4 Evidence and interpretation

The Appraisal Committee (Appendix A) considered evidence from a number of sources (see Appendix B). The remit given to NICE by the Department of Health/Welsh Assembly Government was to advise on the clinical and cost effectiveness of zaleplon, zolpidem and zopiclone relative to benzodiazepines. The appraised evidence was therefore restricted to that informing comparison of the Z-drugs with benzodiazepines that are approved for the treatment of insomnia and may be prescribed within the NHS (diazepam, nitrazepam, loprazolam, lorazepam, lormetazepam and temazepam).

4.1 Clinical effectiveness

4.1.1 The Assessment Report reviewed data from 24 randomised controlled trials (RCTs) that compared the Z-drugs with either a benzodiazepine or with another Z-drug in patients with insomnia. In total, 11 different comparisons were made between benzodiazepines (temazepam, lormetazepam or nitrazepam) and zolpidem (4 RCTs) or zopiclone (13 RCTs). No RCTs were identified that compared zaleplon with a benzodiazepine. Six RCTs were reviewed that compared zaleplon with zolpidem and one that compared zolpidem with zopiclone.

4.1.2 The duration of the studies ranged from 1 night to 6 weeks. Ten studies included a follow-up period, which ranged from 3 to 11 days. The number of patients included in the trials ranged from 10 to 615. The most common comparator used in the RCTs was nitrazepam, which has a prolonged duration of action and may give rise to residual effects on the following day. None of the trials compared the Z-drugs against 10 mg temazepam or 1 mg loprazolam. One of the ten studies used objective PSG recordings; the remaining nine collected data from post-sleep questionnaires and sleep diaries.

4.1.3 Five RCTs restricted their inclusion criteria to people of 60 years of age or older. Although it is recommended that the doses of both the Z-drugs and the benzodiazepine hypnotics should be reduced in older people, only three of the five RCTs used recommended doses for this age group. People over the age of 60 were amongst the population enrolled in a further 12 of the included RCTs. In these studies, standard dose hypnotics (benzodiazepines and Z-drugs) were used with no reported dose reductions for the people over the age of 60.

4.1.4 The Assessment Group reported that it was difficult to compare the results of individual studies because of methodological problems and variations in the outcome measures used. In addition, direct statistical comparisons between the hypnotics included in individual RCTs were not always made, and often insufficient data were reported to permit further analysis. There was also evidence of multiple testing of outcomes, with selective reporting of significant findings.

4.1.5 Although in the individual RCTs there were some statistically significant differences between the Z-drugs and the benzodiazepines for some of the efficacy outcome measures, the differences were not consistent across the trials. In addition, in most cases the absolute difference was small and the clinical significance of the differences was difficult to ascertain. The Assessment Group concluded that there was no consistent pattern of superiority of one drug over another.

4.1.6 Six RCTs compared zaleplon with zolpidem. One RCT found that 10 mg zaleplon per night resulted in statistically significant shorter sleep onset latency than 5 mg zolpidem (median time 31 minutes versus 42 minutes). Pooled data from three RCTs indicated the sleep was of less quality (odds ratio [OR] 0.66; 95% confidence interval [CI]: 0.51 to 0.87) and the median sleep time was statistically significantly less with 5 mg zaleplon per night compared with 5 mg zopiclone (291 minutes versus 309 minutes). Compared with 7.5 mg zopiclone, 10 mg zolpidem per night was associated with shorter sleep onset latency (OR 1.72; 95% CI: 1.04 to 2.84) in the 2-week trial identified. In the cross-over study, there were no statistically significant differences between 10 mg zaleplon and 10 mg zolpidem in the patients' global impression of treatment (38% versus 62%).

4.1.7 There was little consistency in the reporting of adverse events, which prevented comparison of individual event rates or meta-analysis. There were no statistically significant differences in the rates of treatment-emergent adverse events associated with any of the comparisons of Z-drugs versus benzodiazepines. There were no consistent differences between the Z-drugs and the benzodiazepines in the incidence of next-day residual effects.

