"Sleep Drugs Found Only Mildly Effective, but Wildly Popular” proclaimed the headline of a recent New York Times article (S. Saul, October 23, 2007). The Times analyzed the results of a newly published meta-analysis of chronic insomnia treatments and concluded that, on average, sleeping pills reduce the time to go to sleep (sleep latency) by only 12.8 minutes beyond placebo, and increase total sleep time by only 11.4 minutes.

These apparently unimpressive numbers were derived from an NIH-funded study available for free on the web (Buscemi N et al., AHRQ Report, accessed at http://www.ahrq.gov/downloads/pub/evidence/pdf/insomnia/insomnia.pdf), and it’s well worth scanning, although at 135 pages (not including the 154-page appendix), it’s not suitable for between-appointment reading. The bottom line of the report is that when you hook patients up to dozens of wires and put them in a sleep lab bed, you can still demonstrate that sleeping pills beat placebo, but the numbers won’t blow your socks off. In the real world, not only do patients snooze wirelessly in their own Serta Perfect Sleeper, but they benefit from both the drug effect and the placebo effect. Combining placebo and drug effects leads to roughly a half hour of improved sleep latency over taking no pill any kind.

Clearly, sleeping pills work; the more relevant issue is whether there are any significant differences among them, and whether we should pony up 3 dollars per pill for the Maserati medications or just 3 cents per pill for perfectly adequate Civics.

Benadryl (Diphenhydramine)

To begin our review with the Civics, Benadryl is an antihistamine with anticholinergic properties, and is probably being taken by more patients than you realize. Why? Because it’s a popular ingredient in over-the-counter cold and pain remedies, including Tylenol PM (500 mg acetominophen and 25 mg diphenhydramine) and the newer Advil PM (200 mg ibuprofen and 38 mg diphenhydramine). Benadryl definitely makes people drowsy and is an effective sleep aid, but anecdotally, patients are said to quickly develop tolerance and to have nasty cognitive side effects. Is there any actual data to guide us? The only large study to look at side effects was of 426 hospitalized elderly patients (mean age, 80), 114 of whom had been given Benadryl vs. 312 who had not. Patients given Benadryl had about double the rate of subtle symptoms of delirium, including inattention, disorganized speech, and altered consciousness (Agostini JV et al., Arch Int Med 2001;161(17):2091-2097). But this study is unlikely to be generalizable to most physically healthy outpatients in psychiatric practices. Studies of young healthy patients given Benadryl 50 mg BID (a dosing regimen for allergic rhinitis) have shown that any daytime sedation or cognitive impairment wears off quickly, within a couple of days (Richardson GS et al., J Clin Psychopharm 2002;22:511-515). But there are no published studies supporting the anecdotal impression that patients quickly develop tolerance to Benadryl when it is used at bedtime as a sleeping pill.

The bottom line on Benadryl is that you should avoid it in the elderly, especially those who are medically ill, but that it is likely safe and effective for most other patients, at least for short term use.

Desyrel (Trazodone)

First, I’ll lay out my own bias: I think trazodone is a great sleeping pill, I’ve used it with hundreds of patients and have never seen any dangerous side effects, priapism included. Nonetheless, trazodone has been the target of numerous “critical reviews” over the last few years, all funded by the makers of branded hypnotics that lose market share to trazodone (see for example, Mendelson WB, J Clin Psychiatry 2005;66:469-476, and James SP et al.,  J Clin Psychiatry 2004;65:752-755). The gist of these reviews is that trazodone has not undergone the same kinds of rigorous clinical trials for non-depressed primary insomnia as the newer agents, so we simply don’t know how well it really works. This is true, because trazodone is generic and there are no big-money sponsors willing to pay for fancy studies to show it works. Nonetheless, several studies over the years have shown low dose trazodone (50-100 mg HS) to be effective for insomnia in depressed patients (e.g., Saletu-Zyhlarz GM et al., Prog Neuropsychopharmacol Biol Psychiatry 2002;26:249-260). One study, however, funded by Ambien’s original manufacturer, did compare Ambien and trazodone head-to-head, and found that both drugs were better than placebo for primary insomnia (Walsh JK et al., Human Psychopharm 1998;13:191-198). The sponsor has spun the results of the study to argue that Ambien was superior to trazodone, but a close read of the study does not support this conclusion.

Bottom line: For a very cheap, long-half-life sleeping pill, it’s hard to beat trazodone, but be sure to assess your patients for next-day sedation, orthostatic hypotension, and priapism, all of which are possible side effects.

Ambien (Zolpidem) and Ambien CR

We’ll start with the potential dark side of Ambien, simply because this has been on patients’ minds ever since an Ambien-addled Patrick Kennedy crashed his car enroute to a 2 AM “vote” on Capitol Hill (the last vote had occurred at 9 PM). Since then, the FDA has issued alerts regarding possible “sleep-driving” and “sleep-eating” on all sleeping pills.

Is Ambien (and by extension, the other nonbenzodiazepines) “abusable?” Undoubtedly. But is it less abusable then benzodiazepines? Almost certainly. There have been numerous case reports of zolpidem abuse (for example, see Cubala WJ et al., Br J Clin Pharm 2007;online early article). Typically, Ambien abusers take an awful lot of the stuff to get high (these authors reported a range from 160 mg/day to 2000 mg/day – yes, 2000 mg), and in rare cases zolpidem withdrawal seizures have occurred. But one study provided strong evidence that Ambien abuse is much rarer than benzodiazepine abuse. Researchers in Germany looked at the reported rates of ambien and zopiclone (mother of Lunesta) abuse from 1992 to 1997. They found an average of 4.5 cases of nonbenzo abuse per 10,000 doses, vs. a rate of benzo abuse of 106.7 per 10,000 doses (Hajak G et al., Addiction 2003;98:1371-1378). This implies that benzos are about 20 times more likely to be abused than the non-benzos.

