The Prescription Pain Pill Epidemic: A Conversation with Dr. Anna Lembke

back-pain-in-seniors-helped-with-mindfulness-300x200manypills
My colleague, Dr. Anna Lembke is the Program Director for the Stanford University Addiction Medicine Fellowship, and Chief of the Stanford Addiction Medicine Dual Diagnosis Clinic. She is the author of a newly released book on the prescription pain pill epidemic: “Drug Dealer, MD: How Doctors Were Duped, Patients Got Hooked, and Why It’s So Hard to Stop” (Johns Hopkins University Press, October 2016).

I spoke with her recently about the scope of this public health tragedy, how we got here and what we need to do about it.

Dr. Jain: About 15-20 years ago American medicine underwent a radical cultural shift in its attitude towards pain, a shift that ultimately culminated in a public health tragedy. Can you comment on factors that contributed to that shift occurring in the first place?
Dr. Lembke: Sure. So the first thing that happened (and it was really more like the early 1980’s when this shift occurred) was that there were more people with daily pain. Overall, our population is getting healthier, but we also have more people with more pain conditions. No one really knows exactly the reason for that, but it probably involves people living longer with chronic illnesses, and more people getting surgical interventions for all types of condition. Any time you cut into the body, you cut across the nerves and you create the potential for some kind of neuropathic pain problem.
The other thing that happened in the 1980’s was the beginning of the hospice movement. This movement helped people at the very end of life (the last month to weeks to days of their lives) to transition to death in a more humane and peaceful way. There was growing recognition that we weren’t doing enough for people at the end of life. As part of this movement, many doctors began advocating for using opioids more liberally at the end of life.
There was also a broader cultural shift regarding the meaning of pain. Prior to 1900 people viewed pain as having positive value: “what does not kill you makes you stronger” or “after darkness comes the dawn”. There were spiritual and biblical connotations and positive meaning in enduring suffering. What arose, through the 20th century, was this idea that pain is actually something that you need to avoid because pain itself can lead to a psychic scar that contributes to future pain. Today, not only is pain painful, but pain begets future pain. By the 1990’s, pain was viewed as a very bad thing and something that had to be eliminated at all cost.
Growing numbers of people experiencing chronic pain, the influence of the hospice movement, and a shifting paradigm about the meaning and consequences of experiencing pain, led to increased pressures within medicine for doctors to prescribe more opioids. This shift was a departure from prior practice, when doctors were loathe to prescribe opioids, for fear of creating addiction, except in cases of severe trauma, cases involving surgery, or cases of the very end of life.
Dr. Jain: The American Pain Society had introduced “pain as the 5th vital sign,” a term which suggested physicians, who were not taking their patients’ pain seriously, were being neglectful. What are your thoughts about this term?
Dr. Lembke: “Pain is the 5th vital sign” is a slogan. It’s kind of an advertising campaign. We use slogans all the time in medicine, many times to good effect, to raise awareness both inside and outside the profession about a variety of medical issues. The reason that “pain is the 5th vital sign” went awry, however, has to do with the ways in which professional medical societies, like the American Pain Society, and so-called “academic thought leaders”, began to collaborate and cooperate with the pharmaceutical industry. That’s where “pain is the 5th vital sign” went from being an awareness campaign to being a brand for a product, namely prescription opioids.
So the good intentions in the early 1980’s turned into something really quite nefarious when it came to the way that we started treating patients. To really understand what happened, you have to understand the ways in which the pharmaceutical industry, particularly the makers of opioid analgesics, covertly collaborated with various institutions within what I’ll call Big Medicine, in order to promote opioid prescribing.
Dr. Jain: So by Big Medicine what do you mean?
Dr. Lembke: I mean the Federation of State Medical Boards, The Joint Commission (JACHO), pain societies, academic thought leaders, and the Food and Drug Administration (FDA). These are the leading organizations within medicine whose job it is to guide and regulate medicine. None of these are pharmaceutical companies per se, but what happened around opioid pain pills was that Big Pharma infiltrated these various organizations in order to use false evidence to encourage physicians to prescribe more opioids. They used a Trojan Horse approach.. They didn’t come out and say we want you to prescribe more opioids because we’re Big Pharma and we want to make more money, instead what they said was we want you to prescribe more opioids because that’s what the scientific evidence supports.
The story of how they did that is really fascinating. Let’s take The Joint Commission (JACHO) as an example. In 1996, when oxycontin was introduced to the market, JACHO launched a nationwide pain management educational program where they sold educational materials to hospitals, which they acquired for free from Purdue Pharma. These materials included statements which we now know to be patently false. JACHO sold the Purdue Pharma videos and literature on pain to hospitals.
These educational materials perpetuated four myths about opioid prescribing. The first myth was that opioids work for chronic pain. We have no evidence to support that. The second was that no dose is too high. So if your patient responds to opioids initially and then develops tolerance, just keep going up. And that’s how we got patients on astronomical amounts of opioids. The third myth was about pseudo addiction. If you have a patient who appears to be demonstrating drug seeking behavior, they’re not addicted. They just need more pain meds. The fourth and most insidious myth was that there is a halo effect when opioids are prescribed by a doctor, that is, they’re not addictive as long as they’re being used to treat pain.
So getting back to JACHO, not only did they use material propagating myths about the use of opioids to treat pain, but they also did something that was very insidious and, ultimately, very bad for patients. They made pain a “quality measure”. By The Joint Commission’s own definition of a quality measure, it must be something that you can count. So what they did was they created this visual analog scale, also known as the “pain scale”. The scale consists of numbers from one to ten describing pain, with sad and happy faces to match. JAHCO told doctors they needed to use this pain scale in order to assess a patients’ pain. What we know today is that this pain scale has not led to improved treatment or functional outcomes for patients with pain. The only thing that it has been correlated with is increased opioid prescribing.
This sort of stealth maneuver by Big Pharma to use false evidence or pseudo-science to infiltrate academic medicine, regulatory agencies, and academic societies in order to promote more opioid prescribing: that’s an enduring theme throughout any analysis of this epidemic.
Dr. Jain: Can you comment specifically on the breadth and depth of the opioid epidemic in the US? What were the key factors involved?
Dr. Lembke: Drug overdose is now the leading cause of accidental death in this country, exceeding death due to motor vehicle accidents or firearms. Driving this statistic is opioid deaths and driving opioid deaths is opioid pain prescription deaths, which in turn correlates with excessive opioid prescribing. There are more than 16,000 deaths per year due to prescription opioid overdoses.
What’s really important to understand is that an opioid overdose is not a suicide attempt. The vast majority of these people are not trying to kill themselves, and many of them are not even taking the medication in excess. They’re often taking it as prescribed, but over time are developing a low grade hypoxia. They may get a minor cold, let’s say a pneumonia, then they’ll take the pills and they’ll fall asleep and won’t wake up again because their tolerance to the euphorigenic and pain effects of the opioids is very robust, but their tolerance to the respiratory suppressant effect doesn’t keep pace with that. You can feel like you need to take more in order to eliminate the pain, but at the same time the opioid is suppressing your respiratory drive, so you eventually become hypoxemic and can’t breathe anymore and just fall into a gradual sleep that way.
There are more than two million people today who are addicted to prescription opioids. So not only is there this horrible risk of accidental death, but there’s obviously the risk of addiction. We also have heroin overdose deaths and heroin addiction on the rise, most likely on the coattails of the prescription opioids epidemic, driven largely by young people who don’t have reservations about switching from pills to heroin..
Dr. Jain: I was curious about meds like oxycontin, vicodin, and percocet. Are they somehow more addictive than other opioid pills?
Dr. Lembke: All opioids are addictive, especially if you’re dealing with an opioid naive person. But it is certainly true that some of the opioids are more addictive than others because of pharmacology. Let’s consider oxycontin. The major ingredient in oxycontin is oxycodone. Oxycodone is a very potent synthetic opioid. When Purdue formulated it into oxycontin, what they wanted to create was a twice daily pain medication for cancer patients. So they put this hard shell on a huge 12 hours worth of oxycodone. That hard shell was intended to release oxycodone slowly over the course of the day. But what people discovered is that if they chewed the oxycontin and broke that hard shell, then they got a whole day’s worth of very potent oxycodone at once. With that came the typical rush that people who are addicted to opioids describe, as well as this long and powerful and sustained high. So that is why oxycontin was really at the center of the prescription opioid epidemic. It basically was more addictive because of the quantity and potency once that hard shell was cracked.
Dr. Jain: So has the epidemic plateaued? And if so, why?
Dr. Lembke: The last year for which we have CDC data is 2014, when there were more prescription opioid-related deaths, and more opioid prescriptions written by doctors, than in any year prior. This is remarkable when you think that by 2014, there was already wide-spread awareness of the problem. Yet doctors were not changing their prescribing habits, and patients were dying in record numbers.
I’m really looking forward to the next round of CDC data to come out and tell us what 2015 looked like. I do not believe we have reached the end or even the waning days of this epidemic. Doctors continue to write over 250 million opioid prescriptions annually, a mere fraction of what was written three decades ago.
Also, the millions of people who have been taking opioids for years are not easily weaned from opioids.. They now have neuroadaptive changes in their brains which are very hard to undo. I can tell you from clinical experience that even when I see patients motivated to get off of their prescription opioids, it can take weeks, months, and even years to make that happen.
So I don’t think that the epidemic has plateaued, and this is one of the major points that I try to make in my book. The prescription drug epidemic is the canary in the coal mine. It speaks to deeper problems within medicine. Doctors get reimbursed for prescribing a pill or doing a procedure, but not for talking to our patients and educating them. That’s a problem. The turmoil in the insurance system we can’t even establish long term relationships with our patients. So as a proxy for real healing and attachment, we prescribe opioids. ! Those kinds of endemic issues within medicine have not changed, and until they do, I believe this prescription drug problem will continue unabated.

