The following is a guest post by Harold DeMonaco, MS, a member of our editorial team, and Director of the Innovation Support Center at the Massachusetts General Hospital. A graduate of the Massachusetts College of Pharmacy and Allied Health Sciences he holds a bachelors degree in pharmacy and a masters degree in therapeutics. He has a keen interest in the innovation process in medicine and organizational behavior related to change.
An epidemiologic study published in BMJ Open this week created quite a buzz in the media. The study looked at the administrative records of over 10,000 enrollees in an integrated health system who received a prescription for hypnotic drugs (AKA sleeping pills) over a 2.5 year period. Nearly 24,000 matched control enrollees who did not receive a prescription were compared. The conclusion was as follows:
Receiving hypnotic prescriptions was associated with a greater than threefold increased hazard of death even when prescribed <18 pills/year. This association held in separate analyses for several commonly used hypnotics and for newer shorter acting drugs. Control of selective prescription for patients in poor health did not explain the observed excess mortality.
Given the number of Americans who take sleeping pills, the amount of enthusiasm in the press is not very surprising. We counted more than 100 stories using a Google search. What is surprising is the wide variability in the headlines and in the reporting. Here are several examples of the headlines:
New study finds big risks for sleeping pills
Sleeping pills a killer habit, shock study claims
Sleeping pill users at higher risk of death, study finds
Sleeping pills called as risky as cigarettes
Sleeping pill death toll may top 500,000
Study: sleeping pills linked to early death
Study finds statistical link between sleeping tablets and death
And my personal favorite: Could sleeping pills kill you? Theres a high chance according to new research.”
I can only imagine how many telephone calls my primary care physician friends are receiving as a result of the media coverage.
As I read the article, I was struck by the statement that a prescription for as few as 18 doses of a hypnotic was associated with an increased risk of death over a 2.5 year period. Even more interesting is that it did not seem to matter which drug was prescribed. If we are to believe what many stories are saying about the study, there are lots of adults in the US and elsewhere at significant risk if they are prescribed a hypnotic drug.
To be right up front, I have no enthusiasm for the way in which sleeping pills are handed out in the US. Most of the drugs have only been shown to be effective for relatively short periods of time. Despite their limited usefulness they represent cash cows for the pharmaceutical industry and are expected to top $9 Billion in worldwide sales in 2015. Ambien CR and Lunesta racked up almost $2 Billion in sales in the US in 2010. A months supply of Lunesta 3mg will set you back about $230. A similar supply of Ambien CR will cost about $190.
While the headlines are likely to grab the readers attention, they clearly demonstrate that most reporters dont really understand the difference between association and causation.
The study used administrative data from a large integrated health system. There are limitations to this approach. The administrative data sets are notoriously incomplete and people go in and out of coverage. They are also not very granular. Important confounding variables and severity of illness are usually not explicitly included.
Association and causality are not the same. Just because there is an association dose not imply causality. Just because there appears to be a link between death and the use of sleeping pills does not necessarily mean that the sleeping pills are the cause. Other factors not taken into consideration may be playing a significant role, rather than the single factor chosen by the researchers. The authors of the study did look for confounders that would increase the risk of death. Table 2 of the article provides a listing that includes asthma, cerebrovascular disease, coronary heart disease, chronic kidney disease, COPD to name a few. Interestingly, the people who had a prescription for sleeping pills also had a higher prevalence of a significant chronic medical condition as compared to those who were not prescribed.
Here are a couple of examples from that table all with a statistically significant difference (p,0.001) between the two groups:
The Chicken and the Egg question should pop into your mind. Are people with serious illness likely to be prescribed sleeping pills? Is the increased death rate merely a reflection of the underlying illness? Or, do the sleeping pills confer some increased risk?
Given the pharmacologic differences between the available prescription drugs, is it likely that they all are associated with the same risk? As Michael Yurcheshen, MD, an expert quoted by WebMD noted, “It is implausible to think that so many of these medications, spread across several different drug classes, could have the same biological effects. Yet the media rarely noted this important consideration. If in fact the sleeping pills were causative, it is also interesting that the association did not vary with the number of doses prescribed. Or, put another way, there does not appear to be a dose related association.
As we occasionally note, conflicts of interest are important in reporting of published medical studies. Although I did not read through all 100+ stories, the small sample I looked into did not note that the senior author of the study is a longstanding critic of hypnotic drugs and has a website devoted to the topic.
Most of the stories dont give readers any information about alternative approaches. So, the uninformed reader presumably then has two choices, take the sleeping pills and risk a fourfold increase in death or dont sleep. Clearly those are not the correct options.
The study by Kripke and colleagues is provocative to say the least. While there are limitations to the study and perhaps the enthusiasm of the authors, the conclusions are both interesting and deserving of further discussion.