Overview of included studies

We identified 2570 citations from literature searches, reviews of reference lists, and citations from dossiers submitted by 2 pharmaceutical manufacturers, Sanofi-Aventis (zolpidem extended release) and Takeda (ramelteon). After applying the eligibility and exclusion criteria to titles and abstracts, we obtained the full text of 370 publications. After reapplying the criteria for inclusion, we included 205 studies. The flow of study inclusion and exclusion is detailed in Figure 1. A list of excluded trials is provided in Appendix C.

Figure 1

Figure 1

Newer drugs for insomnia: Results of literature search

We included eight head-to-head trials (Table 2).12–19 One trial is published as a poster presentation only;18 additional details were provided by the manufacturer and in the FDA review of eszopiclone.20

Table 2

Table 2

Total numbers of head-to-head trials of newer drugs for insomnia

We included 44 trials in 45 publications comparing newer insomnia drugs compared with benzodiazepines.21–65 Appendix D summarizes the efficacy, safety, and rebound insomnia results of these studies.

We identified two trials comparing trazodone compared with a sedative hypnotic; one (compared with zaleplon)52 was rated poor quality and the other (compared with zolpidem)61 was rated fair.

Sixty-one placebo-controlled trials were included. (Some publications described more than one trial).5, 20, 66–92 93–123

Four good-quality systematic reviews of newer sedative hypnotics were included.1, 124–126 The review most relevant to this report is a comparative review conducted by the National Institute for Clinical Excellence.124 The others were not designed to compare the sedative hypnotics head-to-head. One meta-analysis examined the risks and benefits of sleep agents, including newer sedative hypnotics, in older people with insomnia.126

We included 20 observational studies of adverse events associated with newer drugs for insomnia.127–146

Key Question 1. What is the comparative effectiveness of newer drugs in treating adults and children with insomnia?

Summary of the Evidence

There is no evidence in children.

Direct evidence

  • Direct evidence is from 8 short-term head-to-head trials (7 fair quality, 1 poor; 1 measured withdrawal effects only)
  • Eszopiclone compared with zolpidem (1 trial)
    • - There was no significant difference between eszopiclone 2 mg or 3 mg and zolpidem 10 mg on polysomnography-measured sleep latency, WASO, or number of awakenings.
    • - Both active treatments were more effective than placebo for polysomnography-measured sleep latency
    • - Eszopiclone 2.5 mg and 3 mg were significantly better than placebo for polysomnography-measured WASO
    • - Eszopiclone 1mg or 2 mg (lower doses) and zolpidem 10 mg were no different from placebo on polysomnography-measured WASO
  • Zaleplon compared with zolpidem (4 trials)
    • - Zaleplon and zolpidem were similarly effective for subjective sleep latency in elderly patients and in patients under age 65
    • - There is evidence from 2 head-to-head trials that zaleplon is less likely than zolpidem to cause rebound insomnia in adults under age 65
  • Zolpidem compared with zopiclone (2 efficacy trials, 1 poor quality)
    • - In a fair-quality trial, zolpidem and zopiclone were similarly effective in investigator and patient global assessments of improvement
    • - Subjective sleep outcomes (latency, frequency of awakening, sleep duration) were improved from placebo to a similar extent in both treatment groups
    • - A trial of simulated driving performance was rated poor quality

Indirect evidence

  • Adjusted indirect analysis of 22 placebo-controlled trials found few differences between drugs on subjective sleep outcomes
  • Sleep latency was shorter with eszopiclone than ramelteon (mean difference −11 minutes; 95% CI −21 to −1.3 minutes)
  • Sleep duration was an average of 37 minutes longer with eszopiclone compared with ramelteon (95% CI 17 to 56 minutes)
  • Comparing only manufacturers’ recommended initial doses, there were no differences between drugs on any outcome, with the exception of longer sleep duration with eszopiclone compared to ramelteon (mean difference 28.9 minutes; 95% CI 6.2 to 51.7 minutes)
  • Excluding poor-quality studies, eszopiclone significantly increased sleep duration compared with zolpidem (mean difference 37.1 minutes; 95% CI 23.9 to 50.4 minutes)
  • In a subgroup analysis of elderly patients, eszopiclone significantly increased sleep duration compared with zolpidem (mean difference 32.8 minutes; 95% CI 1.2 to 64.4 minutes)
Zolpidem extended-release
  • There are no head-to-head trials comparing zolpidem extended-release with other newer insomnia drugs
  • Evidence of efficacy comes from 3 fair-quality placebo-controlled trials. We were unable to include data from these trials in the adjusted indirect analysis.
  • In adults under age 65, polysomnography- measured WASO was significantly shorter than placebo on nights 1 and 2 of treatment but not on nights 15 and 16. A post hoc analysis found that WASO was significantly better than placebo through hour 6, although not at hours 7 and 8
  • Results for subjective sleep outcomes were mixed, with zolpidem extended-release showing superiority to placebo at some, but not all, assessment points
  • In patients over age 65, polysomnography-measured WASO was shorter than placebo through the first 6 hours of the night
  • In a 6-month study of intermittent treatment (3 to 7 nights per week), 90% of patients taking zolpidem extended-release said the treatment helped them sleep, compared with 51% of the placebo group

Detailed Assessment

Direct evidence

Patient and study design characteristics of included head-to-head trials are shown in Table 3. No new head-to-head trials were identified for Update #2.

Table 3

Table 3

Head-to-head trials of newer insomnia drugs: Study design and patient characteristics

Eszopiclone compared with zolpidem

One head-to-head trial compares eszopiclone compared with zolpidem.20, 147 According to the study funder, the objective of the study was to evaluate the polysomnographic efficacy and safety of eszopiclone relative to placebo. Zolpidem 10 mg was included as an active control to allow qualitative comparisons to eszopiclone. The primary efficacy outcome was latency to persistent sleep as measured by polysomnography. The study compared 4 doses of eszopiclone (1 mg, 2 mg, 2.5 mg, and 3 mg) to placebo and zolpidem 10 mg in a crossover design over 2 nights of treatment.

Both drugs were more effective than placebo for the primary outcome of polysomnography-measured sleep latency. Eszopiclone 2.5 mg and 3 mg were more effective than placebo for polysomnography-measured WASO, but there was no difference from placebo for eszopiclone at lower doses or for zolpidem 10 mg. There was also no difference between zolpidem and eszopiclone on subjective measures of next-day effects, including morning sleepiness, daytime alertness, and daytime ability to function.20

The main analysis in this study compared eszopiclone with placebo; no analysis comparing eszopiclone with zolpidem was presented. To make a direct comparison between the two drugs, we calculated the weighted mean difference between eszopiclone and zolpidem for polysomnography-measured sleep outcomes using data provided in the FDA review of eszopiclone. Results of these analyses are shown in Table 4.20 There were no significant differences between eszopiclone and zolpidem on polysomnography-measured sleep latency, WASO, or number of awakenings.