‘Replace male doctors with female ones and save at least 32,000 lives each year’?

The authors of a recent article in JAMA Internal Medicine

Physician Gender and Outcomes of Hospitalized Medicare Beneficiaries in the U.S.,” Yusuke Tsugawa, Anupam B. Jena, Jose F. Figueroa, E. John Orav, Daniel M. Blumenthal, Ashish K. Jha, MD, MPH1,2,8, JAMA Internal Medicine, online December 19, 2016, doi: 10.1001/jamainternmed.2016.7875

Stirred lots of attention in the media with direct quotes like these:

“If we had a treatment that lowered mortality by 0.4 percentage points or half a percentage point, that is a treatment we would use widely. We would think of that as a clinically important treatment we want to use for our patients,” said Ashish Jha, professor of health policy at the Harvard School of Public Health. The estimate that 32,000 patients’ lives could be saved in the Medicare population alone is on par with the number of deaths from vehicle crashes each year.

Washington Post: Women really are better doctors, study suggests.

LA  Times: How to save at least 32,000 lives each year: Replace male doctors with female ones.

NPR: Patients cared for by female doctors fare better than those treated by men.

My immediate reactions after looking at the abstract were only confirmed when I delved deeper.

Basically, we have a large, but limited and very noisy data set. It is unlikely that these data allow us to be confident about the strength of any signal concerning the relationship between physician gender and patient outcome that is so important to the authors. The small apparent differences could be just more noise on which the authors have zeroed in so that they can make a statement about the injustice of gender differences in physician pay.

 I am unwilling to relax methodological and statistical standards to manufacture support for such a change. There could be unwanted consequences of accepting that arguments can be made with such weak evidence, even for a good cause.

What if the authors had found the same small differences in noisy data in the reverse direction? Would they argue that we should preserve gender differences in physician pay? What if the authors focus on a different variable in all this noise and concluded that lower pay which women receive was associated with reduced mortality? Would we then advocate that will reduce the pay of both male and female physicians in order to improve patient outcomes?

Despite all the excitement that claim about an effect of physician gender on patient mortality is generating, it is most likely that we are dealing with noise arising from overinterpretation of complex analyses that assume more completeness and precision than can be found in the data being analyzed.

These claims are not just a matter of causal relationships being spun from correlation. Rather, they are causal claims being made on the basis of partial correlations emerging in complex multivariate relationships found in an administrative data set.

  • Administrative data sets, particularly Medicaid data sets like this one, are not constructed with such research questions in mind. There are severe constraints on what variables can be isolated and which potential confounds can be identified and tested.
  • Administrative data sets consist of records, not actual behaviors. It’s reasonable to infer a patient death associated with a record of a death. Association of a physician gender associated with a particular record is more problematic, as we will see. Even if we accept the association found in these records, it does not necessarily mean that physicians engaged in any particular behaviors or that the physician behavior is associated with the pattern of deaths emerging in these multivariate analyses.
  • The authors start out with a statement about differences in how female and male physicians practice. In the actual article and the media, they have referred to variables like communication skills, providing evidence-based treatments, and encouraging health-related behaviors. None of these variables are remotely accessible in a Medicaid data set.
  • Analyses of such administrative data sets do not allow isolation of the effects of physician gender from the effects of the contexts in which their practice occurs and relevant associated variables. We are not talking about a male or female physician encountering a particular patient being associated with a death or not, but an administrative record of physician gender arising in a particular context being interpreted as associated with a death. Male and female physicians may differ in being found in particular contexts in nonrandom fashion. It’s likely that these differences will dwarf any differences in outcomes. There will be a real challenge in even confidently attributing those outcomes to whether patients had an attending male or female physician.

