As most readers know, we’re paying more attention than ever before to conflicts of interest in medicine. If an individual physician, researcher, speaker, or author is known to have a financial relationship with a drug company, we publicize it. It’s actually federal law now. The idea is that doctors might be biased by drug companies who “pay” them (either directly—through gifts, meals, or cash—or indirectly, through research or educational grants) to say or write things that are favorable to their drug.
A recent article on the relationship between sedative/hypnotics and mortality, published this week in BMJ Open (the online version of the British Medical Journal) and widely publicized, raises additional questions about the conflicts and biases that individual researchers bring to their work.
Co-authors Daniel Kripke, of UC San Diego, and Robert Langer, of the Jackson Hole Center for Preventive Medicine, reviewed the electronic charts of over 30,000 patients in a rural Pennsylvania health plan. Approximately 30% of those patients received at least one prescription for a hypnotic (a benzodiazepine like Klonopin or Restoril, or a sleeping agent like Lunesta or Ambien) during the five-year study period, and there was a strong relationship between hypnotics and risk of death. The more prescriptions one received, the greater the likelihood that one would die during the study period. There was also a specifically increased risk of cancer in groups receiving the largest number of hypnotic prescriptions.
The results have received wide media attention. Mainstream media networks, major newspapers, popular websites, and other outlets have run with sensational headlines like “Higher Death Risk With Sleeping Pills” and “Sleeping Pills Can Bring On the Big Sleep.”
But the study has received widespread criticism, too. Many critics have pointed out that concurrent psychiatric diagnoses were not addressed, so mortality may have been related more to suicide or substance abuse. Others point out the likelihood of Berkson’s Bias—the fact that the cases (those who received hypnotic prescriptions) may have been far sicker than controls, despite attempts to match them. The study also failed to report other medications patients received (like opioids, which can be dangerous when given with sedative/hypnotics) or to control for socioeconomic status.
What has not received a lot of attention, however, is the philosophical (and financial) bias of the authors. Lead author Daniel Kripke has been, for many years, an outspoken critic of the sleeping pill industry. He has also widely criticized the conventional wisdom that people need 8 or more hours of sleep per night. He has written books about it, and was even featured on the popular Showtime TV show “Penn & Teller: Bullshit!” railing against drug companies (and doctors) who profit by prescribing sleep meds. Kripke is also one of the pioneers of “bright light therapy” (using high-intensity light to affect circadian rhythms)—first in the area of depression, and, most recently, to improve sleep. To the best of my knowledge, he has no financial ties to the makers of light boxes. Then again, light boxes are technically not medical devices and, therefore, are not regulated by the FDA, so he may not be required to report any affiliation. Nevertheless, he clearly has had a decades-long professional interest in promoting light therapy and demonizing sleeping pills.
Kripke’s co-author, Robert Langer, is an epidemiologist, a past site coordinator of the Women’s Health Initiative, and a staunch advocate of preventive medicine. More importantly, though (and advertised prominently on his website), he is an expert witness in litigation involving hormone replacement therapy (HRT), and also in cancer malpractice cases. Like Kripke, he has also found a place in the media spotlight; he will be featured in “Hot Flash Havoc,” a movie about HRT in menopausal women, to be released later this month.
[Interestingly, Kripke and Langer also collaborated on a 2011 study showing that sleep times >6.5 hrs or <5 hrs were associated with increased mortality. One figure looked virtually identical to figure 1 in their BMJ paper (see below). It would be interesting to know whether mortality in the current study is indeed due to sedative prescriptions or, if the results of their earlier paper are correct, simply due to the fact that the people requesting sedative prescriptions in the first place are the ones with compromised sleep and, therefore, increased mortality. In other words, maybe the sedative is simply a marker for something else causing mortality—the same argument raised above.]
Do the authors’ backgrounds bias their results? If Kripke and Langer were receiving grants and speakers’ fees from Forest Labs, and published an article extolling the benefits of Viibryd, Forest’s new antidepressant, how would we respond? Might we dig a little deeper? Approach the paper with more skepticism? Is the media publicizing this study (largely uncritically) because its conclusion resonates with the “politically correct” idea that psychotropic medications are bad? Michael Thase (a long-time pharma-sponsored researcher and U Penn professor) was put in the hot seat on “60 Minutes” a few weeks ago about whether antidepressants provide any benefit, but Kripke and Langer—two equally prominent researchers—seem to be getting a free ride, as far as the media are concerned.
