“Complex sleep behavior” is a euphemism for various problems that can happen after ingesting a sleeping pill. They range from cooking and emailing to driving a car or even sexual assault, all done in an amnestic state that is not recalled upon awakening. In 2007 the FDA placed warnings about these behaviors on all medications approved for insomnia, and this year they moved that warning up to a black-box level for the z-hypnotics: zolpidem (Ambien), zaleplon (Sonata), and eszopiclone (Lunesta). The new warning also applies to zolpidem’s various forms: CR, sublingual (Intermezzo, Edluar), and oral spray (Zolpimist).

The agency took this step to recognize the gravity of these parasomnias, which are too often the butt of jokes. They reviewed 26 years of adverse event reports and found 66 cases of complex sleep behaviors that resulted in either serious injuries (46 cases) or death (20 cases). Some of the cases involved accidental overdoses, falls, burns, near drowning, exposure to extreme cold temperatures leading to loss of limb, carbon monoxide poisoning, drowning, hypothermia, motor vehicle accidents with the patient driving, and self-injuries such as gunshot wounds and apparent suicide attempts.

Although severe events are very rare, complex sleep behaviors are fairly common and occur in 3%-15% of people on z-hypnotics. Zolpidem is the most notorious only because it is the most frequently prescribed—its risk is actually no different than the others in its class. With all z-hypnotics, the risk goes up as the dose goes higher (Chen LF et al, Neuropsychiatr Dis Treat 2013;9:1159-1162).

The most important part of this warning is the word “contraindication.” The FDA now recommends discontinuing z-hypnotics in anyone who has had a complex sleep behavior after taking them, however mild. Patients may protest that there’s nothing dangerous about microwaving popcorn at 2 a.m., but there’s always a risk that the problem might lead to a fractured skull or burnt arm.

Part of the reason for the absolute tone in this contraindication is that the benefits of z-hypnotics are so meager that it’s hard to justify their use in the face of these risks. On average, they cause people to fall asleep 22 minutes faster than a placebo by polysomnographic measures, and only 7 minutes faster by subjective report, according to a meta-analysis of the FDA-registration trials (Huedo-Medina TB et al, BMJ 2012;17;345:e8343). They do nothing to improve quality of sleep or long-term health outcomes. Patients seem to find these drugs more helpful than the research suggests, in part because there is a large placebo effect. People fall asleep 20-30 minutes faster with a placebo. The amnestic qualities may also explain why these drugs are so well liked: patients forget how poorly they slept.