The validity of complex multivariate analyses are strongly threatened by specification bias and residual confounding. The analyses must assume that all of the relevant confounds have been identified and measured without error. Departures from these ideal conditions can lead to spurious results, and generally do. Examination of the limitations in the variables available in a Medicaid data set and how they were coded can quickly undermine any claim to validity.

Acceptance of claims about effects of particular variables like female physician gender arising in complex multivariate analyses involve assumptions of “all-other-things-being-equal.” If we attempt to move from statistical manipulation to inference about a real world encounter, we no longer talking about a particular female physician, but a construction that may be very different from particular physicians interacting with particular patients in particular contexts.

The potential for counterfactual statements can be seen if we move from the study to one of science nerds and basketball players and hypothesize if John and Jason were of equivalent height, John would not study so hard.

Particularly in complex social situations, it is usually a fantasy that we can change one variable, and only one variable, not others. Just how did John and Jason get of equal height? And how are they now otherwise different?

Associations discovered in administrative data sets most often do not translate into effects observed in randomized trials. I’m not sure how we could get a representative sample of patients to disregard their preferences and accept random assignment to a male or female physician. It would have to be a very large study to detect the effect sizes reported in this observational study, and I’m skeptical this sufficiently strong signal would emerge from all of the noise.

We might relax our standards and accept a quasi-experimental design that would be smaller but encompass a wider range of relevant variables. For instance, it is conceivable that we could construct a large sample in which physicians varied in terms of whether they had formal communication skills training. We might examine whether communications training influenced subsequent patient mortality, independent of physician gender, and vice versa. This would be a reasonable translation of the authors’ hypothesis that communication skills differences between male and female physicians account for what the authors believe is the observed association between physician gender and mortality. I know of no such study having been done. I know of no study demonstrating that physician communication training affects patient mortality. I’m skeptical that the typical communication training is so powerful in its effects. If such a study required substantial resources, rather than relied on data on hand, I would not be encouraged to invest in it by the strength of the results of the present study to marshal those resources.

What I saw when I looked at the article

 We dealing with very small adjusted differences in percentage arising in a large sample.

Patients treated by female physicians had lower 30-day mortality (adjusted mortality, 11.07% vs 11.49%; adjusted risk difference, –0.43%; 95% CI, –0.57% to –0.28%; P < .001; number needed to treat to prevent 1 death, 233).

Assignment of a particular patient to a particular physician is done with a lot of noise.

We assigned each hospitalization to a physician based on the National Provider Identifier in the Carrier File that accounted for the largest amount of Medicare Part B spending during that hospitalization.25 Part B spending comprises professional and other fees determined by the physician. On average, these physicians were responsible for 51.1% of total Part B spending for a given hospitalization.

One commentator quoted in a news article noted:

William Weeks, a professor of psychiatry at Dartmouth’s Geisel School of Medicine, said that the researchers had done a good job of trying to control for other factors that might influence the outcome. He noted that one caveat is that hospital care is usually done by a team. That fact was underscored by the method the researchers used to identify the doctor who led the care for patients in the study. To identify the gender of the physician, they looked for the doctor responsible for the biggest chunk of billing for hospital services — which was, on average, about half. That means that almost half of the care was provided by others.

Actually, much of the care is not provided by the attending physician, but other staff, including nurses and residents.

The authors undertook the study to call attention to gender disparities in physician pay. But could disparities show up in males being able to claim more billable procedures – greater credit administratively for what is done with patients during hospitalization, including by other physicians? This might explain at least some of the gender differences, but could undermine the validity of this key variable in relating physician gender to differences in patient outcome.

The statistical control of differences in patient and physician characteristics afforded by variables in this data set is inadequate.

Presumably, a full range of patient variables is related to whether patients die within 30 days of a hospitalization. Recall the key assumption that all of the relevant confounds have been identified and assessed without error in considering the variables used to characterize patient characteristics:

Patient characteristics included patient age in 5-year increments (the oldest group was categorized as ≥95 years), sex, race/ethnicity (non-Hispanic white, non-Hispanic black, Hispanic, and other), primary diagnosis (Medicare Severity Diagnosis Related Group), 27 coexisting conditions (determined using the Elixhauser comorbidity index28), median annual household income estimated from residential zip codes (in deciles), an indicator variable for Medicaid coverage, and indicator variables for year.

Note that the comorbidity index is based on collapsing 27 other variables into one number. Simplifies the statistics, yes, but with a tremendous loss of information.

Recall the assumption that this set of variables represent not just what is available in administrative data set, but all the patient characteristics relevant to their dying within 30 days after discharge from the hospital. Are we really willing to accept this assumption?

For the physician variables displayed at the top of Table 1, there are huge differences between male and female physicians, relative to the modest difference in patient mortality, adjusted mortality, 11.07% vs 11.49%.

smaller table of patient characiteristics

These authors encourage us to think about the results as simulating a randomized trial, except that statistical controls are serving the function that randomization of patients to physician gender would serve. We are being asked to accept that these difference in baseline characteristics of the practices of female versus physicians can be eliminated through statistics. We would never accept that argument in a randomized trial.

Addressing criticisms of the authors interpretation of their results.

 The senior author provided a pair of blog posts in which he acknowledges criticism of his study, but attempts to defuse key objections. It’s unfortunate that the sources of these objections are not identified, and so we dependent on the author’s summary out of context. I think the key responses are to straw man objections.

Correlation, Causation, and Gender Differences in Patient Outcomes

Do women make better doctors than men?

Correlation is not causation.

We often make causal inferences based on observational data – and here’s the kicker: sometimes, we should.  Think smoking and lung cancer.  Remember the RCT that assigned people to smoking (versus not) to see if it really caused lung cancer?  Me neither…because it never happened.  So, if you are a strict “correlation is not causation” person who thinks observational data only create hypotheses that need to be tested using RCTs, you should only feel comfortable stating that smoking is associated with lung cancer but it’s only a hypothesis for which we await an RCT.  That’s silly.  Smoking causes lung cancer.