I’m not trying to defend the drug industry, and I’m certainly not defending sedatives. My own bias is that I prefer to minimize my use of hypnotics in my patients—although my opposition is not so much because of their cancer or mortality risk, but rather the risk of abuse, dependence, and their effect on other psychiatric and medical symptoms. The bottom line is, I want to believe the BMJ study. But more importantly, I want the medical literature to be objective, fair, and unbiased.
Unfortunately, it’s hard—if not impossible—to avoid bias, particularly when you’ve worked in a field for many years (like Kripke and Langer) and have a strong belief about why things are the way they are. In such a case, it seems almost natural that you’d want to publish research providing evidence in support of your belief. But when does a strongly held belief become a conflict of interest? Does it contribute to a bias in the same way that a psychopharmacologist’s financial affiliation with a drug company might?
These are just a few questions that we’ll need to pay closer attention to, as we continue to disclose conflicts of interest among medical professionals. Sometimes bias is obvious and driven by one’s pocketbook, other times it is more subtle and rooted in one’s beliefs and experience. But we should always be wary of the ways in which it can compromise scientific objectivity or lead us to question what’s really true.
Thought Broadcast 
Sleeping pills are “deadly?” you ask?
I gotta go with Dr. Breggin again.
In this article on the Huffington Post, he talks about “Xanax facts”…
Just the “facts” (no matter how unpopular to his own profession, psychiatry) –
http://www.huffingtonpost.com/dr-peter-breggin/xanax-whitney-houston_b_1288122.html
Duane
Perhaps this study is biased, but we already know that alcohol plays a role in 50% of suicides, and these drugs as depressants and potentially disinhibiting agents operate similarly in the brain to alcohol. Yes, people with substance use and psychiatric disorders may be more at risk, but those are the very patients we treat, so psychiatrists should be particularly conservative, given our target population. Confucius say, “The cautious seldom err.”
The study in the text is unthinkably flawed.
It is more useful to blame incautious prescribing habits than drugs.
Isn’t it clear that somtimes you simply have to create a benzo adict to avoid a suicide or other forms of total destruction?
I propose: if you know what you are doing and why and are properly afraid, then do what you must.
True that the Kripke studies seem questionable (cancer??) but more and more emerges about the dangers of polypharmacy, which so often includes benzos and sleeping medications — see the more credible information in this NY Times article Cocktail of Popular Drugs May Cloud Brain http://well.blogs.nytimes.com/2012/02/27/cocktail-of-popular-drugs-may-cloud-brain/
I don’t know what to think any more.
I’ m a male 50 tears old and take lorazepam for four years now. 1 mg every evening. I feel great.
Now I‘ m going to stop because I’ m afraid of the study results. A depressive thought.
So many people take lorazepam 1 mg every evening. Is it so difficult to case controll lets say a thousand of them and see if the results from mr Kripkes study are correct?
If not, I suggest to put mr Kripke in jail.
Bart,
1 mg of lorazepam at night, if you are not mixing it with high doses of alcohol and/or other central nervous system depressant drugs, will never kill you.
Back in the sixties, benzodiazepines like lorazepam (e.g. valium) quickly became the most prescribed drugs in the country by far. There was no spike in the death rate, and average life expectancy continued to climb.
Some people are just addicted to getting to sleep, but get labeled as drug abusers.
However, Bart may have a great deal of difficulty reducing that chronic dosage of lorazepam, and suffer rebound insomnia and anxiety like he’s never seen as well as other debilitating symptoms.
Technically, it’s not addiction, it’s physical dependency.
With all due respect, Dr. Allen, saying people are addicted to getting to sleep is black humor of the highest caliber! If you’ve ever experienced that rebound insomnia — well, it transcends what’s usually called insomnia. There should be another name for it. And the rebound anxiety can make your teeth themselves shake in their sockets….
Altostrata,
All possible, but there are several ways to prevent rebound symptoms. Also, sometimes it is the original symptoms coming back. One mg of lorazepam is a very low dose to begin with. Most people would not have trouble tapering it off (except perhaps their anxiety about being off the medication – psychological dependence can indeed be a big problem – usually because the medication worked so well to begin with).