Sleep-driving, -eating, and -walking are not really abuse issues, but possible side effects. While not common, they occur frequently enough that most psychiatrists have seen cases in their practices. They can usually (but not always) be prevented by reminding patients to get into bed right after taking their sleeping pill.

On the positive side, Ambien is an effective, FDA-approved sleeping pill. Now that it is available generically, its manufacturer, Sanofi, has been pushing us to switch patients from Ambien to Ambien CR. Well, should we? Theoretically, Ambien CR has the ideal pharmacokinetic profile. After ingestion, 60% of the dose is released immediately, and the remainder is released gradually throughout the night. This means that it should provide enough GABA agonism to get patients to sleep and keep them asleep. Indeed, while Ambien is approved for sleep initiation only, Ambien CR provided sufficient evidence to win an additional FDA approval for “sleep maintenance.”

Nonetheless, it is troubling to many that Sanofi has never funded a head-to-head trial comparing Ambien with Ambien CR. Are they afraid that IR will end up outshining CR? Instead of an appropriate comparative trial, we see research designed to guarantee a favorable impression of Ambien CR, such as one recent study of elderly patients comparing Ambien CR with Dalmane 30 mg, which is double the recommended dose for older patients. Dalmane, with a half-life of between 50-100 hours, is not exactly a first-line sleeping pill for the elderly, and, not surprisingly, it caused more next-day cognitive impairment than Ambien CR.

Bottom line: In the absence of the relevant comparative data, we’re left to our own anecdotal sense of whether Ambien CR actually poses clinical advantages.

Sonata (Zaleplon)

Sonata seems to have completely dropped off everybody’s radar screen, which is too bad, because it helps patients get to sleep and it causes no next day grogginess, even when taken in the middle of the night (Hindmarch I, et al., Hum Psychopharmacol. 2001;16(2): 159-167). It is not approved for sleep maintenance.

Lunesta (Eszopiclone)

Let’s face it: aside from a memorable commercial, Lunesta has no advantages over its competitors and has marked disadvantages. Its half-life is long, at 6 hours, meaning that it likely causes more next-day impairment than its competitors. In its promotional material, Sepracor claims no next-day residual effects in most patients. But if you dig into the actual data behind this claim, you’ll find that they chose a very convenient time point for their assessment-9.5 hours and 12 hours (Lunesta package insert) – whereas the other nonbenzos assessed for cognitive effects at 8 hours (see, for example, Ambien CR package insert). Like the foolish man looking for his car keys far from where he dropped them “because the light’s better here,” Sepracor opted to shed their research light far enough away from the time of ingestion to guarantee a good marketing line.

In fact, the precursor of Lunesta, zopiclone, does cause next-day impairment and is associated with a risk of car accidents (Staner L et al., Psychopharm (Berlin) 2005;181(4):790-798). The nail in Lunesta’s coffin is that it causes an unpleasant taste in 20-40% of people who take it (Lunesta package insert), prompting some Sepracor reps to recommend that patients bite into a lemon when they wake up in the morning. Thanks, but I’ll order the waffles instead.

If you don’t have anything compelling to say about your product, just keep funding opinion leaders to write something about it, no matter how irrelevant – eventually, you’ll build up market share, if only through literature saturation alone. Thus, we have company-funded studies such as a recent one in which 410 menopausal women with insomnia were randomized to receive either Lunesta or placebo. And indeed, Lunesta, a sleeping pill, helped these women sleep better (Soares CN et al., Obset Gynecol 2006;108:1402-1410). This is a contribution to the medical literature…how?

Bottom line: Lunesta has the same half life as Restoril (temazepam), is 100 times more expensive, and has resulted in a scourge of Luna moths fluttering around people’s ears while they are trying to sleep.

Rozerem (Ramelteon)

There are two good things to say about Rozerem: it has no abuse potential (really, none, nada) and it has a different mechanism of action from its competitors. Rather than acting by revving up GABA, it stimulates two subtypes of the melatonin receptors in the brain’s suprachiasmatic nucleus (SCN): MT1 and MT2. Of the two, MT1 is most specific for sedation, and Rozerem has a 15-fold greater affinity for MT1 than straight melatonin, a fact that probably accounts for why melatonin is only mildly effective as a sleep aid (see this meta-analysis of melatonin studies: Brzezinski A et al., Sleep Med Rev 2005;9:41-50).

For some reason, Rozerem has the anecdotal reputation of having a one to two week “lag time” before working, but this is not true. In fact, a recent review of the Rozerem studies, written by two pharmacists who received no payments from Takeda, concluded that it improves sleep latency on nights one and two as well as any of the competing agents (Borja NL et al., Clin Ther 2006;28:1540-1555). It has not been shown to decrease a standard measure of sleep, WASO (wake time after sleep onset), so it did not win the sleep maintenance indication.

According to experts I consulted, Rozerem doesn’t deliver the obvious “knock out punch” that the benzos and nonbenzos do, because it doesn’t act as a generalized CNS depressant at the GABA receptors. On the plus side, this means Rozerem is unlikely to cause the bizarre sleep-related behavior and the dizziness and falls implicated for the nonbenzos.

Bottom line: Rozerem’s worth a try, particularly for substance abusers and for the elderly. But, like all of the newer sleeping pills, it’s expensive, around $3/dose.