No, it is this argument that is silly. We can now look back on the data concerning smoking and lung cancer and benefit from the hindsight provided by years of sorting smoking as a risk factor from potential confounds.  I recall at some point, drinking coffee being related to lung cancer in the United States, whereas drinking tea was correlated in the UK. Of course, if we don’t know that smoking is the culprit, we might miss that in the US, smoking was done while drinking coffee, whereas the UK, while drinking tea.

And isolating smoking as a risk factor, rather than just a marker for risk, is so much simpler than isolating whatever risk factors for death are hidden behind physician gender as a marker for risk of mortality.

Coming up with alternative explanations for the apparent link between physician gender and patient mortality.

The final issue – alternative explanations – has been brought up by nearly every critic. There must be an alternative explanation! There must be confounding!  But the critics have mostly failed to come up with what a plausible confounder could be.  Remember, a variable, in order to be a confounder, must be correlated both with the predictor (gender) and outcome (mortality).

This is similarly a fallacious argument. I am not arguing for alternative substantive explanations, I’m proposing that spurious results were produced by pervasive specification bias, including measurement error. There is no potential confounder I have to identify. I am simply arguing that that the small differences in mortality are dwarfed by specification and measurement error.

This tiny difference is actually huge in its implications.

Several critics have brought up the point that statistical significance and clinical significance are not the same thing.  This too is epidemiology 101.  Something can be statistically significant but clinically irrelevant.  Is a 0.43 percentage point difference in mortality rate clinically important? This is not a scientific or a statistical question.  This is a clinical question. A policy and public health question.  And people can reasonably disagree.  From a public health point of view, a 0.43 percentage point difference in mortality for Medicare beneficiaries admitted for medical conditions translates into potentially 32,000 additional deaths. You might decide that this is not clinically important. I think it is. It’s a judgment call and we can disagree.

The author taking a small difference and magnifying its importance by applying to a larger population. He is attributing the “additional deaths” to patients being treated by men. I feel he hasn’t made a case that physician gender is the culprit and so nothing is accomplished except introducing shock and awe by amplifying the small effect into its implications for the larger population.

In response to a journalist, the author makes a parallel argument:

The estimate that 32,000 patients’ lives could be saved in the Medicare population alone is on par with the number of deaths from vehicle crashes each year.

In addition to what I have already argued, if we know the same number of deaths are attributable to automobile crashes, we at least know how to take steps to reduce these crashes and the mortality associated with them. We don’t know how to change the mortality the authors claim is associated with physician gender. We don’t even know that the author’s claims are valid.

Searching for meaning where meaning no meaning is to be found.

In framing the study and interpreting the results to the media, the authors undertake a search of the literature with a heavy confirmation bias, ignoring the many contradictions that are uncovered with a systematic search. For instance, one commentator on the senior author’s blog notes

It took me about 5 minutes of Google searching to find a Canadian report suggesting that female physicians in that country have workloads around 75% to 80% of male physicians:

https://secure.cihi.ca/free_products/PracticingPhysicianCommunityCanada.pdf

If US data is even vaguely similar, that factor would be a serious omission from your article.

But the authors were looking for what supported the results, not for studies that potentially challenged or contradicted their results. They are looking to strengthen a narrative, not expose it to refutation.

Is there a call to action here?

As consumers of health services, we could all switch to being cared for by female physicians. I suspect that some of the systems and structural issues associated with the appearance that care by male physicians inferior would be spread among females, including increased workloads. The bias in the ability of male physicians to claim credit for the work of others would be redistributed to women. Neither would improve patient mortality.

We should push for reduction in inequalities in pay related to gender. But we don’t need results of this study to encourage us.

I certainly know health care professionals and researchers who have more confidence in communication learning modules producing clinically significant changes in position behavior. I don’t know any of them who could produce evidence that these changes include measurable reductions in patient mortality. If someone produces such data, I’m capable of being persuaded. But the present study adds nothing to my confidence in that likelihood.

If we are uncomfortable with the communication skills or attention to evidence that our personal physicians display, we should replace them. But I don’t think this study provides additional evidence for us doing so, beyond the legitimacy of us acting on our preferences.

In the end, this article reminds us to stick to our standards and not be tempted to relax them to make socially acceptable points.

 

 

 

 

 

Trusted source? The Conversation tells migraine sufferers that child abuse may be at the root of their problems

Patients and family members face a challenge obtaining credible, evidence-based information about health conditions from the web.

Migraine sufferers have a particularly acute need because their condition is often inadequately self-managed without access to best available treatment approaches. Demoralized by the failure of past efforts to get relief, some sufferers may give up consulting professionals and desperately seek solutions on Internet.

A lot of both naïve and exploitative quackery that awaits them.

Even well-educated patients cannot always distinguish the credible from the ridiculous.

One search strategy is to rely on websites that have proven themselves as trusted sources.

The Conversation has promoted itself as such a trusted source, but its brand is tarnished by recent nonsense we will review concerning the role of child abuse in migraines.

Despite some excellent material that has appeared in other articles in The Conversation, I’m issuing a reader’s advisory:

exclamation pointThe Conversation cannot be trusted because this article shamelessly misinforms migraine sufferers that child abuse could be at the root of their problems.

The Conversation article concludes with a non sequitur that shifts sufferers and their primary care physicians away from getting consultation with the medical specialists who are most able to improve management of a complex condition.

 

The Conversation article tells us:

Within a migraine clinic population, clinicians should pay special attention to those who have been subjected to maltreatment in childhood, as they are at increased risk of being victims of domestic abuse and intimate partner violence as adults.

That’s why clinicians should screen migraine patients, and particularly women, for current abuse.

This blog post identifies clickbait, manipulation, misapplied buzz terms, and  misinformation – in the The Conversation article.

Perhaps the larger message of this blog post is that persons with complex medical conditions and those who provide formal and informal care for them should not rely solely on what they find on the Internet. This exercise specifically focusing on The Conversation article serves to demonstrate this.

Hopefully, The Conversation will issue a correction, as they promise to do at the website when errors are found.

We are committed to responsible and ethical journalism, with a strict Editorial Charter and codes of conduct. Errors are corrected promptly.

The Conversation article –

Why emotional abuse in childhood may lead to migraines in adulthood

clickbaitA clickbait title offered a seductive  integration of a trending emotionally laden social issue – child abuse – with a serious medical condition – migraines – for which management is often not optimal. A widely circulating estimate is that 60% of migraine sufferers do not get appropriate medical attention in large part because they do not understand the treatment options available and may actually stop consulting physicians.

Some quick background about migraine from another, more credible source:

Migraines are different from other headaches. People who suffer migraines other debilitating symptoms.