Other than gradual tapering, are there any other strategies (in a nutshell) to avoid the risk of rebound insomnia? This is a significant problem in the use of “Z” drugs, too.
(As nervous systems have different sensitivities, 1mg lorazepam may be a low dose to one person but a substantial dose to another.)
Altostrata – you’ve probably already heard them. Gradual tapering is the primary way. Or you can temporarily use alternative sedatives like trazodone. For people on high doses, very short term use of tregretol is often useful. Some folks just tough it out for a few days, and then symptoms go away (a seizure can be a risk of that, however, so I don’t recommend that).
Of course some people can be extremely sensitive to any particular drug; I was saying that 1 mg of lorazepam is a very low dose for the vast majority of patients.
You should see how often some patients have, on their own, stopped benzo’s cold turkey when they find themselves pregnant. Many times, surprisingly, they have no after-affects at all.
And there are often placebo-like “negative” effects from discontinuing a drug with some patient types.
Dr. Allen,
Patients who are “addicted to going to sleep” is not the problem.
Doctors who are addicted to prescribing drugs is the problem –
http://www.madinamerica.com/2011/11/anxiety/
http://breggin.com/index.php?option=com_docman&task=cat_view&gid=29&Itemid=37
I suppose the “benzodiazepines save lives” argument is next, huh?
We’ve heard that one before.
It doesn’t hold water.
Duane
Rather than questioning their motives I think you could help by dissecting the facts in this study and/or examining the set up. What can we conclude from this study, and what not? If we really want to know if sleeping pills kill – and yes, we certainly do – what would be the best way to get a reliable answer? And why hasn’t this been done yet, considering the facts that hundreds of millions of people worldwide take these pills and there have been serious doubts about their safety for many years?
After reading this observational study in BMJ Open, I have this question:
How can an association be made between sleeping pills and premature mortality when it is unknown whether any patients in the study actually ingested sleeping pills?
Financial ties are only one source, among many, of possible bias. I see nothing wrong with requiring disclosure of financial ties in order to lay bare this “low hanging fruit” of potential bias, while bearing in mind that, leaving money aside, nearly everyone has an axe to grind. Research is rarely rewarded by wealth or fame, so there’s usually some passion for the topic itself.
This is partly why I’m skeptical of financial disclosure as a cure-all for research or CME bias. It’s a starting point at best. “Consumers” of research or medical education still need to employ a critical eye themselves.
Dr. Balt, please get your facts straight and refrain from character assassination when your beliefs are challenged. If you took the time to check, you’d see that I have been on the pharma side of HRT litigation because that’s what’s supported by the science. I have no interest in pandering to anyone — but I do hold the best interests of patients and the public health uppermost in my practice. When the evidence supports treatment, I’m for it. When the evidence suggests meaningful harm, I remember the Hippocratic Oath. I’d suggest you do the same. Look again at our study. You’ll find that we tried every available way to remove confounders and still found striking associations. Other studies that we cited have looked at the impact of things we could not control, including psychiatric diagnoses, on these hazards and found that they do not account for the harm. Our paper is not an isolated piece, we cite 18 others that have also found harm. Your suggestion that the survival plots for our sleep duration paper and the hypnotics analysis look the same is particularly disingenuous. If you honestly understand survival analysis, you’d recognize that nothing about these is similar other than the fact that they were both generated from Cox models.
Well said. Perhaps you are a better person to ask (see my earlier comment): how would you set up research to prove without doubt what the effects of these pills on mortality are? I know you’ve tried your best with this study, but apparently it is not enough to convince critics.
Thanks for asking. One of the things we dearly hope our paper will do is to prompt further research on this question. So far about 20 observational studies, including ours, have found harm. The RCTs done for licensing, typically with about 6 months of exposure per subject, have not been long enough to detect a mortality hazard. Of course the best test would be a randomized clinical trial. But that will not happen because an RCT with sufficient length and representation is unethical given the consistent finding of harm in the available literature. It would also be prohibitively expensive because the background mortality rate in the population of greatest interest — otherwise healthy people in the prime of life — is tiny, so it would take an enormous number of person years experience to achieve statistical significance. So we’re left to find the best strategies using observational data — hopefully in a variety of populations since none will be perfect — so that the missing elements or residual confounding in each will be offset by strengths in others. Matched case-control or matched cohort designs like the one we employed are the most efficient strategies in this context.