Table 4

Table 4

Head-to-head comparison of eszopiclone compared with zolpidem on polysomnography-measured outcomes

Subjective measures were also reported, but standard deviations were not provided, so we could not calculate a mean difference.

Zaleplon compared with zolpidem

Four fair-quality head-to-head studies compared zolpidem with zaleplon and placebo.12, 14, 15, 17 Two of these were conducted in adults under age 65 and had identical designs.14, 15 Another was conducted in older adults.12 The fourth head-to-head study17 was a small, single-dose crossover trial that measured patient preference as a primary outcome.

In the 3 studies with sleep outcomes, comparisons between zaleplon and placebo were the primary comparisons. Published reports do not provide a head-to-head analysis comparing zaleplon with zolpidem, and it was not possible to conduct an analysis of zaleplon compared with zolpidem from data provided.

Sleep latency

Sleep latency was the primary outcome in two studies in adults.14, 15 Both compared zaleplon at three fixed doses (5 mg, 10 mg, or 20 mg) with zolpidem 10 mg for 4 weeks. A placebo arm was also included, and analyses are presented for the comparison to placebo. Neither publication provided a head-to-head analysis of zolpidem compared with zaleplon, but a head-to-head analysis is provided in the FDA statistical review of zaleplon5 for one trial.15 At weeks 1 through 4,15 there was no difference between zaleplon 5 mg or 10 mg and zolpidem 10 mg on the median number of minutes to sleep onset. The only significant difference between the drugs on this outcome was a shorter latency with zaleplon 20 mg compared with zolpidem 10 mg. There was no difference in the comparison of recommended starting doses zaleplon 10 mg and zolpidem 10 mg. These results are not from intention-to-treat analyses.

For the second trial,14 intention-to-treat results using the last observation carried forward method are presented in the FDA review of zaleplon.5 Analyses compared zaleplon with placebo. Results were mixed. Zaleplon at all three doses had a shorter latency than placebo at all time points, with the exception of 5 mg at week 4. For zolpidem 10 mg, at weeks 2 and 3 latency was significantly shorter than for placebo but was not significantly different at week 4. At week 1, there was a trend for shorter latency, but this was not significant (−10 minutes; P=0.07).

In a 2-week head-to-head trial of zaleplon 5 mg or 10 mg compared with zolpidem 5 mg conducted in 549 older adults (65 years or older),12 results were similar to those of the trials in younger patients. There was no difference in sleep latency for zaleplon 5 mg and zolpidem 5 mg, but zaleplon at a higher dose (10 mg) was associated with a shorter latency than zolpidem 5 mg. Zolpidem, but not zaleplon, was associated with rebound sleep latency on the first night of discontinuation.

Sleep duration

Duration of sleep was a secondary outcome in three head-to-head trials of zaleplon compared with zolpidem.12, 14, 15 Zolpidem 5 mg and 10 mg increased sleep duration more than placebo in all three studies. In two studies in adults, zaleplon 5 mg and 10 mg were no different from placebo on this outcome at any time period. Zaleplon 20 mg was more effective than placebo at weeks 1 and 3, but not weeks 2 and 4.

Number of awakenings

The difference from placebo in the median number of awakenings during the night was another secondary outcome in head-to-head trials. In one trial,14 there was no difference from placebo for any dose of either zaleplon or zolpidem at any time point. The other trial in adults15 had mixed results. Zaleplon 5 mg and 10 mg was no different from placebo, zaleplon 20 mg was more effective than placebo at weeks 2, 3, and 4, and zolpidem 10 mg was better than placebo at weeks 1, 2, and 3. In older adults, only zolpidem 5 mg was more effective than placebo.12

Sleep quality

In a pooled analysis of three trials of zaleplon compared with zolpidem,12, 14, 15 the National Institute for Clinical Excellence review124 found that patients on zaleplon were less likely to experience improvement in sleep quality at the end of treatment than patients taking zolpidem (odds ratio 0.66; 95% CI 0.51 to 0.87).

Rebound insomnia

Two head-to-head trials found zolpidem 10 mg to be associated with more rebound insomnia than zaleplon as measured by an increase in median sleep latency on the first night after discontinuation.14, 15 Zolpidem 10 mg was associated with a 20- to 22-minute increase in sleep latency compared with placebo on the first night of discontinuation. Rebound sleep latency was not seen with zaleplon at any dose. Zaleplon at all doses (5 mg, 10 mg, and 20 mg) was less likely to cause rebound sleep latency than zolpidem 10 mg. The mean difference between zolpidem 10 mg and zaleplon 10 mg was 34 minutes (95% CI 10.5 to 57.5 minutes). Head-to-head studies also found zolpidem to be associated with rebound decrease in sleep duration on the first night of discontinuation. Zaleplon was not associated with rebound on this outcome, except at the 10 mg dose in older adults. In two studies in adults14, 15 zolpidem, but not zaleplon, was associated with an increase in awakenings compared to placebo on the first night after withdrawal. In older adults, neither drug was associated with rebound insomnia on this measure.12

Other outcomes

A small (N=53) single-dose crossover study comparing zolpidem 10 mg with zaleplon 10 mg was designed to measure patient preference for a drug as the primary outcome.17 This was measured by a questionnaire filled in by the patient the evening following administration of the drug. More patients preferred zolpidem, but the difference was not statistically significant (62% compared with 32%; P=0.81). Secondary outcomes were mean scores on the Leeds sleep evaluation questionnaire, and “day quality,” a visual analog scale (0–100, higher is better) measuring 7 factors on the day following the administration of the drug. Zolpidem patients improved more on two of four factors on the Leeds sleep evaluation questionnaire (Getting to Sleep and Quality of Sleep); there was no difference between drugs on the other two factors (Ease of Waking Up and Behavior Following Wakefulness). Only one of 7 factors of the day-quality measure was significantly different between drugs. Zolpidem patients reported better quality of sleep (mean score 68.8 compared with 50.2, P<0.0001), but there were no differences on other factors.

