TST, and WASO in the general adult population. In a single study, mean SOL was the same (not significant); mean TST was 27.00 minutes longer (significant), and WASO was 1.00 minute longer (not significant) in the zolpidem group than the temazepam group.

Figure 52

Comparative effectiveness of zolpidem versus temazepam: subjective sleep outcomes. CI = confidence interval; IV = inverse variance; SD = standard deviation

Functioning, Mood, and Quality of Life

No functioning, mood, and quality of life outcomes were reported.

Adverse Effects

Overall withdrawals, withdrawals due to adverse effects, and participants with at least one adverse effect were not reported according to treatment arm. Nine participants withdrew due to an adverse effect. No participant experienced a major adverse effect.

Zolpidem Versus Zaleplon

Overview of Studies

We identified two 4-week RCTs evaluating zaleplon versus placebo that also included a zolpidem arm (Table 24).111,112 Head-to-head comparisons between zaleplon and zolpidem were not provided, which limited our assessment of comparative effectiveness. Among the 965 participants randomized to zaleplon or zolpidem, mean age was 42 years, 62 percent were female, and most were white (91%). One trial was conducted in the United States112 and one was conducted in Canada and Europe.111 Participants were randomized to zaleplon 5, 10, or 20 mg doses and zolpidem 10 mg. Both trials reported industry sponsorship and had moderate risk of bias.

Table 24. Efficacy and comparative effectiveness of nonbenzodiazepines: overview and strength of evidence.

Table 24

Efficacy and comparative effectiveness of nonbenzodiazepines: overview and strength of evidence.

Global Outcomes

The included trials did not report global outcomes.

Sleep Outcomes

Sleep outcomes from included trials are presented in Table 24, and Figures 53 and 54. Zolpidem 10 mg improved sleep onset latency compared with zaleplon 5 mg by approximately 14 minutes.112 Improvements in sleep onset latency were similar between the zolpidem and zaleplon 10 mg dose groups (insufficient evidence). We could not evaluate the comparative effectiveness of the two nonbenzodiazepine agents for total sleep time from the data reported (insufficient evidence). Both trials reported that zaleplon and zolpidem did not consistently improve median total sleep time compared with placebo over the 4 week study durations.

Figure 53 is a forest plot displaying data presented in Table 24 comparing the efficacy of zolpidem and zaleplon for the outcome SOL in the general adult population. In a single study, mean SOL was the 13.70 minutes shorter with zolpidem 10 mg than zaleplon 5 mg (significant); mean SOL was 1.30 minutes shorter with zolpidem 10 mg than zaleplon 10 mg (not significant).

Figure 53

Comparative effectiveness of zaleplon versus zolpidem: sleep onset latency. CI = confidence interval; IV = inverse variance; SD = standard deviation

Figure 54 is a forest plot displaying data presented in Table 24 comparing the efficacy of zolpidem and zaleplon for the outcome improvement in sleep quality in the general adult population. When data from two studies were pooled, the risk ratio for improvement was 0.90 in the zaleplon group compared with the zolpidem group (not significant).

Figure 54

Comparative effectiveness of zaleplon versus zolpidem: sleep quality, participants reporting improvement. CI = confidence interval; SD = standard deviation; M-H = Mantel-Haenszel

Sleep quality with zaleplon was similar to zolpidem at week 4 (57% vs. 64%) (moderate strength of evidence). There were also no significant differences between the individual zaleplon doses versus zolpidem at week 4.

Functioning, Mood, and Quality of Life

No functioning, mood, and quality of life outcomes were reported in the included trials.

Adverse Effects

Adverse effects were reported in both trials. There were no differences in withdrawals for any reason (12% each) and the proportion of participants reporting at least one adverse event (7% each) between the zaleplon and zolpidem groups. Withdrawals due to adverse effects were comparable between groups. Incidences of withdrawal symptoms and rebound insomnia following discontinuation were reported for zolpidem. Neither trial reported evidence of tolerance or withdrawal symptoms associated with zaleplon use.

Long-Term Adverse Effects: Analysis of Observational Studies

We used data from 12 observational studies including open-label extensions of RCTs to assess long-term harms of pharmacological treatments of insomnia.139-150 We included studies that reported harms if: (1) study population included adults with chronic insomnia without other major diagnoses such as cancer, Parkinson's, etc. or the hypnotics evaluated were only those that were FDA-indicated for insomnia and were likely administered for sleep disorders; (2) study duration was at least 6 months; and (3) study reported on at least 100 persons. Outcomes included percentage of individuals withdrawing from pharmacological treatments, reasons for withdrawal (lack of efficacy, adverse effects, other), any serious adverse effects (i.e., mortality), and specific adverse effects associated with the drug of interest. Followup duration ranged from 6 months to 12 years.

Any Hypnotic Drug

Four studies provided information on long-term harms with hypnotic drugs. Results suggest a correlation between hypnotic use and dementia and fractures; results regarding a correlation with mortality were mixed.