The ideal population would be unselected, that is all people with potential exposure to these drugs — not people studied because they have a particular kind of insurance, or met some criteria that made them eligible for a particular study of heart disease, cancer, diabetes, etc. The available data would include drug dispensing information (with dates) for all prescribed medications and the indication for that drug; all diagnoses (with dates) classified by ICD-9 or ICD-10 codes; vital status with last follow-up date, and for decedents the date of death and primary cause of death, reviewed and recoded if appropriate according to accepted standards. For reference, the population we studied was unselected; had drug order data with dates and indication, but not dispensing data; had all diagnoses with ICD-9 codes except that we were required to delete the mental health codes because of a state law protecting that information; and had vital status data verified by the Social Security Death Index, but not cause of death.
Dr. Langer,
I think you nailed the issue. You say, “the background mortality rate in the population of greatest interest — otherwise healthy people in the prime of life — is tiny, so it would take an enormous number of person years experience to achieve statistical significance.”
According to the book “Stats.con” by James Penston, the larger the sample size needed to find a difference, the less clinically significant any difference is.
I assume your finding of the increased risk is stated in terms of relative risk, not absolute risk. Big difference
And of course, with such small numbers of people dying from the increased risk, any finding must be tempered with the knowledge that the population of people taking sedatives probably contains of a higher percentage of people who do other things that, when mixed with sedatives (e.g. excessive alcohol intake, or driving while sedated, to name but two) can lead to death. Therefore, the taking of sedatives all by itself may confer little risk.
While obviously caution should be excercized in taking any drug with potential side effects, I think you are needlessly frightening the public.
Dr. Allen, The hazard was present in all age groups and within all subgroups defined by specific diagnoses. I suggested that an RCT should focus on low risk people because Elmar Veerman asked how I would “set up research that would provide without doubt” the mortality association. An RCT in low risk people would provide a virtually irrefutable answer. We did use hazard ratios in our paper since that is the convention in the literature, but I agree that absolute risks are probably a better measure and I have calculated them. The absolute risk of mortality in our study population overall was 19 per 1000 person years. Among patients with no known disease it was 7 per 1000 person years. Another way of looking at this is Number Needed to Harm. In our study overall, one person was harmed for every 52 people who took sleeping pills, and among those with no known disease one person was harmed for every 137 who took sleeping pills.
Jackie, Yes, your reading is correct. We had information on orders, not pick-up at a pharmacy. That was because this is a large outpatient population spread over hundreds of square miles in central Pennsylvania. These folks, like most of us, get their prescriptions filled at neighborhood pharmacies and in the US we have no central reporting of that information. That said, we found a trend for increasing harm with increasing numbers of pills ordered, which suggests that the association was linked to consumption of the drugs. We agree that cause of death information would have been very useful. It simply was not available. However, the absence of that specificity does not give us license to ignore a stunningly high risk.
Altostrata, I strongly agree with your perspective. The important take-home from this report is that these drugs are associated with substantial risks that most patients, and most prescribing doctors, have discounted. There are lots of reasons for this. Among them was the notion that the newer, shorter-acting drugs would be safer. In our study they were as risky as the older drugs. I do not believe that these drugs should be ditched. I do believe that the risks need to be acknowledged and managed responsibly by both patients and doctors.
Dr. Allen, that to me is an argument for making wider distribution of the risk elevation, providing the conditions you just mentioned are explicit.
Alcohol use is widespread and patients are either ignorant of or cut corners on the dangers of combination with sedatives. They do keep on accidentally (or accidentally on purpose) killing themselves while on them.
This has already emerged as an epidemic in “abuse” of prescription drugs.
While it may be true the drugs as inanimate objects don’t kill people — people kill people — it is prudent for doctors to minimize throwing the ammunition out into the general public. Making a big public issue of the risk serves to inform doctors who for some reason are convinced prescribing psychiatric drugs is a no-brainer.
So much of the debate about psychiatric drug dangers comes down to excessive and irresponsible prescribing, but putting the red flag on the drugs avoids pointing the finger at doctors, who so richly deserve more intense scrutiny themselves.