Zopiclone compared with zolpidem

Two fair-quality13, 16 and one poor-quality study19 compared zolpidem with zopiclone. One was designed to assess the effect of withdrawal in patients already taking the drugs for insomnia and did not report efficacy outcomes; it is discussed under Key Question 2.13

A two-week, double-blind trial in 479 patients at multiple centers in Japan16 is the only head-to-head trial of zolpidem compared with zopiclone in which efficacy is the primary outcome. The primary outcome was the investigator’s global assessment of improvement, based on patient sleep diaries and reported as the proportion of patients who were “moderately improved” or “markedly improved.” At the end of treatment, there were no significant differences between treatment groups in the number of patients “markedly improved” (18.7% zolpidem compared with 16.4% zopiclone) or “moderately improved” (49.3% zolpidem compared with 45.2% zopiclone). Patients’ ratings of treatment efficacy were similar and did not differ between treatment groups. Sleep outcomes (sleep onset latency, frequency of awakening, sleep duration, daytime mood, and daytime physical condition) were improved from placebo to a similar extent in both treatment groups, but data were not reported. Rebound insomnia was reported as the percentage of patients with an aggravation of sleep onset latency by one grade or more after 2 weeks of treatment.16 More patients who took zopiclone had rebound insomnia by this definition than those who took zolpidem (15.4% compared with 4.5%, P<0.005).

Another head-to-head study comparing zolpidem with zopiclone measured next-day simulated driving performance as the primary outcome and reported subjective sleep parameters as secondary outcomes.19 This study was rated poor quality because no baseline demographic or clinical data were reported, and so it cannot be determined if groups were comparable at baseline, and there is no information about withdrawals, so it is impossible to determine if an intention-to-treat analysis was conducted.

Indirect evidence: Meta-analysis of placebo-controlled trials

Because of the limited amount of direct evidence available to compare efficacy across drugs, we conducted a meta-analysis of placebo-controlled trials reporting subjective sleep outcomes. Studies were included in the meta-analysis if they reported endpoint mean scores and standard deviations (or the data to calculate these) for subjective sleep latency, sleep duration, number of awakenings, or WASO in patients with chronic insomnia. Studies were excluded if they reported only objective outcomes, if they used an intermittent dosing strategy (for example, treatment as needed), or if they were conducted in patients with acute insomnia or co-morbid conditions such as depression or sleep apnea.

Twenty-two trials contributed data to the meta-analysis; their characteristics are shown in Table 5.5, 15, 29, 32, 40, 61, 71, 72, 80, 84, 90, 96, 106, 110, 111, 118, 119, 122, 123, 148 We included 4 trials of eszopiclone, 4 of ramelteon, 4 of zaleplon, 7 of zolpidem, and 4 of zopiclone (one trial included both zaleplon and zolpidem15). Five trials were conducted in older adults40, 80, 106, 110, 118 and the rest in adults under age 65. Sample sizes ranged from 14 to 848, and treatment durations ranged from 2 nights to 6 months. Three studies were rated poor quality71, 84, 118 and the rest were fair; no study met all quality criteria.

Table 5

Table 5

Placebo-controlled trials included in meta-analysis of subjective sleep outcomes

Subjective sleep latency

Figure 2 shows the pooled estimates of subjective sleep latency from 22 placebo-controlled trials of individual insomnia drugs. Ramelteon, whether at the 8 mg or the 16 mg dose, and zaleplon 5 mg, were not statistically significantly better than placebo. In the remaining studies, the active drug decreased the time to fall asleep by about 13 to 20 minutes compared with placebo. There was significant heterogeneity among the ramelteon 8 mg, zolpidem 10 mg, and zopiclone 7.5 mg trials (P>0.10).

Figure 2

Figure 2

Subjective sleep latency in placebo-controlled trials of newer insomnia drugs

Sleep duration

Figure 3 shows the pooled estimates for subjective sleep duration in 16 placebo-controlled trials reporting this outcome. Eszopiclone was significantly better than placebo for increasing total sleep time. At the 3 mg dose, the difference from placebo was an increase of 46 minutes (95% CI 36 to 56 minutes). Eszopiclone 2 mg increased the total time slept by 24 minutes over placebo (95% CI 10 to 38 minutes). Ramelteon 8 mg and zopiclone 7.5 mg were also significantly better than placebo. For ramelteon, the difference from placebo was only 3 minutes, however. There was significant heterogeneity among the zopiclone studies (P=0.04), and the estimate was imprecise, with a very wide confidence interval (1.79 to 89.84 minutes).

Figure 3

Figure 3

Subjective sleep duration in placebo-controlled trials of newer insomnia drugs

Subjective wake time after sleep onset

Subjective WASO was reported in 9 trials (Figure 4). Only eszopiclone 3 mg was significantly better than placebo on this outcome. Eszopiclone 3 mg shortened the time spent awake after sleep onset by 11 minutes compared with placebo (95% CI −20 to −2 minutes).

Figure 4

Figure 4

Subjective wake time after sleep onset in placebo-controlled trials of newer insomnia drugs

Subjective number of awakenings

Figure 5 shows results from 12 placebo-controlled trials reporting subjective number of awakenings. Only eszopiclone was significantly better than placebo for this outcome. The difference was less than one awakening per night (mean difference −0.52; 95% CI −0.57 to −0.03).

Figure 5

Figure 5

Subjective number of awakenings in placebo-controlled trials of newer insomnia drugs

Results of adjusted indirect meta-analysis

Results of the adjusted indirect meta-analysis are shown in Table 6. Data from varying doses of the same drug were combined for this indirect comparison. There were very few significant differences between the drugs on any outcomes. The exceptions were significantly shorter sleep latency and longer sleep duration with eszopiclone compared to ramelteon. On average, sleep latency was 11 minutes shorter with eszopiclone than ramelteon (95% CI −21 to −1.3 minutes). Sleep duration was an average of 37 minutes longer with eszopiclone than ramelteon (95% CI 17 to 56 minutes). Patients taking eszopiclone had significantly fewer awakenings than those taking zolpidem, but the difference was less than one time per night (mean difference 0.6 times per night; 95% CI −1.0 to −0.2).

Table 6

Table 6

Adjusted indirect meta-analysis: Summary of results

We performed several subgroup analyses to determine if meta-analysis results varied by population or study design characteristics. These analyses were planned a priori.

When studies conducted in adult and elderly patients were analyzed separately, adjusted indirect analysis showed no significant differences between any of the drugs in subjective sleep latency or WASO (Table 7). In elderly patients, sleep duration was significantly longer with eszopiclone than with ramelteon and zolpidem in elderly patients, but there was no difference between any of the drugs in adult patients under age 65.

Table 7

Table 7

Subgroup analysis by elderly and non-elderly adult patients

We also performed a subgroup analysis excluding studies that used doses other than the manufacturers’ recommended initial dose. Recommended initial doses are eszopiclone 2 mg, ramelteon 8 mg, zaleplon 10 mg, zolpidem 10 mg, and zopiclone 7.5 mg. (Note: Although the eszopiclone product label states that eszopiclone can be started at 3 mg if clinically indicated, the 2 mg dose is recommended for ‘most non-elderly adults.’) In this subgroup analysis, there were no differences between any drugs on any outcome, with the exception of longer sleep duration for eszopiclone than ramelteon (mean difference 28.9 minutes; 95% CI 6.2 to 51.7 minutes).