Using data from longitudinal electronic medical records, a matched cohort survival analysis identified 10,529 patients who received hypnotic prescriptions and compared them with 23,676 matched controls who did not receive hypnotic prescriptions.144 The study was conducted in the United States. Participants were matched by age, sex and smoking status and were followed for a mean of 2.5 years. Overall mean age was 54 years, 63 percent were female, and most were white. During the study interval from 2002 to 2006, zolpidem was the most commonly prescribed hypnotic (41%, n=4338), followed by temazepam (20%, n=2076). The participants were stratified into tertiles based on the number of doses prescribed per year. For the participants prescribed 0.4-18 pills per year (n=3491) of any hypnotic, the hazard ratio (HR) of death was 3.60 [95% CI, 2.92 to 4.44] compared with participants who were not prescribed hypnotics. For the participants prescribed 18-132 (n=3548) and >132 pills (n=3490) per year the HRs were 4.43 [95% CI, 3.67 to 5.36] and 5.32 [95% CI, 4.50 to 6.30], respectively, indicating a dose-response association. Kripke et al. also reported increased hazards of incidental major cancers for the middle and upper dose tertile groups. A major limitation to this study was that residual confounding could not be fully excluded, due to possible biases that affected which patients were prescribed hypnotics in addition to possible imbalances in surveillance.

A prospective cohort study examined the association between hypnotic usage and mortality over a 12-year period.142 The study, conducted in France, included older participants aged at least 65 years without dementia at baseline. Median age was 73 years, and 59 percent were female. Among the 6696 participants, 1454 were confirmed regular users of hypnotics, mainly benzodiazepines, and 5242 did not use hypnotics. Overall, 72 percent of participants reported at least one insomnia complaint (82% for hypnotic users vs. 70% for nonhypnotic users). Mortality was not significantly associated with hypnotic use. Over a median followup time of 8.9 years, all-cause mortality was not significantly different between groups, 22 percent (326/1454) in hypnotic user group compared with 19 percent (981/5242) in the nonhypnotic user group. Following adjustment for confounders, the HR was 1.03 [95% CI, 0.84 to 1.28]. Results were similar when limited to benzodiazepine use only. Study limitations included the unavailability of hypnotic dose data, low participation rate at baseline, and the nonrandom exclusion of participants with missing data at baseline.

A retrospective, case-control study from Korea evaluated the risk of fractures related with zolpidem in elderly insomnia patients.143 The 1508 study participants were mostly female (80%) and 31 percent had a history of osteoporosis. Cases were defined as subjects who had a diagnosis of a fracture, mainly in the femur. Hazard period exposures (1 day length prior to the fracture date) and control exposures (periods of the same length at 5, 10, 15, and 20 weeks prior to the fracture date) were established at a one-to-four ratio, resulting in 1508 hazard period exposures and 6032 control period exposures. During the hazard and control periods, 431 had used zolpidem more than once. Analysis of the data found use of zolpidem was associated with significant increase in the risk of with a fracture. The crude odds ratio was 1.84 [95% CI, 1.47 to 2.30]. Following adjustment of the effect of other drugs that can increase the risk of fall or fracture, the odds ratio was 1.72 [95% CI, 1.37 to 2.16]. Among the 703 patients that had used benzodiazepines more than once during the same exposure periods, there was no difference in risk of fracture (adjusted odds ratio 1.00 [95% CI, 0.83 to 1.21]).

A retrospective cohort study from Taiwan aimed to examine whether hypnotic use increased the risk of dementia in older adults.140 Using a large population database, the study cohort was comprised of 5693 subjects, median age 65 years and 56 percent female, with long-term insomnia who had been prescribed hypnotics, mainly nonbenzodiazepines (49%) followed by benzodiazepines (34%). The control group, in a five-to-one-ratio and matched by age and sex, comprised 28,465 subjects without insomnia. All subjects were examined over a 3-year period. Over the 3 year interval, 4 percent (220/5693) of 5693 subjects with insomnia and prescribed hypnotics were diagnosed with dementia compared with 1.5 percent (424/28,465) of the controls. Following adjustment for confounders, the HR was 2.34 [95% CI, 1.92 to 2.85]. Risk of dementia with hypnotic use was also greater in both male and female subgroups. Subjects aged between 50 and 65 years had the highest risk of dementia with an HR of 5.22 [95% CI, 2.62 to 10.41]. There was no difference in risk between nonbenzodiazepine versus benzodiazepine use (HR 1.01 [95% CI 0.76 to 1.33]. Limitations to the study included the inability to control for all confounders (educational level, personal history of smoking and alcohol consumption, body mass index, socioeconomic status) and a relatively short followup period that may not have been long enough for patients to develop dementia.

Specific Nonbenzodiazepines

Six studies provided longer term harms information on specific nonbenzodiazepines. Results suggest a correlation between nonbenzodiazapine use and mortality, major injury, fracture,

The previously discussed study by Kripke et al. also compared zolpidem alone (n=4338) with no hypnotic use (n=23,671).144 Comparable to any hypnotic prescribed, doses were stratified into tertiles of 5-130 (n=1453), 130-800 (n=1456), and >800 (n=1427) mg per year. Zolpidem use was associated with increased hazard of death for all tertiles, with HRs of 3.93 [95% CI, 2.98 to 5.17], 4.54 [95% CI, 3.46 to 5.95], and 5.69 [95% CI, 4.58 to 7.07], respectively. As with the findings of any hypnotic prescribed, a dose-response association was demonstrated. Increased hazard of incidental major cancers was only found in the upper dose tertile group (HR 1.28 [95%CI 1.03 to 1.59]).