Dr. Langer,
Two questions from a layperson:
1. It seemed to me that your study equated “prescriptions written” with “pills taken.” Is that a correct reading?
2. Why didn’t your study look at “cause of death” to exclude things like drive-by shootings, mining accidents, and other events that have no connections to sleeping pills?
Thanks
Mining accidents may very well have a connection to the use of sleeping pills. If someone causes an accident because he is still sleepy from taking a pill, for example.
@Elmar Veerman
That’s why I added the qualifier ” … that have no connections to sleeping pills.”
For example, fatalities due to a mine collapse caused by an earthquake would have nothing to do with sleeping pills.
—————
My apologies for originally posting this reply in the wrong place above.
Dr Langer,
Thank you for your response. My intent is decidedly not to assassinate your character, and as for my “beliefs,” they essentially align with yours and Dr Kripke’s (see my 3rd-to-last paragraph), so they (i.e., I) do not feel challenged at all.
I appreciate your clarification of your stance on HRT. (And I have corrected the relevant statement in my post.) The reason I pointed out your involvement in this matter was to show that simply focusing on pharmaceutical-company conflicts of interest doesn’t always tell “the whole story.” Does a reader need (or deserve) to know whether an author is a speaker for a drug firm? Does a reader need to know that an author has provided his services as an expert witness in legal proceedings? Does a reader need to know that an author has published numerous studies reaching a similar conclusion as the present one? These are the open questions I pose with this post. And there are no easy answers.
Regarding the survival plots, the fact that they look similar is not disingenuous (nor ingenuous [?]). It is simply apparent to the naked eye. But the underlying question (which was, BTW, not the point of this post, but something that came to my mind as I looked at the data) is whether increased sleep time in and of itself (i.e., with or without sedatives) is a mortality risk factor. Obviously, the BMJ piece didn’t take sleep duration into account. Maybe the mortality culprit was sleep duration and not sedative use? That seems unlikely to me, but not clear with the available data.
Ultimately, I appreciate and am humbled by your (and Dr Kripke’s) presence here. But if you read other posts in my blog you’ll realize that I, too, have a distinctly anti-sedative, anti-benzodiazepine bias, so (and, yes, this is my conflict of interest) I have no reason to question your findings. The whole point of the post was to show that a study emphasizing an anti-medication point of view (a point of view that is currently in vogue, especially in the mass media) can receive extensive media coverage without any question of authors’ conflicts or biases, while papers with a pro-drug opinion get run through the conflict-of-interest ringer (the best recent example is Ian Hickie’s Lancet article on agomelatine), sometimes obscuring the underlying science.
Readers of the BMJ Open article on sleeping pills at:
http://bmjopen.bmj.com/content/2/1/e000850.full will find my financial interest disclosed: since the family owns a mutual fund which has invested a bit in Sanofi and J&J, I will lose money if our article makes the stock go down. It is worth the personal loss to save lives.
Readers of http://www.DarkSideOfSleepingPills.com will find in Chapter 11 a more detailed disclosure of my financial interests.
Daniel F. Kripke, M.D.
Dr Kripke,
Thanks for your comment. The purpose of my post was not to emphasize your financial relationships (or lack thereof), but rather to point out the curious double-standard in which we feel an obligation to reveal every financial interest (however small) of medical researchers, while ignoring philosophical biases or opinions that may have just as much of an impact on what they (choose to) publish.
While in college at Stanford in the early 1990s I worked as a research assistant for a fairly well-known behavioral neuroscientist, and helped him with some pop-science books and magazine articles. He would ask me a question, I’d go to the library and dig up whatever I could on the topic, and then he would conveniently ignore the information that didn’t support his argument. Granted, we weren’t writing for peer-reviewed journals, but it gave me my first look at how one’s own beliefs can easily lead to selective reporting of data, which, in turn, can skew outcomes.
I don’t really understand what you want. No obligation to reveal financial interests? Or an obligation to reveal all relevant opinions with every research article?
I share EV’s confusion. Currently we aim to control for possible biases (eg, those stemming from financial ties) that can be addressed in a systematic way, while recognizing many other possible biases we can’t account for. Is there a better way?
This situation parallels actually doing clinical research. Good clinical studies control for extraneous variables whenever possible, but it’s rarely possible to control for all of them. In both domains, doing research and reading about it, we do the best we can. We then discuss the remaining shortcomings in the discussion section of the paper or the blogosphere respectively.