Exclusion of poor-quality studies71, 84, 118 did not change results of the sleep latency, WASO, or number-of-awakenings analyses. In fair-quality studies, eszopiclone significantly increased sleep duration compared with zolpidem (mean difference 37.1 minutes; 95% CI 23.9 to 50.4 minutes).

PSG-measured outcomes in trials of ramelteon

PSG-measured sleep outcomes were reported in three placebo-controlled trials of ramelteon.77, 106, 122

PSG- measured sleep outcomes were measured at weeks 1, 3, and 5 in a trial of ramelteon 8 mg or 16 mg compared with placebo.122 The primary outcome, sleep latency at week 1 was reduced for both the 8 mg (32 minutes) and 16 mg (29 minutes) groups compared to placebo (48 minutes, P<0.001); improvements were also shown at weeks 3 and 5. Total sleep time was improved with ramelteon compared with placebo at weeks 1 and 3 but not week 5. There were no differences in WASO or number of awakenings.

In a crossover study of 2 nights of treatment with ramelteon 4 mg, 8 mg, 16 mg, or 32 mg, all doses of ramelteon resulted in reductions in PSG-measured sleep latency (P<0.001) and increases in total sleep time (P<0.05).77 There were no differences in WASO for any of the treated groups compared to placebo.

In a 2-night crossover study conducted in patients over age 65, there were significant improvements in PSG-measured sleep latency with ramelteon 4 mg (28.7 minutes, P<0.001) and 8 mg (30.8 minutes, P=0.005) compared with placebo (38.4 minutes).106 PSG-measured total sleep time was also improved with ramelteon (359 minutes for 4 mg and 362 minutes for 8 mg compared with 350 minutes for placebo; P=0.018). There was no difference in objective WASO with either dose of ramelteon compared to placebo, and there was an increase in number of awakenings with ramelteon 4 mg (but not with the 8 mg dose).

Zolpidem extended-release

There are no head-to-head trials comparing zolpidem extended-release with other newer drugs for insomnia. Evidence for the efficacy of zolpidem extended-release comes from three fair-quality placebo-controlled trials.89, 115, 121 Additional information is provided in the FDA statistical review of zolpidem extended-release79 Table 8 summarizes the results of these trials. Because they did not report means for subjective sleep outcomes at endpoint, we were not able to include their data in our meta-analysis.

Table 8

Table 8

Placebo-controlled trials of zolpidem extended-release

A placebo-controlled trial of zolpidem extended-release 12.5 mg was conducted in 212 adults with primary insomnia.107 This study included 2 nights of polysomnography recording, 12 nights of outpatient treatment, 2 more nights of polysomnography recording, 5 nights of outpatient treatment, and a 2-night placebo run-out to measure rebound. The primary outcome measure was polysomnography-recorded WASO in the first 8 hours of the night, measured on nights 1 and 2, and nights 15 and 16, with scores averaged over each 2-night period. WASO was significantly shorter with zolpidem-XR than placebo on nights 1 and 2, but not on nights 15 and 16. A post hoc analysis found that WASO was significantly better than placebo through hour 6, but not during hours 7 and 8 of the night, suggesting that the effects of zolpidem extended-release did not persist past 6 hours.79 The publication of this trial reports only 6-hour results.107 The 8-hour results are reported in the FDA review.79

Data for subjective sleep outcomes are reported graphically only. Results for subjective WASO, subjective number of awakenings, subjective sleep duration, and subjective sleep latency were mixed. Zolpidem extended-release was superior to placebo (P<0.05) at some, but not all, assessment points. Both groups had worse outcomes in the sleep laboratory than at home.

A second placebo-controlled trial, with an identical design was conducted in elderly patients.121 In this trial, the primary outcome was polysomnography-measured WASO in the first 6 hours of the night. This was significantly better than placebo at both nights 1 and 2, and nights 15 and 16. WASO through 8 hours was not measured.

There was a rebound effect in both studies after discontinuation on the first night after discontinuation (night 22), but not on night 23.

A third placebo-controlled trial studied the effect of 6 months of treatment with zolpidem extended-release 12.5 mg in 1018 adults with chronic primary insomnia.89 Patients were instructed to take the medication only as needed, but were required to take a minimum of 3 doses per week. By month 6, the mean number of doses patients took per month was 19.6 (SD 5.2; median 20.0). The primary outcome was the patients’ assessment of the treatment’s aid to sleep, measured by the Patient’s Global Impression at week 12. At week 12, 89.8% of patients treated with zolpidem extended-release said the treatment helped them sleep, compared with 51.4% of the placebo group (P<0.0001).

Key Question 2. What are the comparative tolerability and safety of newer drugs for insomnia when used to treat patients with insomnia?

Summary of the Evidence

Direct evidence

  • Eszopiclone compared with zolpidem
    • - In one head-to-head trial, there was no difference between zolpidem and eszopiclone on subjective measures of next-day effects, including morning sleepiness, daytime alertness, and daytime ability to function
  • Zaleplon compared with zolpidem
    • - In head-to-head trials, total withdrawals and withdrawals due to adverse events were similar for zaleplon and zolpidem and increased with longer duration of trials
  • Zolpidem compared with zopiclone
    • - In a study that measured withdrawal effects over 2 weeks, the incidence of adverse events was higher in withdrawal groups than in continued treatment groups but was similar for zolpidem and zopiclone (38% and 41%, respectively)

Indirect evidence

  • There was no increased risk of withdrawal due to adverse events in placebo-controlled trials of eszopiclone, ramelteon, zaleplon, zolpidem, or zopiclone
  • In a pooled analysis of 3 placebo-controlled trials, the risk of withdrawal due to adverse events was higher with zolpidem extended-release than placebo (relative risk 1.93; 95% CI 1.17 to 3.21)
  • Adjusted indirect analysis of placebo-controlled trials found no differences between the newer sedative hypnotics in rates of withdrawals due to adverse events
  • There is no comparative evidence about long-term safety

Detailed Assessment

Direct evidence

Eszopiclone compared with zolpidem

In an unpublished head-to-head trial, there was no difference between zolpidem and eszopiclone on subjective measures of next-day effects, including morning sleepiness, daytime alertness, and daytime ability to function.20 A comparison of eszopiclone and zolpidem on other adverse events in this trial is not available.

Zaleplon compared with zolpidem

Rates of overall adverse events and withdrawals due to adverse events were similar in short-term head-to-head trials of zaleplon compared with zolpidem and increased with longer duration of the trials (Table 9).