A retrospective matched cohort study conducted in Taiwan examined the risk of major injury (head injury or fracture) requiring hospitalization in patients prescribed zolpidem.145 Participants and data were obtained from the Taiwan National Health Insurance population-based cohort database. Investigators identified 8188 participants who were at least 18 years of age and received a first prescription for zolpidem between January 2000 and December 2009 and compared them with 32,752 age- and sex-matched patients who were not prescribed hypnotic therapies. Overall mean age was 39 years and 49 percent were female. Use of zolpidem was found to be associated with risk of a major head injury or fracture requiring hospitalization compared with nonusers of hypnotics (adjusted HR 1.67 [95% CI, 1.19 to 2.34]). The incidence rate of major injury was 60.1 cases per 10,000 person-years in the zolpidem user group versus 36.7 cases per 10,000 person-years in the nonuser control group. A dose-response association was demonstrated when zolpidem dosage was increased. The adjusted HRs for the 71-800, 801-1600, and >1600 mg per year dosage groups were 2.04 [95% CI, 1.32 to 3.13], 4.37 [95% CI, 2.12 to 9.01], and 4.74 [95% CI, 2.38 to 9.42], respectively. The HR for major injury in zolpidem users in the younger cohort (aged 18-54 years) was 1.70 [95% CI, 1.15 to 2.51] after adjusting for diabetes, sleep disorder, alcohol-related disorders and other variables. The adjusted HR for major injury in the older zolpidem user cohort (aged >55 years) was not statistically significant. Limitations of the study included possible unmeasured or unknown confounders and the data in NHI claims are primarily intended for administrative billing purposes and have not been verified scientifically.

A case-control study conducted in the United States explored the association of zolpidem use and risk of hip fracture in older adults (≥ 65 years of age).150 The participants were enrolled in the New Jersey Medicaid program and information was extracted from January 1993 to June 1995. The cases included 1222 patients who underwent surgical repair of a hip fracture. They were matched by age and sex to controls in a 4 to 1 ratio (n=4888). Overall mean was 83 years, 84 percent were female, and most were white race. Zolpidem use was reported in 1.6 percent of the cases compared with 0.7 percent of the controls. Zolpidem use was found to be associated with a significant increased risk of hip fracture compared with no zolpidem use. The adjusted odd ratio (OR) was 1.95 [95% CI, 1.09 to 3.51]. An increased risk of hip fracture was also observed with benzodiazepines (adjusted OR 1.46 [95% CI 1.21 to 1.76]). A possible limitation of the study was confounding by indication (i.e., selection bias). The study authors attempted to control for a patient's underlying risk of hip fracture by adjusting their analyses for age, sex, and several markers of frailty, but it is possible that residual confounding by indication may have remained.

One open-label extension of two RCTs conducted in the United States and Europe evaluated the long-term use of zaleplon 5-10 mg doses in 576 older adults with insomnia disorder.139 Participants were followed 6 to 12 months following initial double-blind treatment phases. Mean ages of the participants were 73 and 72 years for the U.S. and European populations, respectively. No other demographic details were provided. No deaths occurred during the study. The most commonly reported specific adverse effects were headache (27%) and infection (13%).

One open-label extension of an RCT conducted in the United States evaluated the long-term use of eszopiclone 3 mg in 471 adults with chronic insomnia.147 Participants were followed an additional 6 months following an initial 6-month double-blind treatment phase. Mean age of the participants was 46 years, and 63 percent were female. Among the participants, 111 were previously randomized to placebo during the double-blind phase of the RCT and then switched to eszopiclone for the open-label period (the placebo-eszopiclone (PBO-ESZ) group). The remaining 360 participants remained on eszopiclone for the open-label period (ESZ-ESZ group). Overall, 19 percent withdrew (89/471) from the study for any reason. Approximately 4 percent withdrew due to adverse effects. Incidence of withdrawal due to adverse effects and of treatment-related adverse effects was higher in the PBO-ESZ group compared with the ESZ-ESZ group (6% and 44% vs. 3% and 28%, respectively). The most common reasons for withdrawal due to adverse effects were unpleasant taste and anxiety, reported in two participants each. Among the 471 participants, the most common adverse effects considered treatment-related were unpleasant taste (7%), headache (5%), somnolence (4%), abnormal dreams and dizziness (3% each). Incidence of was much greater in the unpleasant taste in the PBO-ESZ group compared with the ESZ-ESZ group (20% vs. 3%). A serious adverse effect was reported for 11 participants (2%) leading to study withdrawal in two participants. These events included chest pain, accidental injury, atrial fibrillation, and diabetes.