SR and EV,
There is no easy answer, and I don’t know of any “better way.” I think Dr Reidbord’s comments sum it up nicely: it’s rarely possible to control for every potential conflict. Thus, a sole focus on financial ties with “big pharma” might prevent the critical reader from recognizing other conflicts, which could be substantial (if, at first glance, “merely” philosophical). Should we “reveal all relevant opinions”? Of course not… but a long history of articles saying the same darn thing might make the critical reader sit up and take notice.
Dr Reidbord’s comment about discussing shortcomings “in the blogosphere” deserves particular attention. The fact that the two lead authors of this widely-read study have directly responded to readers here (and within days of publication of the original article) is impressive, and speaks to the power of online discourse. The potential impact of this medium on clinical practice– and patient empowerment– is profound, and (as yet) largely unrealized.
Dr. Kripke,
Your stoic willingness to suffer “personal loss” is no doubt admirable to some, but count me out of that number.
Pray tell, what exactly are your financial and personal interests in furthering alternatives to pharma-therapy for sleep disorders? Is it possible that an uptick in the demand for cognitive behavioural therapy might put a few sheckles in the old family purse? Speaking honoraria? Book sales/Web traffic?
Or is this really about self-aggrandizement? The fact that you’re stalking the comments sections of blogs to defend your work shows me that you’re on some sort of a crusade. More power to you, I guess, but please drop the patronizing charade that somehow publishing your work will bring your family hardship.
Real scientists defend their work in the peer reviewed literature.
Bretty
Personally, I wouldn’t read anything into participating in the comments sections of blogs. All authors are interested in reaction to their work.
Here http://www.medpagetoday.com/MeetingCoverage/AAAAIMeeting/31482 allergy researchers have found multiple drug exposure, adverse events, and polypharmacy (explicitly including narcotics and antidepressants) promote iatrogenic medication sensitivity. (They don’t have a medication ax to grind; they complain drug sensitivities are misdiagnosed as drug “allergy”).
While drug sensitivities are not life-threatening in the anaphylaxis sense, they can produce a lot of wear and tear on organs such as the heart, such as overstimulation. (Full disclosure: I’m coping now with arrhythmia following years of severe and prolonged Paxil withdrawal syndrome.)
Another argument for minimizing the prescription of any chronic medication, particularly psychiatric medications that reverberate throughout the nervous system and for which non-drug treatments are available.
@ Robert D. Langer: Wait, you say a randomized controlled trial can’t be done ‘because an RCT with sufficient length and representation is unethical given the consistent finding of harm in the available literature.’ Wouldn’t that make the present situation – giving millions of people these drugs without a placebo group – even more unethical?
Couldn’t you do a study comparing these pills to melatonin or behavioral therapy? Would be very expensive, of course. And I don’t expect the drug companies to pay for that.
Elmar Veerman, Use of these drugs can be ethical if the hazards are acknowledged. I hope that the findings in our paper will help change the perspective from the current discounting of risk, to recognition that these drugs must be used with considerable caution. The study you suggest comparing these drugs to melatonin and/or cognitive behavioral therapy would be interesting. But if it were intended to assess potential differences in mortality, I believe it would be unethical given the strong evidence for increased mortality associated with the drugs. Ultimately that decision would be made by Human Subjects review committees who would be charged with assessing the potential dangers to participants based on available data. In my experience this would not pass such review.
[…] especially after long-term, after controlling for many factors such as smoking. There are some criticisms of the methods used and the fact that the authors were well known critics of hypnotic use for […]
mobile phones…
[…]Sleeping Pills Are Deadly? Says Who, Exactly? « Thought Broadcast[…]…
Get over yourself Dr. Kripke, there are more important things causing loss of life. Life is tough and some people need to sleep period! Let them sleep and don’t try to create such panic over some studies which lack a causal relationship!
Thank you Dr. Balt. At last, an objective voice of reason!
Because the metal can tarnish over time, gold objects are often not appreciated and sold for pennies.
There was very little in the news, however, about the
horrors of his concentration camps what would eventually take the lives of six million Jews and
four million others. America’s War Years on the Home Front (1941-1945)An Unexpected
View of History – By: J.