  • visual disturbances (flashing lights, blind spots in the vision, zig zag patterns etc).
  • nausea and / or vomiting.
  • sensitivity to light (photophobia).
  • sensitivity to noise (phonophobia).
  • sensitivity to smells (osmophobia).
  • tingling / pins and needles / weakness / numbness in the limbs.

Persons with migraines differ greatly among themselves in terms of the frequency, intensity, and chronicity of their symptoms, as well as their triggers for attacks.

Migraine is triggered by an enormous variety of factors – not just cheese, chocolate and red wine! For most people there is not just one trigger but a combination of factors which individually can be tolerated. When these triggers occur altogether, a threshold is passed and a migraine is triggered. The best way to find your triggers is to keep a migraine diary. Download your free diary now!

Into The Conversation article: What is the link between emotional abuse and migraines?

Without immediately providing a clicklink so that  readers can check sources themselves, The Conversation authors say they are drawing on “previous research, including our own…” to declare there is indeed an association between past abuse and migraines.

Previous research, including our own, has found a link between experiencing migraine headaches in adulthood and experiencing emotional abuse in childhood. So how strong is the link? What is it about childhood emotional abuse that could lead to a physical problem, like migraines, in adulthood?

In invoking the horror of childhood emotional abuse, the authors imply that they are talking about something infrequent – outside the realm of most people’s experience.  If “childhood emotional abuse” is commonplace, how could  it be horrible and devastating?

In their pursuit of click bait sensationalism, the authors have only succeeded in trivializing a serious issue.

A minority of people endorsing items concerning past childhood emotional abuse actually currently meet criteria for a diagnosis of posttraumatic stress disorder. Their needs are not met by throwing them into a larger pool of people who do not meet these criteria and making recommendations based on evidence derived from the combined group.

Spiky_Puffer_Fish_Royalty_Free_Clipart_Picture_090530-025255-184042The Conversation authors employ a manipulative puffer fish strategy [1 and  2 ] They take what is a presumably infrequent condition and  attach horror to it. But they then wildly increase the presumed prevalence by switching to a definition that arises in a very different context:

Any act or series of acts of commission or omission by a parent or other caregiver that results in harm, potential for harm, or threat of harm to a child.

So we are now talking about ‘Any act or series of acts? ‘.. That results in ‘harm, potential for harm or threat’? The authors then assert that yes, whatever they are talking about is indeed that common. But the clicklink to support for this claim takes the reader behind a pay wall where a consumer can’t venture without access to a university library account.

Most readers are left with the authors’ assertion as an authority they can’t check. I have access to a med school library and I checked. The link is  to a secondary source. It is not a systematic review of the full range of available evidence. Instead, it is a  selective search for evidence favoring particular speculations. Disconfirming evidence is mostly ignored. Yet, this article actually contradicts other assertions of The Conversation authors. For instance, the paywalled article says that there is actually little evidence that cognitive behavior therapy is effective for people whose need for therapy is only because they  reported abuse in early childhood.

Even if you can’t check The Conversation authors’ claims, know that adults’ retrospective of childhood adversity are not particularly reliable or valid, especially studies relying on checklist responses of adults to broad categories, as this research does.

When we are dealing with claims that depend on adult retrospective reports of childhood adversity, we are dealing with a literature with seriously deficiencies. This literature grossly overinterprets common endorsement of particular childhood experiences as strong evidence of exposure to horrific conditions. This literature has a strong confirmation bias. Positive findings are highlighted. Negative findings do not get cited much. Serious limitations in methodology and inconsistency and findings generally ignored.

[This condemnation is worthy of a blog post or two itself. But ahead I will provide some documentation.]

The Conversation authors explain the discrepancy between estimates based on administrative data of one in eight children suffering abuse or neglect before age 18 versus much higher estimates from retrospective adult reports on the basis of so much abuse going unreported.

The discrepancy may be because so many cases of childhood abuse, particularly cases of emotional or psychological abuse, are unreported. This specific type of abuse may occur within a family over the course of years without recognition or detection.

This could certainly be true, but let’s see the evidence. A lack of reporting could also indicate a lack of many experiences reaching a threshold prompting reporting. I’m willing to be convinced otherwise, but let’s see the evidence.

The link between emotional abuse and migraines

The Conversation authors provide links only to their own research for their claim:

While all forms of childhood maltreatment have been shown to be linked to migraines, the strongest and most significant link is with emotional abuse. Two studies using nationally representative samples of older Americans (the mean ages were 50 and 56 years old, respectively) have found a link.

The first link is to an article that is paywalled except for its abstract. The abstract shows  the study does not involve a nationally representative sample of adults. The study compared patients with tension headaches to patients with migraines, without a no-headache control group. There is thus no opportunity to examine whether persons with migraines recall more emotional abuse than persons who do not suffer headaches.  Any significant associations in a huge sample disappeared after controlling for self-reported depression and anxiety.

My interpretation: There is nothing robust here. Results could be due to crude measurement, confounding of retrospective self-report by current self-report anxious or depressive symptoms. We can’t say much without a no-headache control group.

The second of the authors’ studies is also paywalled, but we can see from the abstract:

We used data from the Adverse Childhood Experiences (ACE) study, which included 17,337 adult members of the Kaiser Health Plan in San Diego, CA who were undergoing a comprehensive preventive medical evaluation. The study assessed 8 ACEs including abuse (emotional, physical, sexual), witnessing domestic violence, growing up with mentally ill, substance abusing, or criminal household members, and parental separation or divorce. Our measure of headaches came from the medical review of systems using the question: “Are you troubled by frequent headaches?” We used the number of ACEs (ACE score) as a measure of cumulative childhood stress and hypothesized a “dose–response” relationship of the ACE score to the prevalence and risk of frequent headaches.

Results — Each of the ACEs was associated with an increased prevalence and risk of frequent headaches. As the ACE score increased the prevalence and risk of frequent headaches increased in a “dose–response” fashion. The risk of frequent headaches increased more than 2-fold (odds ratio 2.1, 95% confidence interval 1.8-2.4) in persons with an ACE score ≥5, compared to persons with and ACE score of 0. The dose–response relationship of the ACE score to frequent headaches was seen for both men and women.

The Conversation authors misrepresent this study. It is about self-reported headaches, not the subgroup of these patients reporting migraines. But in the first of their own studies they just cited, the authors contrast tension headaches with migraine headaches, with no controls.

So the data did not allow examination of the association between adult retrospective reports of childhood emotional abuse and migraines. There is no mention of self-reported depression and anxiety, which wiped out any relationship with childhood adversity in headaches in the first study. I would expect that a survey of ACES would include such self-report. And the ACEs equate either parental divorce and separation (the same common situation likely occur together and so are counted twice) with sexual abuse in calculating an overall score.