An older open-label study conducted in France evaluated zolpidem use in 107 adults with insomnia over 6 months.148 The initial dose of zolpidem was 20 mg, but the dose could be adjusted downward or upward according to efficacy and tolerability. Mean age of the participants was 63 years, and 69 percent were female. The trial was not completed by 19 percent (20/107) of the participants, with adverse effects accounting for 37 percent (7/20) of the withdrawals. Among the seven participants withdrawing due to adverse effects, two withdrew during the 7-day placebo run-in period before active treatment was initiated. Reasons for withdrawal were not indicated. There were 42 adverse effects experienced by 24 patients (22%) which were possibly or probably associated with treatment. These events included malaise, vertigo, and anterograde amnesia (five events each). All participants reporting vertigo (five) or confusion (two) were at least 70 years of age. An additional 22 participants (27 events) reported adverse effects considered unrelated to the study drug. The five withdrawals during active treatment were due to these events. Specific adverse events were not described.

Specific Benzodiazepines

The previously discussed study by Kripke et al. also compared temazepam alone (n=2076) with no hypnotic use (n=23,671).144 Comparable to any hypnotic prescribed, doses were stratified into tertiles of 1-240 (n=798), 240-1640 (n=613), and >1640 (n=665) mg per year. Temazepam use was associated with increased hazard of death for all tertiles, with HRs of 3.71 [95% CI, 2.55 to 5.38], 4.15 [95% CI, 2.88 to 5.99], and 6.56 [95% CI, 5.03 to 8.55], respectively. As with the findings of any hypnotic prescribed, a dose-response association was demonstrated. Increased hazard of incidental major cancers was found in the middle and upper dose tertile groups (HR 1.28 [95% CI 1.03 to 1.59] and 1.99 [95% CI 1.57 to 2.52]).

Melatonin Agonists

Two studies reported longer-term harms related to ramelteon. Adverse effects were common but rarely severe or requiring study withdrawal. However, study withdrawal for any reason was common. One open-label study evaluated the ramelteon in 190 Japanese participants with chronic insomnia.149 The participants had a mean age of 48 years and 69 percent were female. Participants received ramelteon 4 or 8 mg (titrated to 16 mg according to efficacy or lowered due to tolerability issues) for 24 weeks. Seven participants (4%) were withdrawn from the study due to adverse effects. Types of adverse effects that led to withdrawal were not described. An additional 21 participants withheld or discontinued treatment due to adverse effects. A total of 358 adverse effects were reported by 147 participants (77%), most deemed mild in severity. The most common specific adverse events were nasopharyngitis (24%), upper respiratory tract infections (6%), eczema (6%), and headache (4%).There were two serious adverse effects that required hospitalization: pyelonephritis and synovitis.

One open-label study conducted in the United States evaluated the long-term use of ramelteon in 1213 participants with chronic insomnia.146 The participants were divided into two groups. The adult group of 965 participants was aged 18 to 64 years and received ramelteon 16 mg over a 48-week treatment phase. The older adult group, aged at least 65 years, received ramelteon 8 mg. Most participants were female (59%). In the adult group, 62 percent had withdrawn for any reason by the end of the 48-week interval. Primary reasons for withdrawal were adverse effects (12%) and lack of efficacy (18%). Adverse effects associated with study withdrawal were not reported. For adult participants taking ramelteon for 6 months or 1 year, 81 percent reported at least one adverse effect at both time intervals. The most common adverse effects included nasopharyngitis (14% at 6 months, 15% at 1 year) and headache (13% and 14%). Somnolence was reported for 8 percent of the participants at both intervals. Two participants in the 18 to 64 year adult group died in motor vehicle accidents (neither was reported driving). Other serious adverse effects included prolactinoma and brain neoplasm in one participant each, and uterine fibroids in three participants. The prolactinoma was considered possibly treatment related.

In the older adult group, 58 percent had withdrawn for any reason by the end of the 48-week interval. Primary reasons for withdrawal were adverse effects (12%) and lack of efficacy (25%). For the older adult participants taking ramelteon for 6 months or 1 year, the incidence of at least one adverse effect was 83 and 85 percent, respectively. The most common adverse effects included nasopharyngitis (10% at 6 months, 11% at 1 year) and somnolence (9% and 10%). One participant was diagnosed with bladder cancer and one with colon cancer.

Efficacy and Comparative Effectiveness of Complementary and Alternative Medicine Treatments

Key Points

  • A previous high quality systematic review found insufficient evidence on the efficacy of acupuncture as a treatment alone or as an adjunctive treatment. Updating results from this review, we conclude that the evidence remains insufficient to draw conclusions about the efficacy of acupuncture used alone or as an adjunctive treatment for insomnia disorder.
  • A variety of other complementary and alternative interventions have been studied with RCTs and systematic reviews to determine efficacy in treating insomnia. These include homeopathy, valerian, bright light therapy, isoflavones, magnesium supplementation, chamomile extract, Simillimum, and Wuling capsule. Evidence is insufficient to draw conclusions regarding their efficacy in treating insomnia disorder because similar comparisons across studies did not exist and trials were often small with methodologic limitations.

Efficacy of Acupuncture

Overview of Included Studies

We identified one relevant systematic review addressing efficacy of acupuncture for insomnia disorder that was of sufficient quality to include in lieu of de novo extraction (Table 25). Cheuk et al.27 searched bibliographic databases through October 2012, had no language restrictions, distinguished different types of acupuncture, and included 33 primary studies. Twenty one of 33 trials included trials involved treatments lasting 4 or more weeks.

Table 25. Efficacy of acupuncture: description and conclusions from previous systematic review.