Table 9

Table 9

Adverse events in head-to-head studies of zaleplon compared with zolpidem

The most common treatment-emergent adverse events were headache and dizziness. In a 2-week trial in older adults,12 somnolence was significantly more common (P<0.05) with zolpidem 5 mg (10%) than with placebo (2%) or zaleplon 5 mg (4%). In one of two 4-week trials in adults,15 dizziness was significantly more frequent in 10 mg and 20 mg treatment groups than in the placebo group (P<0.001), occurring in 8% of patients in the placebo group, 3% in the zaleplon 5 mg group, 9% in the zaleplon 10 mg group, 14% in the zaleplon 20 mg group, and 14% in the zolpidem 10 mg group.

In a single-dose study conducted in 53 general practice patients,17 3 adverse events occurred in zolpidem 10 mg group (sluggish tongue, impaired concentration, leg complaints), and 4 in the zaleplon 10 mg group (2 headache, 1 abdominal fullness, 1 vertigo).

Zolpidem compared with zopiclone

Zolpidem was compared with zopiclone in a study designed to measure withdrawal effects.13 This was not a head-to-head trial, but 2 trials with the same design conducted simultaneously. The comparison in each trial was the effect of withdrawing treatment compared with continuing treatment. During the 2 weeks following withdrawal from treatment, the incidence of adverse events was higher in the withdrawal groups than the continued treatment groups, but was similar for zolpidem and zopiclone (38% and 41%, respectively). Most events were sleep-related.

In a two-week head-to-head study conducted in Japan, more patients in the zopiclone group than the zolpidem group had an adverse event “related,” “probably related,” or “possibly related” to treatment (31.3% compared with 45.3%; P=0.004).16 There were no significant differences in the proportion of patients who withdrew due to any adverse event (8.5% zolpidem compared with 10.2% zopiclone) or due to a drug-related adverse event (6.6% compared with 8.9%). The frequency of specific adverse events was similar between groups, with the exception of bitter taste, which occurred in 3% of the zolpidem group and 31% of the zopiclone group.

Indirect evidence

Figure 6 shows withdrawals due to adverse events reported in placebo-controlled trials. There was no difference between active drugs and placebo with the exception of zolpidem extended-release. Using a pooled analysis of 3 trials of zolpidem extended-release, we found that risk of withdrawal due to adverse events was higher with zolpidem than placebo (relative risk 1.93; 95% CI 1.17 to 3.21).89, 107, 121

Figure 6

Figure 6

Withdrawals due to adverse events reported in placebo-controlled trials of newer drugs for insomnia

We conducted an adjusted indirect analysis of withdrawals due to adverse events and found no differences between any of the drugs (Table 10).

Table 10

Table 10

Adjusted indirect analysis of placebo-controlled trials: Withdrawals due to adverse events

Newer insomnia drugs compared with benzodiazepines

No additional studies comparing newer insomnia drugs with benzodiazepines were identified for Update #2. Appendix D summarizes results of good- and fair-quality studies of newer drugs compared with benzodiazepines in the general population of adults and older adults with insomnia. We also included 6 active-control trials in subgroups of patients with comorbid conditions.

No trials compare eszopiclone, ramelteon, or zolpidem extended-release with benzodiazepines. Comparison of zaleplon with benzodiazepines is limited to 2 fair-quality trials comparing zaleplon with triazolam.29, 62

Zolpidem

Zolpidem was compared with flurazepam in 1 included study,32 with temazepam in 2,40, 60 and with triazolam in 4.40, 44, 50, 53

In the study comparing zolpidem 10 mg or 20 mg with flurazepam 30 mg, zolpidem was more effective for sleep outcomes.32 Adverse events were similar for zolpidem 10 mg and flurazepam, but zolpidem 20 mg was associated with more adverse events.

The 2 studies comparing zolpidem with temazepam40, 60 found the drugs similar in efficacy and rebound insomnia.

In 2 studies comparing zolpidem 10 mg with triazolam 0.25 mg,50, 53 sleep outcomes were similar for the two drugs, but zolpidem caused less rebound insomnia. There was also less rebound insomnia with zolpidem 5 mg than triazolam 0.25 mg50 and with zolpidem 10 mg than triazolam 0.5 mg.44

A review by the National Institute for Clinical Excellence124 presents an analysis of 2 studies comparing zolpidem with nitrazepam that were excluded from our review because they are not written in English.(Kazamatsuri, 1993 and Kudo, 1993) The 2 studies found no significant differences between drugs in sleep latency or duration. In one study more patients reported improved sleep quality with zolpidem than nitrazepam (66.7% compared with 37.5%, P=0.031) (Kudo, 1993). The other study found fewer awakenings with zolpidem (Kazamatsuri, 1993). Both studies found no differences in rates of adverse events (odds ratio 0.70; 95% CI 0.37 to 1.30) and no difference in daytime alertness or global impression of treatment.

Zaleplon

In 2 trials comparing zaleplon with triazolam, the drugs were similar in most sleep outcomes and short-term adverse events.29, 62 In 1 study triazolam 0.25 mg was associated with more nausea than zaleplon 5 mg, a low dose.62 The same study found no difference in nausea between triazolam 0.25 mg and zaleplon 10 mg, the manufacturer’s suggested initial dose.62

Zopiclone

Zopiclone has been compared with four benzodiazepines (flurazepam, nitrazepam, temazepam, and triazolam). In 5 studies comparing zopiclone with flurazepam,26, 31, 43, 45, 54 most comparisons found the two drugs to be similar in efficacy and adverse effects.

Zopiclone and triazolam were similar in efficacy and adverse events.28, 37, 38 For rebound insomnia, results were mixed in 2 studies, with 1 finding that triazolam causes more rebound33 and the other finding no difference.36

In studies comparing zopiclone with nitrazepam,22, 39 efficacy and safety were similar, but nitrazepam was associated with more rebound insomnia.

The National Institute for Clinical Excellence review124 presents an analysis of 4 studies comparing zopiclone with temazepam. No significant differences were found by the 2 studies that directly compared the drugs’ sleep outcomes (sleep latency, sleep duration, number of awakenings, and sleep quality). Adverse events were similar in the study that directly compared them.

Newer insomnia drugs compared with trazodone

We identified 1 short-term, fair-quality study comparing zolpidem 10 mg with trazodone 50 mg.61 Sleep latency was shorter with zolpidem after 1 week of treatment (48.2 compared with 57.7 minutes, P=0.037), but the difference was not significant at week 2 (48.1 compared with 54.5 minutes, P not reported). Sleep duration, number of awakenings, sleep quality, and patients’ global impression of treatment were similar for the drugs at weeks 1 and 2. The number of total adverse events and withdrawals due to adverse events were similar for the drugs. However, more patients reported somnolence with trazodone (16% compared with 23%).