4.1.8 In the RCT comparisons between the Z-drugs and benzodiazepines in people with insomnia, no data were identified on the frequencies of symptoms associated with withdrawal or dependency. In their submissions, the manufacturers also referenced a number of other studies that examined the rates of tolerance and dependency associated with the Z-drugs. The studies were not considered to be methodologically robust and there were no direct comparisons between the Z-drugs and the benzodiazepines used in the NHS.

4.1.9 The Assessment Group also searched for studies of other designs that specifically evaluated rates of dependence and withdrawal symptoms following treatment with the Z-drugs. Six studies were identified, four of which evaluated patients after extended treatment periods. Two placebo-controlled studies examined relative rates of withdrawal symptoms after patients receiving zolpidem were switched to placebo after 3 or 4 weeks of treatment. In one study, no patient in either group reported more than one symptom after 4 weeks treatment with 10 mg zolpidem or placebo and in the second study, three older patients who had received zolpidem at doses of 10–20 mg experienced adverse events. In addition, information on cases of dependence reported to the Committee on Safety of Medicines was sought and 16 case reports were identified in the literature relating to zolpidem (11) and zopiclone (5). There are problems with the interpretation of such reports, as rates of reporting are dependent on the publicity and awareness of certain adverse reactions and the pattern of use of the drugs.

4.1.10 In addition to the RCTs conducted in people with insomnia, a further 9 RCTs, which were conducted in healthy volunteers in whom insomnia had been induced, were submitted by the manufacturers. Most of these 9 studies had very small sample sizes (less than 30 people) or were of very short duration (for example 1–3 nights). The largest study was conducted in 630 people and compared 10 mg zolpidem with 15 mg temazepam, 0.25 mg triazolam or placebo. Data were collected from multiple outcome measures. There were no statistically significant differences between zolpidem and temazepam in objective measures of sleep latency and sleep efficiency. The zolpidem group had statistically significantly fewer awakenings than the temazepam group. For the subjective measures, the group receiving zolpidem reported greater ease in falling asleep, more sleep time and less wake time than those receiving temazepam. There were no statistically significant differences in subjects' ratings of their ability to concentrate or morning sleepiness.

4.1.11 A manufacturer submitted evidence from two RCTs that compared continuous zolpidem use (10 mg per night) with intermittent use. Data for each night were not reported but the submission reported that there were no statistically significant differences in the sleep efficacy outcomes, other than the investigators' assessment of sleep onset latency, which was statistically significantly greater in the continuous zolpidem group. No similar studies were located for zaleplon, zopiclone or any benzodiazepine.

4.1.12 In summary, the Assessment Group did not find any RCTs that appropriately compared the Z-drugs with shorter-acting benzodiazepines, used at appropriate doses. In the RCTs that were reviewed by the Committee, which had been conducted in both healthy volunteers and people with insomnia, there were no consistent differences between the drugs. However, this lack of consistency was attributed in part to the poor quality of reporting.

4.2 Cost effectiveness

4.2.1 None of the submissions contained an economic evaluation that compared the costs and effects of the short-term use of Z-drugs with benzodiazepines. In addition, the Assessment Group was unable to identify any evaluations in the health economics literature. No comparative data on the health-related quality of life associated with Z-drugs and benzodiazepines using generic health status measures were identified, and there was no evidence to link the clinical endpoints from the trials with quality of life.

4.2.2 The manufacturer of zaleplon submitted two models based upon the key assumption that zaleplon does not cause 'mental impairment' the day after administration.

4.2.2.1 The first model consisted of a cost–consequence algorithm comparing the costs and additional road traffic accidents (RTAs) associated with the residual effects of zopiclone and zaleplon. This model was based on a study mapping residual effects of zopiclone and zaleplon to RTAs using data from three other studies. The first stage of the mapping procedure was to estimate the impact of the residual effects of the Z-drugs on driving performance. This was taken from a doubleblind study of 28 healthy volunteers given zaleplon, zopiclone or placebo in the evening. The volunteers were woken in the middle of the night and given either placebo (those who had earlier taken active medication) or zaleplon, zopiclone or placebo (those who had earlier taken placebo). Participants undertook a series of tests, including a driving performance test, the following morning. The results found that driving performance was statistically significantly worse in the zopiclone group, and performance after zaleplon was similar to that for placebo. These results were linked to data from a similar study designed to measure the driving performance associated with differing levels of blood alcohol content. The relationship between blood alcohol content and the residual effects of the Z-drugs was then estimated. The relationship between relative risk of RTAs and blood alcohol content was estimated using data from a case–control study. The results of the model suggest that the expected excess accident costs, combined with drug purchase costs, over a 2-week period were US$71 per person for 10 mg zaleplon and US$92 per person for 10 mg zopiclone.