The authors make a big deal of the “dose-response” they found. But this dose-response could just represent uncontrolled confounding  – the more ACEs indicates the more confounding, greater likelihood that respondents faced other social, person, economic, and neighborhood deprivations.  The higher the ACE score, the greater likelihood that other background characteristic s are coming into play.

The only other evidence the authors cite is again another one of their papers, available only as a conference abstract. But the abstract states:

Results: About 14.2% (n = 2,061) of the sample reported a migraine diagnosis. Childhood abuse was recalled by 60.5% (n =1,246) of the migraine sample and 49% (n = 6,088) of the non-migraine sample. Childhood abuse increased the chances of a migraine diagnosis by 55% (OR: 1.55; 95% CI 1.35 – 1.77). Of the three types of abuse, emotional abuse had a stronger effect on migraine (OR: 1.52; 95% CI 1.34 – 1.73) when compared to physical and sexual abuse. When controlled for depression and anxiety, the effect of childhood abuse on migraine (OR: 1.32; 95% CI 1.15 – 1.51) attenuated but remained significant. Similarly, the effect of emotional abuse on migraine decreased but remained significant (OR: 1.33; 95% CI 1.16 – 1.52), when controlled for depression and anxiety.

The rates of childhood abuse seem curiously high for both the migraine and non-migraine samples. If you dig a bit on the web for details of the National Longitudinal Study of Adolescent Health, you can find how crude the measurement is.  The broad question assessing emotional abuse covers the full range of normal to abnormal situations without distinguishing among them.

How often did a parent or other adult caregiver say things that really hurt your feelings or made you feel like you were not wanted or loved? How old were you the first time this happened? (Emotional abuse).

An odds ratio of 1.33 is not going to attract much attention from an epidemiologist, particularly when it is obtained from such messy data.

I conclude that the authors have made only a weak case for the following statement: While all forms of childhood maltreatment have been shown to be linked to migraines, the strongest and most significant link is with emotional abuse.

Oddly, if we jump ahead to the closing section of The Conversation article, the authors concede:

Childhood maltreatment probably contributes to only a small portion of the number of people with migraine.

But, as we will  see, they make recommendations that assume a strong link has been established.

Why would emotional abuse in childhood lead to migraines in adulthood?

This section throws out a number of trending buzz terms, strings them together in a way that should impress and intimidate consumers, rather than allow them independent evaluation of what is being said.

got everything

The section also comes below a stock blue picture of the brain.  In web searches, the picture  is associated with social media where the brain is superficially brought into  in discussions where neuroscience is  not relevant.

An Australian neuroscientist commented on Facebook:

Deborah on blowing brains

The section starts out:

The fact that the risk goes up in response to increased exposure is what indicates that abuse may cause biological changes that can lead to migraine later in life. While the exact mechanism between migraine and childhood maltreatment is not yet established, research has deepened our understanding of what might be going on in the body and brain.

We could lost in a quagmire trying to figuring out the evidence for the loose associations that are packed into a five paragraph section.  Instead,  I’ll make some observations that can be followed up by interested readers.

The authors acknowledge that no mechanism has been established linking migraines and child maltreatment. The link for this statement takes the reader to the authors own pay walled article that is explicitly labeled “Opinion Statement ”.

The authors ignore a huge literature that acknowledges great heterogeneity among sufferers of migraines, but points to some rather strong evidence for treatments based on particular mechanisms identified among carefully selected patients. For instance, a paper published in The New England Journal of Medicine with well over 1500 citations:

Goadsby PJ, Lipton RB, Ferrari MD. Migraine—current understanding and treatment. New England Journal of Medicine. 2002 Jan 24;346(4):257-70.

Speculations concerning the connections between childhood adversity, migraines and the HPA axis are loose. The Conversation authors their obviousness needs to be better document with evidence.

For instance, if we try to link “childhood adversity” to the HPA axis, we need to consider the lack of specificity of” childhood adversity’ as defined by retrospective endorsement of Adverse Childhood Experiences (ACEs). Suppose we rely on individual checklist items or cumulative scores based on number of endorsements. We can’t be sure that we are dealing with actual rather than assumed exposure to traumatic events or that there be any consistent correlates in current measures derived from the HPA axis.

Any non-biological factor defined so vaguely is not going to be a candidate for mapping into causal processes or biological measurements.

An excellent recent Mind the Brain article by my colleague blogger Shaili Jain interviews Dr. Rachel Yehuda, who had a key role in researching HPA axis in stress. Dr. Yehuda says endocrinologists would cringe at the kind of misrepresentations that are being made in The Conversation article.

A recent systematic review concludes the evidence for specific links between child treatment and inflammatory markers is of limited and poor quality.

Coelho R, Viola TW, Walss‐Bass C, Brietzke E, Grassi‐Oliveira R. Childhood maltreatment and inflammatory markers: a systematic review. Acta Psychiatrica Scandinavica. 2014 Mar 1;129(3):180-92.

The Conversation article misrepresents gross inconsistencies in the evidence of biological correlates representing biomarkers. There are as yet no biomarkers for migraines in the sense of a biological measurement that reliably distinguishes persons with migraines from other patient populations with whom they may be confused. See an excellent funny blog post by Hilda Bastian.

Notice the rhetorical trick in authors of The Conversation article’s assertion that

Migraine is considered to be a hereditary condition. But, except in a small minority of cases, the genes responsible have not been identified.

Genetic denialists like Oliver James  or Richard Bentall commonly phrased questions in this manner to be a matter of hereditary versus non-hereditary. But complex traits like height, intelligence, or migraines involve combinations of variations in a number of genes, not a single gene or even a few genes.. For an example of the kind of insights that sophisticated genetic studies of migraines are yielding see:

Yang Y, Ligthart L, Terwindt GM, Boomsma DI, Rodriguez-Acevedo AJ, Nyholt DR. Genetic epidemiology of migraine and depression. Cephalalgia. 2016 Mar 9:0333102416638520.

The Conversation article ends with some signature nonsense speculation about epigenetics:

However, stress early in life induces alterations in gene expression without altering the DNA sequence. These are called epigenetic changes, and they are long-lasting and may even be passed on to offspring.

Interested readers can find these claims demolished in Epigenetic Ain’t Magic by PZ Myers, a biologist who attempts to rescue an extremely important development concept from its misuse.

Or Carl Zimmer’s Growing Pains for Field of Epigenetics as Some Call for Overhaul.

What does this mean for doctors treating migraine patients?

The Conversation authors startle readers with an acknowledgment that contradicts what they have been saying earlier in their article:

Childhood maltreatment probably contributes to only a small portion of the number of people with migraine.