A trial of comparing trazodone with zaleplon in psychiatric inpatients was rated poor quality and does not provide additional comparative information about newer insomnia drugs compared with trazodone.52

Long-term safety

There is limited evidence about the long-term safety of newer drugs for insomnia and no direct evidence about their comparative long-term safety.

Eszopiclone

In a 6-month placebo-controlled trial of eszopiclone 3 mg,90 the rate of serious adverse events was 2.9% for eszopiclone and 1.0% for placebo. The most common serious adverse events were gastrointestinal disorder (0.5% per group) and chest pain (0.5% per group). After discontinuation of the drug, similar overall rates of “new” events (defined as either not seen during the treatment period or worsening after the treatment period) were seen in placebo (10.7%) and eszopiclone (11.2%) groups. There were no reports of seizures, hallucinations, or perceptual-disturbance events. Anxiety was reported once in the eszopiclone group. Adverse events occurred in 81.1% of the eszopiclone group and 70.8% of the placebo group. The most common adverse event was unpleasant taste (26.1% eszopiclone compared with 5.6% placebo). Over 6 months, the rate of discontinuation due to adverse events was 12.8% in the eszopiclone group and 7.1% in the placebo group. The most common reasons for discontinuation were somnolence (2.2% eszopiclone compared with 1.5% placebo), depression (2.0% compared with 0.0%), unpleasant taste (1.7% compared with 0.5%), headache (0% compared with 2%), asthenia (1% compared with 1.5%), and insomnia (0.0% compared with 1.5%).

A 6-month, open-label extension study of this trial has also been conducted.108 All patients who completed the double-blind phase were eligible to participate in the open-label extension. Of the 788 patients enrolled in the 6-month double-blind phase, 471 patients continued into the 6-month open-label extension study (59.8%), and 382 completed a full 12 months of treatment (48.5%). Improvements in sleep outcomes were sustained; rebound insomnia and withdrawal effects were not reported. During the extension study108 3.8% of patients discontinued due to adverse events. The most common treatment-related adverse events were unpleasant taste (6.8%), headache (4.7%), somnolence (3.8%), abnormal dreams (3.0%), and dizziness (2.5%).

A more recently published 6-month study of nightly treatment with eszopiclone 3 mg was conducted in 828 patients with chronic insomnia.119 Rates of withdrawals due to adverse events were similar in the eszopiclone (9%) and placebo (8%) groups. By the end of the study, 37% of eszopiclone and 52% of placebo patients withdrew. There were more reports of somnolence, unpleasant taste, and myalgia in the eszopiclone group than the placebo group. Most events were mild or moderate. There was no evidence of withdrawal symptoms or rebound insomnia.

Zaleplon

A one-year open-label extension of a head-to-head trial12 was conducted to assess the longer-term safety of zaleplon 5 mg in older patients.128 In order to qualify for the extension phase, patients must have completed the trial and a 7-day placebo run-out without adverse effects. Thus this extension was limited to a sample of patients highly selected to be less likely to experience discontinuation effects. Sixty-four percent of patients who completed the 2-week trial enrolled in the extension study. Results of this open-label extension are reported in combination with those of an extension of a different, unpublished trial, also conducted in older people. The most frequent adverse events were headache (27%) and infection (13%). The most frequent adverse events resulting in discontinuation were pain (5%), somnolence or dizziness (4%), and gastrointestinal disturbance (2%). There was a significant increase in sleep latency, number of awakenings, and reduced total time slept on the first night after discontinuation, but these did not approach original baseline levels.

Zolpidem

Two open-label studies in general practice patients in France assessed the safety of 6 months of treatment with zolpidem.137, 142 One looked at zolpidem 10 mg or 20 mg137 in 96 patients over age 40. All 96 patients were followed for 6 months; 49 of these patients continued treatment for an additional 6 months. Patients were evaluated every 30 days. About 70% of patients used the 10 mg dose. In the first 6 months, 7.3% of patients withdrew due to adverse events considered related to the drug, including a feeling of strangeness (1 patient), a feeling of drunkenness (1 patient), anterograde amnesia (2 patients), nausea (1 patient), a confusional episode (1 patient), malaise (1 patient), vertigo (4 patients), daytime drowsiness (2 patients), unpleasant awakening (1 patient), and diplopia (1 patient). Four of the 49 patients who continued treatment after 6 months withdrew (8%); two experienced nightmares, but these were not considered to be related to the study drug. No withdrawal or rebound phenomena were reported.

In the second study, 107 patients were enrolled, and 20 patients withdrew before 6 months (18.7%): two because of inefficacy, seven because of adverse events, two because of hospital admission for other reasons, and two because of resolution of sleeping disturbances.142 Adverse events included malaise (5 events), vertigo (5 events), and anterograde amnesia (5 events). Patients experiencing vertigo and confusion were all over age 70. There was no evidence of tolerance over the 6-month course of the study, and no rebound insomnia.

Zolpidem extended-release

In a 6-month placebo-controlled trial of zolpidem extended-release 12.5 mg, administered for 3 to 7 nights per week, 8.5% of patients in the treatment group and 4.6% of patients in the placebo group withdrew due to adverse events.89 The most common adverse events associated with zolpidem extended-release were headache (10.5%), anxiety (6.3%), somnolence (5.7%), dizziness (4.8%), fatigue (4.5%), disturbance in attention (4.3%), irritability (3.7%), nausea (3.4%), and sinusitis (3.3). Most events were mild or moderate in severity. There was no evidence of tolerance to treatment over the 6-month study period and no rebound insomnia on the first 3 nights after discontinuation of medication.

Zopiclone

We identified no prospective studies that assessed the long-term safety of zopiclone.

Abuse and dependence

Abuse and dependence have been associated with zolpidem and zopiclone.149–166 A review of case reports and epidemiological data found most patients abusing or becoming dependent on zolpidem had a history of drug or alcohol abuse or other psychiatric conditions.167 A study of French data on zolpidem collected by the Centers for Evaluation and Information on Pharmacodependence found that from 1993 to 2002, the period of the study, health professionals spontaneously reported an increasingly higher number of cases of abuse or dependence associated with zolpidem.144 In 1993 <1% of abuse and dependence reports included zolpidem, and by 2002 almost 5.5% included zolpidem. An epidemiological survey of falsified or forged prescriptions shows that the popularity of zolpidem among forged prescriptions has increased: It was the 6th most common drug for which prescriptions were falsified in 1998 and had risen to #1 by 2004. The ratio of the number of forged zolpidem prescriptions to the number of legitimate zolpidem prescriptions indicates that zolpidem’s falsification ratio is moderate, although higher than that of the leading benzodiazepine in France (specific data not reported). Finally, annual surveys of drug abusers show that the number of patients using zolpidem increased from <1% in 1998 to 4% in 2001. Nearly all patients abusing zolpidem were abusing more than one drug, 1 of 2 also using a benzodiazepine and 4 out of 10 using cannabis. Until 1998, 100% of patients obtained zolpidem through medical prescriptions; since 2001 nearly 15%–20% of users bought it through street deals.