4.2.2.2 In support of the economic model, the manufacturer of zaleplon also cited a UK-based study which examined a sample of drivers involved in RTAs and compared the odds of having an accident whilst exposed or not exposed to specified drugs. The study found that zopiclone and anxiolytic benzodiazepines, but not hypnotic benzodiazepines, were associated with increased risk of RTAs. The manufacturer also cited a Canadian study estimating the relationship between hypnotic drugs and RTAs in older people. The study found that benzodiazepines with a long half-life were associated with an increased risk of RTAs, but that those with a short-half life were not. Concern was expressed regarding this study at the time of publication, particularly with regard to the failure to adequately control for confounding effects and the lack of distinction between benzodiazepines used as hypnotics and those used as anxiolytics.

4.2.2.3 The manufacturer of zaleplon also submitted a model designed to estimate the costs associated with hip fractures caused by falls as a result of the residual effects of zolpidem, nitrazepam, temazepam and zaleplon. The model assumed treatment on a basis of 2 weeks on therapy followed by 2 weeks off therapy over a one-year period and did not take into account benefits relating to treatment. This model was based on a published retrospective study, which examined the use of sedative hypnotics using Medicare data in a sample of older patients. Each case, defined as a patient who underwent surgical care for hip fracture, was matched by four age–gender matched controls from the Medicare database. Confounding factors are likely to bias results in studies of this type, particularly as insomnia may be a result of co-morbid conditions, which in turn could increase the risk of falls. Attempts to reduce confounding were made by adjusting crude odds ratios by a number of factors, including a measure of co-morbid illness severity. The study found that zolpidem use and benzodiazepine use were associated with a statistically significant increase in the risk of hip fracture. The model used the adjusted odds ratios from this study to represent the increased risk of hip fracture. The manufacturer assumed no additional risk of hip fracture was associated with zaleplon (that is, adjusted OR 1.0), although this drug had not been evaluated in the published study. The results of the model suggest that per year zolpidem, nitrazepam and temazepam therapy were more expensive than zaleplon when treating a cohort of 10,000 patients by £821,000, £129,000 and £111,000, respectively.

4.2.3 The Assessment Group reanalysed the hip fractures model after correcting some errors and adding illustrative qualityadjusted life year (QALY) values, but retaining the key assumption that zaleplon was not associated with increased risk of hip fractures. The Assessment Group concluded from the results of the amended model that at current acquisition costs, after taking into account the uncertainty surrounding all of the model inputs, the drugs included in the analysis appeared similarly cost effective compared with no treatment.

4.2.4 A manufacturer of zolpidem submitted a discussion of a model to estimate the costs of long-term zolpidem treatment compared with temazepam treatment. The model used estimates of dependence based on national prescribing data and concluded that despite having higher costs per dependent and non-dependent patient, zolpidem was shown to cost less on average because of its lower likelihood of high-cost dependency. However, the manufacturer stated that, "in the absence of robust data on the incidence of dependence, the author had to rely on a weak surrogate indicator of dependence in the form of continuous use".

4.2.5 In summary, the manufacturers suggest that the additional acquisition costs of Z-drugs would be offset by reduced consumption of other healthcare resources or lead to an improvement in health outcomes as a result of decreased dependence or reduced residual effects.

4.3 Consideration of the evidence

4.3.1 The Committee reviewed the evidence available on the clinical and cost effectiveness of zaleplon, zolpidem and zopiclone, having considered evidence on the nature of the condition and the value placed by users on the benefits of zaleplon, zolpidem and zopiclone from people with insomnia, those who represent them, and clinical experts. It was also mindful of the need to take account of the effective use of NHS resources.