Table 25

Efficacy of acupuncture: description and conclusions from previous systematic review.

We identified two RCTs assessing the efficacy of acupuncture for insomnia that were not included in the previous systematic review (Table 26).151,152 Hatchel et al. randomized participants to acupuncture or sham acupuncture and had moderate risk of bias; the study was underpowered but could be pooled with one comparison in the previous systematic review.151 (Acupuncture versus sham acupuncture was included in the Cheuk et al. review.) We also identified one trial that assessed acupuncture as an adjunct therapy.152 Adjunctive acupuncture versus other treatment alone was compared in Cheuk et al. Huo et al.152 randomized participants to acupuncture using meridian and Anmian acupoints or to acupuncture using only meridian acupoints, had moderate risk of bias, and can be used to update the Cheuk et al. analysis for one outcome.

Table 26. Efficacy of acupuncture in the general adult population: overview and strength of evidence.

Table 26

Efficacy of acupuncture in the general adult population: overview and strength of evidence.

Hachul et al.151 was conducted in Brazil, enrolled only females, randomized 18 participants, and had a study duration of 5 weeks.151 Huo et al., a 4-week study, was conducted in China.152

Global Outcomes

Cheuk et al.,27 Hachul et al.,151 and Huo et al.152 reported PSQI scores. In the trial of acupuncture versus sham acupuncture, Hachul et al.151 found no significant differences in PSQI scores and no significant change from baseline in either group. In contrast, in the trial of acupuncture at meridian and Anmian acupoints versus at meridian acupoints alone, Huo et al.152 found significantly better (lower) PSQI scores with acupuncture at meridian and Anmian acupoints (5.49 vs. 7.77), but no significant improvements from baseline within either group. Updating the Cheuk et al. review strengthens the evidence for these two comparisons (Figures 55 and 56). However, because all the trials in that review were rated high risk of bias, we maintain that this evidence is insufficient.

Figure 55 is a forest plot displaying data presented in Table 26 comparing the efficacy of acupuncture and sham acupuncture for the outcome PSQI score in the general adult population. When data from eight studies were pooled, the mean PSQI score was 2.11 lower (better) in the acupuncture group than the sham acupuncture group (significant).

Figure 55

Efficacy of acupuncture in the general adult population: PSQI score.

Figure 56 is a forest plot displaying data presented in Table 26 comparing the efficacy of adjunctive acupuncture and adjunctive sham acupuncture for the outcome PSQI score in the general adult population. When data from four studies were pooled, the PSQI score was 2.50 lower (better) in the adjunctive acupuncture than the adjunctive sham acupuncture group (significant).

Figure 56

Efficacy of adjunctive acupuncture in the general adult population: PSQI score.

Sleep Outcomes

Neither of the studies published since Cheuk et al. reported sleep outcomes; therefore, we could not update those outcomes.

Functioning, Mood, and Quality of Life

In their trial of acupuncture versus sham acupuncture, Hachul et al.151 reported the Beck Depression Inventory but found no significant difference between groups in scores (33.28 vs. 32.5) and no significant improvement from baseline within either group. Hachul et al.151 also reported the World Health Organization Quality of Life score. They found no significant difference between treatment groups in any component and significant improvement from baseline for only the psychological component within the acupuncture group. In their trial of acupuncture at meridian and Anmian acupoints versus meridian acupoints alone, Huo et al.152 found significantly better therapeutic efficacy and lower self-rating depression scores (25.53 vs. 30.80) but not self-rating anxiety scores (31.23 vs. 32.00) for meridian and Anmian acupoints. Self-rating depression scores and self-rating anxiety scores improved significantly from baseline within both groups. Huo et al.152 also found significantly better treatment efficacy in the meridian plus Anmian acupuncture group.

Adverse Effects

Fewer than half of the studies included in Cheuk et al. reported adverse effects, and the adverse effects that were reported were minor. Compared with an intervention used by an RCT in this review, the Cheuk et al. systematic review27 found no significantly greater risk for adverse effects for needle acupuncture versus placebo or sham acupuncture (OR 3.19 [95% CI, 0.12 to 84.43]) or for needle acupuncture with other treatment versus other treatment alone (OR 0.05 [95% CI, 0.00 to 1.03]).

Huo et al.152 reported withdrawals by treatment group. No withdrawals occurred in either treatment group. Updating the data from the systematic review provides insufficient evidence to draw conclusions about the rates of adverse effects between groups.

Efficacy of Homeopathy

Overview and Summary of Previous Systematic Review

We identified one relevant systematic review that examined homeopathy for insomnia.153 The review was assessed as having fair quality and was therefore used in lieu of de novo extraction. Cooper et al. identified five RCTs of homeopathy for insomnia, all had high risk of bias; they identified another RCT in an update.153,154 Only one RCT showed a significant difference in the sleep impairment index with homeopathy compared with placebo. Evidence is insufficient to draw conclusions about the efficacy of homeopathy for treating insomnia disorder.