A 2003 survey of 297 patients admitted to addiction treatment sites in the United Kingdom136 found that while zopiclone was used by many more subjects than zolpidem (53.7% compared with 5.8%), the drugs were similar in their use to induce sleep (88% compared with 82%) or to get high (23% compared with 24%).

Eszopiclone, zaleplon, zolpidem extended-release, and ramelteon have been in use for a shorter period than zolpidem and zaleplon, so there is less information about their effects over the long term. The newer insomnia drugs, with the exception of ramelteon, are classified by the US Drug Enforcement Administration as controlled substances. Because of its different mechanism of action, ramelteon is not considered to have the potential for abuse and dependence of the other newer sedative hypnotics.

Case reports

We identified 64 case reports of adverse events: 1 with eszopiclone,168 3 with zaleplon,169–171 13 with zopiclone,150, 152, 157, 160, 162, 166, 172–178 and 46 with zolpidem.149, 151, 153–156, 158, 159, 161, 163–165, 179–213 Overall, the most commonly reported adverse event was hallucination, reported with eszopiclone (1 of 1 case), zaleplon (1 of 3 cases), and zolpidem (15 of 46 cases). The next most common adverse effect was dependence, with 7 cases reported for zolpidem and 6 for zopiclone. Somnambulism was also reported for zaleplon (1 of 3 cases) and zolpidem (6 of 46 cases). Finally, several cases of some form of amnesia were reported with zolpidem (4 of 46 cases).

Key Question 3. Are there subgroups of patients for which one newer drug for insomnia is more effective or associated with fewer adverse events?

Summary of the Evidence

  • Older adults (age ≥65 years)
    • - In a 2-week head-to-head trial comparing zolpidem with zaleplon in older adults, efficacy was similar to that in younger adults
    • - Somnolence was more common (P<0.05) with zolpidem 5 mg (10%) than with placebo (2%) or zaleplon 5 mg (4%), with no difference in overall adverse events or in withdrawals due to adverse effects
    • - A case-control study in 6110 older women found that use of zolpidem was associated with an increased risk of hip fracture (adjusted odds ratio compared with nonuse 1.95; 95% CI 1.09–3.51). The risk was higher than with benzodiazepines (adjusted odds ratio compared with nonuse 1.46; 95% CI 1.21–1.76)
  • Race and gender
    • - We found no evidence that one newer insomnia drug is safer or more effective in any subgroup based on gender or race
  • Pregnancy
    • In a prospective cohort study in 40 women with exposure to zopiclone in the first trimester of pregnancy, zopiclone use was associated with lower mean birth weight (3249 ± 676 grams compared with 3624 ± 536 grams; P=0.01) and gestational age (38.3 ± 2.7 weeks compared with 40.0 ± 1.6 weeks; P=0.002), but there were no differences in other pregnancy outcomes
    • - A prescription event monitoring study in the United Kingdom found no congenital anomalies among 18 births in women who had taken either zolpidem or zopiclone during the first trimester of pregnancy
    • - No evidence is available about use of other newer insomnia drugs during pregnancy
  • Comorbid conditions
    • - Active-control trials suggest that zopiclone is similar to benzodiazepines for sleep outcomes and adverse effects in patients withdrawing from alcohol, patients with generalized anxiety disorder, and in patients with stroke living in a residential care facility
    • - Zolpidem 5 mg, but not 10 mg, was more effective than triazolam 0.25 mg for some sleep outcomes in patients with chronic obstructive pulmonary disease
    • - Eszopiclone, ramelteon, and zolpidem, have been studied in short-term placebo-controlled trials of patients with obstructive sleep apnea or upper obstructive sleep apnea or upper airway resistance syndrome and symptoms of inadequate sleep. In mild to moderate sleep apnea, sleep laboratory outcomes were better with eszopiclone than placebo, but not with ramelteon compared with placebo. In severe sleep apnea, zolpidem was significantly better on 2 of 5 sleep laboratory outcomes.
    • - In overweight patients with upper airway resistance syndrome, zopiclone was superior to placebo on 2 of 5 sleep laboratory measures

Detailed Assessment

Older adults

One head-to-head trial compared zaleplon with zolpidem in adults at least 65 years old.12 In this 2-week trial,12 somnolence was significantly more common (P<0.05) with zolpidem 5 mg (10%) than with placebo (2%) or zaleplon 5 mg (4%). There was no difference in overall adverse events or in withdrawals due to adverse events. A one-year open-label extension of this trial was conducted to assess the longer-term safety of zaleplon in older patients.128 Adverse events were mild. (See long-term-safety section for more details of this extension study).

In a subgroup analysis of our adjusted indirect meta-analysis, there was no difference between any of the newer insomnia drugs in sleep latency in older patients. (See Table 7.) Eszopiclone significantly increased sleep duration compared with ramelteon (mean difference 29.5 minutes; 95% CI 0.8–58.2) and zolpidem (mean difference 32.8 minutes; 95% CI 1.2–64.4) in older patients.

In a subgroup analysis of a study of ramelteon in older adults with severe sleep-onset insomnia (>60 minutes), there were significant reductions in subjective sleep latency with ramelteon 8 mg (−23.2 minutes) compared with placebo (−7.5 minutes; P=0.002) at week 1.94 Improvement over placebo was also evident at weeks 3 and 5.

A case-control study (N=6110) of the relationship between use of zolpidem or other medications and occurrence of hip fracture in older women found an increased risk of fracture in patients using zolpidem (adjusted odds ratio 1.95; 95% CI 1.09–3.51).145 This risk was higher than the risk with benzodiazepines (adjusted odds ratio 1.46; 95% CI 1.21–1.76). The study did not include other newer insomnia drugs, and so it provides no information for comparing the risk associated with zolpidem with the risk associated with other newer drugs for insomnia.

An observational study used data from a representative survey of Medicare beneficiaries to determine if the increased risk of hip fracture observed with sedative hypnotic use might be due to confounding factors that are not available from claims data.143 Potential confounders were body mass index, current smoking status, activities-of-daily-living score, cognitive impairment, and Rosow-Breslau physical impairment scale. The authors found that the activities-of-daily-living score was the strongest confounder, causing an overestimation of 10% in comparisons of zolpidem users with benzodiazepine users. They conclude, however, that the magnitude of the effect of unmeasured confounders is unlikely to explain completely the greater incidence of hip fracture observed in older users of sedative hypnotic.