4.3.2 The Committee heard testimony that insomnia is generally not well managed and that there is a lack of availability of training for healthcare providers in this field. It was advised that, despite national recommendations and restrictions specified in the SPCs, hypnotic agents are commonly used for minor degrees of insomnia and also prescribed for long periods. Use of these agents for extended periods is associated with increased likelihood of dependence.

4.3.3 The Committee heard evidence that some non-pharmacological strategies, including simple techniques such as the provision of advice on appropriate routines to encourage good sleep, (for example, avoiding stimulants and maintaining regular sleeping hours with a suitable environment for sleep) had been shown to be effective in the management of insomnia. The Committee considered that it was likely that such strategies could replace some of the current prescribing of hypnotics. However, nonpharmacological therapies did not fall within the remit of this appraisal and therefore their clinical and cost effectiveness had not been determined.

4.3.4 The Committee appreciated that there were differences in the pharmacokinetics of the individual hypnotics (both Z-drugs and benzodiazepines) which may have some benefits in specific clinical situations. For example, a hypnotic that is rapidly absorbed and rapidly cleared will inevitably result in shorter sleep onset latency, but it may not extend the total sleep duration as its effects will rapidly wear off. The Committee was not however persuaded that these differences resulted in any overall benefit for the majority of patients in terms of perceived quality of sleep, daytime functioning or quality of life. The Committee was also aware that when comparing the Z-drugs with nitrazepam and diazepam, consideration needed to be given to the fact that it was inevitable that the longer half-life of these benzodiazepines would be associated with an increased likelihood of persistence of the sedative effects into the next day.

4.3.5 The Committee considered that the RCTs available, in both people with insomnia and healthy volunteers, did not reflect current NHS practice: none of the Z-drugs had been compared with appropriate hypnotic doses of temazepam and the most common comparator used in the RCTs was nitrazepam, which has a prolonged duration of action and may give rise to residual effects on the following day. The Committee also appreciated that the effects of both the Z-drugs and the benzodiazepines were dose-related and that inappropriate comparisons, particularly in older people, would confound the results of the RCTs.

4.3.6 The Committee was made aware by the patient organisation that warnings regarding potential dependence associated with extended use of hypnotics are often not passed to patients. The Committee was particularly concerned that patients may be preferentially prescribed Z-drugs or transferred from benzodiazepines to the Z-drugs because of a perception that they are less likely to induce dependency than the benzodiazepines. In addition, the Committee considered that the substitution of the Z-drugs for patients who were being withdrawn from benzodiazepines was inappropriate and not supported by available evidence of reduced potential for dependency.

4.3.7 The Committee recognised that the benzodiazepines are abused and was informed by both the experts and the patient representatives that, although there was limited epidemiological evidence, abuse of the Z-drugs was increasing.

4.3.8 Having considered the results of the RCTs and healthy volunteer studies, together with the testimony from the professional and patient experts, the Committee concluded that currently there was no compelling evidence of a clinically useful difference between the Z-drugs and shorter-acting benzodiazepine hypnotics from the point of view of their effectiveness, adverse effects, or potential for dependence or abuse. There was no evidence to suggest that if a patient did not respond to one of these hypnotic drugs, they were likely to respond to another and this was supported by testimony from the clinical and patient experts. The Committee therefore concluded that 'switching' between these hypnotics was not an appropriate management strategy.

4.3.9 The Committee considered the economic models submitted by one of the manufacturers. The Committee fully discussed the basis of the manufacturer's models and recognised that they were based on the premise that the use of individual Z-drugs in preference to other Z-drugs or benzodiazepines would prevent road traffic accidents or hip fractures caused by falls. The Committee did not accept these models as it considered that the evidence used in them was not robust and the additional assumptions underpinning the models were not appropriate. In addition, the Committee considered that there was no reliable evidence to support the claim that the higher acquisition costs of the Z-drugs would be offset by the reductions in the use of other health service resources.

4.3.10 In summary, given the lack of compelling evidence on any clinically useful differences between the Z-drugs and the shorter-acting benzodiazepine hypnotics, the Committee concluded that, unless a patient experiences adverse effects considered to be directly related to a specific hypnotic, the drug with the lowest purchase cost (taking into account daily required dose and product price per dose) should be used in preference to more expensive alternatives.