One additional trial of homeopathy that was not included in the previous systematic review was identified.155 This trial studied a population different than other studies, so is not used to update results from the previous systematic review. Harrison et al. randomized South African men between the ages of 18 and 40 to homeopathic complex or placebo before supper and at bedtime for 4 weeks.155 The authors report that an intergroup analysis showed a significant difference in sleep onset latency (median 7 minutes lower in the experimental group). Overall withdrawals did not differ significantly between homeopathic complex and placebo.

Efficacy of Valerian

Overview and Summary of Previous Systematic Review

We identified one relevant systematic review that examined valerian for insomnia.156 The review was assessed as having fair quality and was therefore used in lieu of de novo extraction (Table 27). Taibi et al.156 identified 29 clinical trials and eight open-label studies of valerian for insomnia. Most studies found no significant difference in sleep outcomes between valerian and the control treatment.

Table 27. Efficacy of complementary and alternative medicine treatments: description and conclusions from previous systematic reviews.

Table 27

Efficacy of complementary and alternative medicine treatments: description and conclusions from previous systematic reviews.

Efficacy and Comparative Effectiveness of Bright Light Therapy

Overview and Summary of Studies

We identified two trials that compared different exposures to bright light for insomnia disorder.157,158 Evidence for all populations and outcomes was insufficient to draw conclusions because no two studies analyzed similar comparisons.

Friedman et al. randomized 61 older adults to bright (∼4,000 lux) or dim light in the morning or evening and reports on 51 completers.157 Mean age was 64.0 and 69 percent were female; mean insomnia duration was 15 years. Friedman et al. found that mean sleep onset latency and total sleep time were significantly different at both 3 and 6 months post-treatment. Subjective measures were similar in bright light and dim light groups postintervention.

Kirisoglu et al. randomized older adults to 20 or 45 minutes of daily exposure to 10,000 lux for 60 days.158 Longer exposure (45 minutes compared with 20 minutes daily) resulted in shorter sleep latencies and longer total sleep times. Outcomes measured at 3 months and 6 months showed that exposure to 45 minutes of bright light was associated with shorter sleep onset latency and longer total sleep time.

Efficacy of Other CAM Treatments

Overview and Summary of Other Eligible CAM Trials

We identified four trials of other complementary and alternative medicine interventions that met our inclusion criteria and were not included in one of the eligible previous systematic reviews.159-162 Interventions included Wuling capsule, isoflavones, magnesium supplementation, and chamomile extract. Evidence for these interventions is insufficient to draw conclusions about their efficacy in treating insomnia disorder.

Lin et al. randomized volunteers from the general adult population to three Wuling capsules or placebo three times a day for 4 weeks.161 Lin et al. found no significant difference between Wuling capsule and placebo in PSQI.161 Lin et al.161 found no significant difference between Wuling capsule and placebo groups in physical, psychological, social, or environmental domains of the World Health Organization Quality of Life Brief Scale. No significant differences were seen in overall withdrawals, withdrawals due to adverse effects, or the proportion of participants with at least one adverse effect between Wuling capsule and placebo groups. One participant withdrew from the Wuling capsule group because of an adverse effect. The most common adverse effects were dry mouth, dizziness, constipation, stomach bloating, stomach pain, and diarrhea.

Hachul et al. randomized post-menopausal females aged 50 to 65 to isoflavone 80 mg or placebo (frequency not reported); Hachul et al.160 found a smaller proportion of females reported moderate or intense insomnia with isoflavone than with placebo. Overall withdrawals by treatment group, withdrawals due to adverse effects, or the proportion of participants with at least one adverse effect were not reported.

Abbasi et al. randomized 46 older adults to 500 mg magnesium or placebo daily for 8 weeks.159 Compared with placebo, magnesium supplementation improved ISI scores, decreased sleep onset latency, and increased sleep efficiency. Total sleep time remained similar across groups.

Zick et al. randomized 34 patients ages 18 to 65 to 270 mg chamomile twice daily or placebo for 28 days. Sleep and daytime functioning outcomes were similar with chamomile and placebo postintervention.162

Comparative Effectiveness of Interventions of Different Types

We identified several trials that assessed the comparative effectiveness of interventions across intervention classes (psychological versus pharmacologic) or combination treatments across intervention classes.

Comparative Effectiveness of Pharmacologic Versus Psychological Interventions and Combination Treatments

Key Points

  • Evidence was insufficient to draw conclusions regarding the comparative effectiveness of CBT-I versus hypnotic medication.

Overview of Included Studies

We identified three trials with moderate risk of bias that compared CBT-I to a commonly used sleep medication or the combination treatment to either CBT-I or drug therapy alone and/or CBT-I alone (Table 28).60,72,74,163

Table 28. Comparative effectiveness of drug versus CBT or combined drug/CBT: overview and strength of evidence.

Table 28

Comparative effectiveness of drug versus CBT or combined drug/CBT: overview and strength of evidence.

Zolpidem Versus CBT-I or Combined Zolpidem/CBT-I Therapy

One RCT compared nonbenzodiazepine zolpidem with CBT-I and combined zolpidem and CBT-I therapy.60 A total of 63 participants were randomized, 15 each in the zolpidem, CBT-I, and placebo arms, and 18 in the combined therapy arm. Among the 48 participants randomized to zolpidem, CBT-I, and the combined therapy arms, mean age was 47 years, and 69 percent were female. Mean duration of insomnia was 10 years. Zolpidem 10 mg was administered nightly for 28 days, then 5 mg nightly for 7 days, and then 5 mg was taken every other night for the next 7 days. The trial was conducted in the United States and received support from industry. Jacobs et al. had moderate risk of bias.