It is therefore puzzling when they next say:

But because research indicates that there is a strong link between the two, clinicians may want to bear that in mind when evaluating patients.

Cognitive behavior therapy is misrepresented as an established effective treatment for migraines. A recent systematic review and meta-analysis  had to combine migraines with other chronic headaches and order to get ten studies to consider.

The conclusion of this meta-analysis:

Methodology inadequacies in the evidence base make it difficult to draw any meaningful conclusions or to make any recommendations.

The Conversation article notes that the FDA has approved anti-epileptic drugs such as valproate and topiramate for treatment of migraines. However, the article’s claim that the efficacy of these drugs are due to their effects on epigenetics is quite inconsistent with what is said in the larger literature.

Clinicians specializing and treating fibromyalgia or irritable bowel syndrome would be troubled by the authors’ lumping these conditions with migraines and suggesting that a psychiatric consultation is the most appropriate referral for patients who are having difficulty achieving satisfactory management.

See for instance the links contained in my blog post, No, irritable bowel syndrome is not all in your head.

The Conversation article closes with:

Within a migraine clinic population, clinicians should pay special attention to those who have been subjected to maltreatment in childhood, as they are at increased risk of being victims of domestic abuse and intimate partner violence as adults.

That’s why clinicians should screen migraine patients, and particularly women, for current abuse.

It’s difficult to how this recommendation is relevant to what has preceded it. Routine screening is not evidence-based.

The authors should know that the World Health Organization formerly recommended screening primary care women for intimate abuse but withdrew the recommendation because of a lack of evidence that it improved outcomes for women facing abuse and a lack of evidence that no harm was being done.

I am sharing this blog post with the authors of The Conversation article. I am requesting a correction from The Conversation. Let’s see what they have to say.

Meanwhile, patients seeking health information are advised to avoid The Conversation.

Is risk of Alzheimer’s Disease reduced by taking a more positive attitude toward aging?

Unwarranted claims that “modifiable” negative beliefs cause Alzheimer’s disease lead to blaming persons who develop Alzheimer’s disease for not having been more positive.

Lesson: A source’s impressive credentials are no substitute for independent critical appraisal of what sounds like junk science and is.

More lessons on how to protect yourself from dodgy claims in press releases of prestigious universities promoting their research.

If you judge the credibility of health-related information based on the credentials of the source, this article  is a clear winner:

Levy BR, Ferrucci L, Zonderman AB, Slade MD, Troncoso J, Resnick SM. A Culture–Brain Link: Negative Age Stereotypes Predict Alzheimer’s Disease Biomarkers. Psychology and Aging. Dec 7 , 2015, No Pagination Specified. http://dx.doi.org/10.1037/pag0000062

alzheimers
From INI

As noted in the press release from Yale University, two of the authors are from Yale School of Medicine, another is a neurologist at Johns Hopkins School of Medicine, and the remaining three authors are from the US National Institute on Aging (NIA), including NIA’s Scientific Director.

The press release Negative beliefs about aging predict Alzheimer’s disease in Yale-led study declared:

“Newly published research led by the Yale School of Public Health demonstrates that                   individuals who hold negative beliefs about aging are more likely to have brain changes associated with Alzheimer’s disease.

“The study suggests that combatting negative beliefs about aging, such as elderly people are decrepit, could potentially offer a way to reduce the rapidly rising rate of Alzheimer’s disease, a devastating neurodegenerative disorder that causes dementia in more than 5 million Americans.

The press release posited a novel mechanism:

“We believe it is the stress generated by the negative beliefs about aging that individuals sometimes internalize from society that can result in pathological brain changes,” said Levy. “Although the findings are concerning, it is encouraging to realize that these negative beliefs about aging can be mitigated and positive beliefs about aging can be reinforced, so that the adverse impact is not inevitable.”

A Google search reveals over 40 stories about the study in the media. Provocative titles of the media coverage suggest a children’s game of telephone or Chinese whispers in which distortions accumulate with each retelling.

Negative beliefs about aging tied to Alzheimer’s (Waltonian)

Distain for the elderly could increase your risk of Alzheimer’s (FinancialSpots)

Lack of respect for elderly may be fueling Alzheimer’s epidemic (Telegraph)

Negative thoughts speed up onset of Alzheimer’s disease (Tech Times)

Karma bites back: Hating on the elderly may put you at risk of Alzheimer’s (LA Times)

How you feel about your grandfather may affect your brain health later in life (Men’s Health News)

Young people pessimistic about aging more likely to develop Alzheimer’s later on (Health.com)

Looking forward to old age can save you from Alzheimer’s (Canonplace News)

If you don’t like old people, you are at higher risk of Alzheimer’s, study says (RedOrbit)

If you think elderly people are icky, you’re more likely to get Alzheimer’s (HealthLine)

In defense of the authors of this article as well as journalists, it is likely that editors added the provocative titles without obtaining approval of the authors or even the journalists writing the articles. So, let’s suspend judgment and write off sometimes absurd titles to editors’ need to establish they are offering distinctive coverage, when they are not necessarily doing so. That’s a lesson for the future: if we’re going to criticize media coverage, better focus on the content of the coverage, not the titles.

However, a number of these stories have direct quotes from the study’s first author. Unless the media coverage is misattributing direct quotes to her, she must have been making herself available to the media.

Was the article such an important breakthrough offering new ways in which consumers could take control of their risk of Alzheimer’s by changing beliefs about aging?

No, not at all. In the following analysis, I’ll show that judging the credibility of claims based on the credentials of the sources can be seriously misleading.

What is troubling about this article and its well-organized publicity effort is that information is being disseminated that is misleading and potentially harmful, with the prestige of Yale and NIA attached.

Before we go any further, you can take your own look at a copy of the article in the American Psychological Association journal Psychology and Aging here, the Yale University press release here, and a fascinating post-publication peer review at PubPeer that I initiated as peer 1.

Ask yourself: if you encountered coverage of this article in the media, would you have been skeptical? If so what were the clues?

spoiler aheadcure within The article is yet another example of trusted authorities exploiting entrenched cultural beliefs about the mind-body connection being able to be harnessed in some mysterious way to combat or prevent physical illness. As Ann Harrington details in her wonderful book, The Cure Within, this psychosomatic hypothesis has a long and checkered history, and gets continually reinvented and misapplied.

We see an example of this in claims that attitude can conquer cancer. What’s the harm of such illusions? If people can be led to believe they have such control, they are set up for blame from themselves and from those around them when they fail to fend off and control the outcome of disease by sheer mental power.