A good-quality systematic review and meta-analysis compared the risks and benefits of a variety of pharmacological treatments for insomnia in people at least 60 years old.126 The review included studies of newer sedative hypnotics, benzodiazepines, and over-the-counter medications such as antihistamines. Only subjective sleep measures were included. Results were combined for all sleep agents for most outcomes, so this review cannot be used to make conclusions about the comparative efficacy and safety between newer sedative hypnotics or between newer sedative hypnotics and other sleep agents. Studies comparing zaleplon, zopiclone, and zolpidem (combined) with benzodiazepines found no significant difference in cognitive adverse events (odds ratio 1.12; 95% CI 0.16 to 7.76) or psychomotor-type adverse events (odds ratio 1.48; 95% CI 0.75 to 2.93).126 For all sedative hypnotics (newer and older) compared with placebo, the number needed to harm for all adverse events was 6 (95% CI 4.7 to 7.1), and the number needed to treat for improved sleep quality was 13 (95% CI 6.7 to 62.9). On the basis of these results, the authors concluded that in older people the benefit of sleep agents may not outweigh their risks.

Pregnancy

A prospective cohort study in Canada evaluated pregnancy outcomes after first-trimester exposure to zopiclone in 40 women.133 The sample consisted of women who had initiated contact with a program that provides counseling for pregnant women, thus it is not representative of the total population of women who were exposed to zopiclone during pregnancy.

Newborns in the zopiclone group had a significantly lower mean birth weight than newborns never exposed to the drug (3249 ± 676 grams compared with 3624 ± 536 grams; P=0.01). They also had a lower gestational age (38.3 ± 2.7 weeks compared with 40.0 ± 1.6 weeks; P=0.002). Once birth weight was adjusted for gestational age, the differences were no longer significant. There was no difference in outcome of pregnancy, delivery method, assisted deliveries, fetal distress, presence of meconium at birth, preterm deliveries, or neonatal intensive care admissions between zopiclone and control groups.

A 1998 report of prescription-event monitoring studies of newly marketed drugs, conducted in general practices in the UK, includes information on pregnancy outcome in 23 women exposed to zolpidem and 18 exposed to zopiclone during pregnancy.146 In women who had taken zolpidem, there were 2 spontaneous and 6 legal abortions. In women who had taken zopiclone, there were 3 spontaneous and 3 legal abortions, and in one the outcome is unknown. There were no congenital anomalies among the 18 live births in women exposed to either drug.

Comorbid conditions

Active-control trials show that zopiclone is similar to benzodiazepines for sleep outcomes and adverse effects in patients withdrawing from alcohol,23 patients with generalized anxiety disorder,34 and in patients with stroke living in a residential care facility.41

Zolpidem 5 mg, but not 10 mg, was more effective than triazolam 0.25 mg for some sleep outcomes in a trial in patients with chronic obstructive pulmonary disease.55

Zolpidem has been studied in placebo-controlled trials in patients with depression68 and other psychiatric conditions,113 in peri- and postmenopausal women,74 and in patients with fibromyalgia.95 Zaleplon has been studied in placebo-controlled trials in patients undergoing kidney dialysis.109 Zopiclone has been compared with placebo in trials of patients with rheumatoid arthritis or76 fibromyalgia75, 82 and in patients who are shiftworkers.97 Eszopiclone was more effective than placebo for insomnia in patients with rheumatoid arthritis,112 in patients with depression who were also taking fluoxetine,78 in patients with generalized anxiety disorder who were also taking escitalopram,214 and in peri- and postmenopausal women.114 In a single-dose study, ramelteon 16 mg improved polysomnographic sleep duration, total sleep time, and WASO in patients with mild to moderate chronic obstructive pulmonary disease; there was no difference between ramelteon and placebo on subjective sleep measures or on objective sleep latency.215 While these studies provide evidence that these drugs are effective for some sleep outcomes in patients with particular comorbid conditions, they do not provide evidence about the comparative efficacy of newer insomnia drugs in these subgroups.

Three studies evaluated newer insomnia drugs in patients with obstructive sleep apnea and continuing symptoms of inadequate sleep: Eszopiclone104 and ramelteon86 were studied in patients with mild to moderate sleep apnea; zolpidem was studied in patients with severe sleep apnea;69 and zopiclone was studied in patients with upper airway resistance syndrome.92 These were all small (N = 8 to 26), short-term crossover studies conducted in sleep laboratories. We rated all of them fair quality. Patients enrolled were predominantly male and 45 to 55 years old. Three studies enrolled patients whose body mass index was in the obese or severely obese range (mean body mass indexes 30, 32, and 36),69, 86, 104 and one study’s patient population had a mean body mass index in the “overweight” category (26.3).92 These studies were conducted with 1 to 7 nights of treatment.

In the studies of mild to moderate sleep apnea, sleep lab outcomes were better with eszopiclone than placebo, but ramelteon was no better than placebo. Latency to persistent sleep time and number of awakenings were similar for eszopiclone and placebo nights, but WASO, sleep efficiency, total sleep time, and wake time during sleep were statistically significantly better during the eszopiclone nights. Total sleep time was 15 minutes longer with eszopiclone.104 No other measures were used. With ramelteon, sleep lab measures of latency to persistent sleep, number of awakenings, total sleep time, and WASO were similar during drug and placebo nights.86 In addition, patient assessment of sleep (number of awakenings, total sleep time, sleep quality, sleep latency, level of alertness, awake time, and ability to concentrate) were also not different between the 2 sessions. However, the very small size of this study could have led to a type II error. For example, differences seen in total sleep time (14.6 minutes) and sleep latency (10.1 minutes) were similar to those seen in the eszopiclone study where statistical significance was found.

In 16 patients with severe obstructive sleep apnea, zolpidem was found to be significantly better on 2 of 5 sleep lab outcomes.69 Sleep latency was 10.4 minutes shorter and sleep arousal was lower by 2.5 arousals per hour. Total sleep time was 17.2 minutes longer with zolpidem, but this difference did not reach statistical significance. Again, the small size of this study may have lead to type II error.

In upper airway resistance syndrome, a condition related to sleep apnea, patients who are habitual snorers but do not experience apnea have sleep disturbances as a response to airway obstruction.92 Zopiclone was compared with placebo in 26 overweight patients in a 7-day crossover study. Zopiclone was superior to placebo in 2 of 5 measures taken in a sleep lab. In sleep efficiency and measures of daytime sleepiness, zopiclone was significantly better than placebo. Measures for which differences did not reach statistical significance were total sleep time (22 minutes longer with zopiclone) and sleep latency (23 minutes shorter with zopiclone).