Jacobs et al. did not report global outcomes. Post-treatment following 8 weeks of therapy, there were no differences in sleep onset latency and TST between the zolpidem, CBT-I, or combined therapy groups. Evidence was insufficient for both outcomes. There were significantly more participants in the CBT-I group who met the considered normal sleep criterion of a sleep latency of 30 minutes or less compared with the zolpidem group, 57 percent (8/14) versus 15 percent (2/13), respectively. The proportions of participants who met the considered normal sleep criterion of a sleep efficiency of 85 percent or more were not significantly different among the three treatment groups. There were eight withdrawals among the three arms of interest, two in the zolpidem group (13%), five in the combined therapy group (28%), and one CBT-I participant (7%). None of the withdrawals were attributed to adverse effects. Specific adverse effects were not reported. Strength of evidence was insufficient for withdrawals and adverse effects. At the 12-month followup assessment, improvements in sleep outcomes were maintained in the CBT group. Outcomes in the temazepam group were not reported.

Temazepam Versus CBT-I or Combined Temazepam/CBT-I Therapy

Two RCTs compared benzodiazepine temazepam with CBT-I and combined temazepam and CBT-I therapy.72,74

Wu et al. randomized 77 participants, 20 in the temazepam and 19 each in the CBT-I, combined therapy, and placebo arms. Demographic information was not reported for the temazepam, CBT-I, or combined therapy arms separately, but among the four treatment arms, the mean age was 38 years, and 53 percent were female. Temazepam recipients initially received 7.5 mg nightly with gradual increases up to 30 mg and then a decrease to 15 mg in the last treatment week for a total of 8 weeks. The trial was conducted in China and had government funding. Wu et al. had moderate risk of bias. Global outcomes were not reported. Post-treatment, temazepam was better than CBT-I in reducing sleep onset latency and increasing TST. There was no difference in sleep efficiency between the two groups. Evidence was insufficient for all outcomes. Insufficient evidence found no differences in sleep outcomes between the temazepam and combined therapy groups. Post-treatment, the proportions of participants who met normal sleep criteria, based on a sleep-onset latency ≤30 minutes and sleep efficiency ≥85 percent, were not significantly different among the three treatment groups. There were no differences among the three groups in the daytime dysfunction component of the PSQI. There were no significant differences in overall withdrawals or withdrawals due to adverse effects among the three groups. Three participants in the temzepam group withdrew due to adverse effects (15%). Specific adverse effects were not reported. Strength of evidence was insufficient for withdrawals and adverse effects. At the 8-month followup assessment, improvements in sleep outcomes were maintained in the CBT group while outcomes in the temazepam and combined therapy groups regressed to pretreatment conditions.

Morin et al. randomized 78 participants, 20 each in the temazepam, combined therapy, and placebo arms and 18 in the CBT-I arm.74 Compared with Wu et al., the participants were older; the mean age was 65 years. Most participants were female (64%) and white (90%). The mean duration of insomnia was 17 years. Temazepam recipients initially received 7.5 mg nightly with gradual increases up to 30 mg as needed (participants were to use sleep medication at least 2 to 3 nights per week, but medication was made available all 7 nights) for a total of 8 weeks. The trial was conducted in the United States and had government sponsorship. Morin et al. had moderate risk of bias. Self-rated global improvements were greater in the combined therapy and CBT-I groups compared with the temzepam group at post-treatment. There were no differences in sleep outcomes assessed between the temazepam and CBT-I groups. TST was greater in the temazepam group compared with the combined therapy group. However, there was an imbalance in baseline TSTs, 340 minutes for the temazpam group versus 290 minutes in the combined therapy group and the mean changes from baseline to post-treatment were comparable between groups, approximately 40 minutes. Strength of evidence was insufficient for all outcomes. Four participants in the three active treatment arms withdrew from the trial, three in the temazepam group (due to adverse effects), one in the combined therapy group, and none in the CBT-I group. Specific adverse effects were not reported within this publication. Strength of evidence was insufficient for withdrawals and adverse effects. Long-term (24-month followup), improvements in sleep out outcomes were maintained in the CBT group but not in the temazepam group.

Comparative Effectiveness of Combined Pharmacologic and Psychological Interventions Versus Psychological Interventions

Key Points

  • Evidence was mostly insufficient to draw conclusions regarding the comparative effectiveness of combined hypnotic medication and CBT-I versus CBT-I alone.

Overview of Included Studies

We identified four trials with moderate risk of bias that compared combined drug and CBT-I therapy to CBT-I alone (Table 29).60,72,74,163

Table 29. Comparative effectiveness of combined drug and CBT-I versus CBT-I: overview and strength of evidence.

Table 29

Comparative effectiveness of combined drug and CBT-I versus CBT-I: overview and strength of evidence.

Combined Zolpidem/CBT-I Therapy Versus CBT-I

Two RCTs compared combined zolpidem and CBT-I therapy to CBT-I therapy alone 60,163 Morin et al. randomized 160 adults, 80 to combined CBT-I with 10 mg of zolpidem taken daily at bedtime and 80 to CBT-I alone.163 Zolpidem was taken during the initial 6 weeks of therapy. Mean age was 50 years and 61 percent were female. The mean duration of insomnia was 16 years. The baseline ISI score was 17, indicating moderate severity. The trial was conducted in Canada. Morin et al. had moderate risk of bias. The 8-week trial by Jacobs et al. randomized 18 adults to combined CBT-I with 10 mg of zolpidem taken daily and 15 adults CBT-I alone.60 Mean age was 48 years, and 69 percent were female. Mean duration of insomnia was 10 years. The trial was conducted in the United States and received support from industry. Jacobs et al. had moderate risk of bias.

At week 6, the proportions of participants who responded to treatment, defined as an ≥8 point reduction in ISI scores, did not differ between groups (61% for combined vs. 60% for CBT-I alone. Proportions of participants who remitted to treatment, defined as an ISI score <8 points, did not differ between the combined and CBT-I alone groups (44% vs. 39%).

The mean difference in ISI scores at 6 weeks of was -0.50 points (95% CI, -1.58 to 0.58) between groups. Mean reductions in ISI scores were 8.8 and 8.3 points for the combined and CBT-I alone groups, respectively. Strength of evidence for global outcomes was insufficient. Patients rated moderately or markedly improved by an independent assessor also did not differ between groups, 83 percent (60/72) versus 89 percent (66/74) in the combined and CBT alone groups, respectively. Jacobs et al. did not report global outcomes.

Overall, combined therapy was not better than CBT-I alone in improving sleep onset latency, TST, or sleep efficiency. Strength of evidence was low to insufficient. Morin et al. reported combined therapy significantly improved WASO by 14 minutes compared with CBT-I alone. Morin et al reported 11 withdrawals during the initial 6 weeks of therapy, six in the combined arm and five in the CBT-I alone arm. No adverse effects were reported. Jacobs et al. reported five withdrawals in the combined therapy group (28%) and one CBT-I participant (7%). None of the withdrawals were attributed to adverse effects. Specific adverse effects were not reported.

Combined Temazepam/CBT-I Therapy Versus CBT-I

Two RCTs compared combined temazepam and CBT-I therapy with CBT-I alone.72,74

Wu et al. randomized 38 participants, 19 each in the combined therapy and CBT-I arms. Demographic information was not reported for the temazepam, CBT-I, or combined therapy arms separately, but among the four treatment arms, the mean age was 38 years, and 53 percent were female. Temazepam 7.5 mg nightly was initially administered with gradual increases up to 30 mg and then a decrease to 15 mg in the last treatment week for a total of 8 weeks. The trial was conducted in China and had government funding. Wu et al. had moderate risk of bias. Global outcomes were not reported. Post-treatment, combined therapy was better than CBT-I in reducing sleep onset latency. There was no difference in TST and sleep efficiency between the two groups. Evidence was insufficient for all outcomes. Post-treatment, the proportions of participants who met normal sleep criteria, based on a sleep-onset latency ≤30 minutes and sleep efficiency ≥85 percent, were not significantly different between the combined and CBT-I groups (50% vs. 36%). There were no differences among the groups in the daytime dysfunction component of the PSQI. Only one withdrawal was reported in the combined group, none in the CBT-I arm. There were no withdrawals due to adverse effects. Specific adverse effects were not reported. Strength of evidence was insufficient for withdrawals and adverse effects.

Morin et al. randomized 38 older adults, 20 in the combined temazepam and CBT-I arm and 18 in the CBT-I arm.74 The mean age was 65 years and most participants were female (68%) and white. The mean duration of insomnia was 18 years. Temazepam was initially administered at 7.5 mg nightly with gradual increases up to 30 mg as needed (participants were to use sleep medication at least 2 to 3 nights per week, but medication was made available all 7 nights) for a total of 8 weeks. The trial was conducted in the United States and had government sponsorship. Morin et al. had moderate risk of bias. There were no differences in sleep outcomes between the combined and CBT-I groups. Strength of evidence was insufficient for all outcomes. One withdrawal was reported in the combined therapy group, none in the CBT-I group. Specific adverse effects were not reported within this publication. Strength of evidence was insufficient for withdrawals and adverse effects.

Comparative Effectiveness and Combination Treatments: Unique Comparisons

Key Points

  • Evidence was insufficient to draw conclusions regarding the efficacy or comparative effectiveness of CBT-I versus Tai Chi, acupuncture versus estazolam, or the adjunctive efficacy of a traditional Chinese medicine approach combined with sleep medication.

Overview of Included Studies

Wang et al. randomized 90 insomnia disorder patients to an intervention based upon traditional Chinese medicine called Low Resistance Thought Induction Sleep-regulating Technique combined with 1-2 mg estazolam nightly or estazolam alone.164

Guo et al. randomized 180 patients with insomnia disorder to three arms, verum acupuncture plus placebo, estazolam plus sham acupuncture, or sham acupuncture plus placebo.165

Irwin et al randomized 123 older adults to CBT-I, Tai Chi Chih, or a sleep seminar education control.76

Views

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...