The myth of “fighting spirit” overcoming cancer that has survived despite the accumulation of excellent contradictory evidence. Cancer patients are vulnerable to blaming themselves for being blamed by loved ones when they do not “win” the fight against cancer. They are also subject to unfair exhortations to fight harder as their health situation deteriorates.

onion composite
                                                        From the satirical Onion

 What I saw when I skimmed the press release and the article

  • The first alarm went off when I saw that causal claims were being made from a modest sized correlational study. This should set off anyone’s alarms.
  • The press release refers to this as a “first ever” d discussion section of the article refer to this as a “first ever” study. One does not seek nor expect to find robust “first ever” discoveries in such a small data set.
  • The authors do not provide evidence that their key measure of “negative stereotypes” is a valid measure of either stereotyping or likelihood of experiencing stress. They don’t even show it is related to concurrent reports of stress.
  • Like a lot of measures with a negative tone to items, this one is affected by what Paul Meehl calls the crud factor. Whatever is being measured in this study cannot be distinguished from a full range of confounds that not even being assessed in this study.
  • The mechanism by which effects of this self-report measure somehow get manifested in changes in the brain lacks evidence and is highly dubious.
  • There was no presentation of actual data or basic statistics. Instead, there were only multivariate statistics that require at least some access to basic statistics for independent evaluation.
  • The authors resorted to cheap statistical strategies to fool readers with their confirmation bias: reliance on one tailed rather than two-tailed tests of significance; use of a discredited backwards elimination method for choosing control variables; and exploring too many control/covariate variables, given their modest sample size.
  • The analyses that are reported do not accurately depict what is in the data set, nor generalize to other data sets.

The article

The authors develop their case that stress is a significant cause of Alzheimer’s disease with reference to some largely irrelevant studies by others, but depend on a preponderance of studies that they themselves have done with the same dubious small samples and dubious statistical techniques. Whether you do a casual search with Google scholar or a more systematic review of the literature, you won’t find stress processes of the kind the authors invoke among the usual explanations of the development of the disease.

Basically, the authors are arguing that if you hold views of aging like “Old people are absent-minded” or “Old people cannot concentrate well,” you will experience more stress as you age, and this will accelerate development of Alzheimer’s disease. They then go on to argue that because these attitudes are modifiable, you can take control of your risk for Alzheimer’s by adopting a more positive view of aging and aging people

The authors used their measure of negative aging stereotypes in other studies, but do not provide the usual evidence of convergent  and discriminant validity needed to establish the measure assesses what is intended. Basically, we should expect authors to show that a measure that they have developed is related to existing measures (convergent validity) in ways that one would expect, but not related to existing measures (discriminate validity) with which it should have associations.

Psychology has a long history of researchers claiming that their “new” self-report measures containing negatively toned items assess distinct concepts, despite high correlations with other measures of negative emotion as well as lots of confounds. I poked fun at this unproductive tradition in a presentation, Negative emotions and health: why do we keep stalking bears, when we only find scat in the woods?

The article reported two studies. The first tested whether participants holding more negative age stereotypes would have significantly greater loss of hippocampal volume over time. The study involved 52 individuals selected from a larger cohort enrolled in the brain-neuroimaging program of the Baltimore Longitudinal Study of Aging.

Readers are given none of the basic statistics that would be needed to interpret the complex multivariate analyses. Ideally, we would be given an opportunity to see how the independent variable, negative age stereotypes, is related to other data available on the subjects, and so we could get some sense if we are starting with some basic, meaningful associations.

Instead the authors present the association between negative age stereotyping and hippocampal volume only in the presence of multiple control variables:

Covariates consisted of demographics (i.e., age, sex, and education) and health at time of baseline-age-stereotype assessment, (number of chronic conditions on the basis of medical records; well-being as measured by a subset of the Chicago Attitude Inventory); self-rated health, neuroticism, and cognitive performance, measured by the Benton Visual Retention Test (BVRT; Benton, 1974).

Readers get cannot tell why these variables and not others were chosen. Adding or dropping a few variables could produce radically different results. But there are just too many variables being considered. With only 52 research participants, spurious findings that do not generalize to other samples are highly likely.

I was astonished when the authors announced that they were relying on one-tailed statistical tests. This is widely condemned as unnecessary and misleading.

Basically, every time the authors report a significance level in this article, you need to double the number to get what is obtained with a more conventional two-tailed test. So, if they proudly declare that results are significant p = .046, then the results are actually (non)significant, p= .092. I know, we should not make such a fuss about significance levels, but journals do. We’re being set up to be persuaded the results are significant, when they are not by conventional standards.

So the authors’ accumulating sins against proper statistical techniques and transparent reporting: no presentation of basic associations; reporting one tailed tests; use of multivariate statistics inappropriate for a sample that is so small. Now let’s add another one, in their multivariate regressions, the authors relied on a potentially deceptive backwards elimination:

Backward elimination, which involves starting with all candidate variables, testing the deletion of each variable using a chosen model comparison criterion, deleting the variable (if any) that improves the model the most by being deleted, and repeating this process until no further improvement is possible.

The authors assembled their candidate control/covariate variables and used a procedure that checks them statistically and drop some from consideration, based on whether they fail to add to the significance of the overall equation. This procedure is condemned because the variables that are retained in the equation capitalize on chance. Particular variables that could be theoretically relevant are eliminated simply because they fail to add anything statistically in the context of the other variables being considered. In the context of other variables, these same discarded variables would have been retained.

The final regression equation had fewer control/covariates then when the authors started. Statistical significance will be calculated on the basis of the small number of variables remaining, not the number that were picked over and so results will artificially appear stronger. Again, potentially quite misleading to the unwary reader.

The authors nonetheless concluded:

As predicted, participants holding more-negative age stereotypes, compared to those holding more-positive age stereotypes, had a significantly steeper decline in hippocampal volume

The second study:

examined whether participants holding more negative age stereotypes would have significantly greater accumulation of amyloid plaques and neurofibrillary tangles.

The outcome was a composite-plaques-and-tangles score and the predictor was the same negative age stereotypes measure from the first study. These measurements were obtained from 74 research participants upon death and autopsy. The same covariates were used in stepwise regression with backward elimination. Once again, the statistical test was one tailed.

Results were:

As predicted, participants holding more-negative age stereotypes, compared to those holding more-positive age stereotypes, had significantly higher composite-plaques-and-tangles scores, t(1,59) = 1.71 p = .046, d = 0.45, adjusting for age, sex, education, self-rated health, well-being, and number of chronic conditions.

Aha! Now we see why the authors commit themselves to a one tailed test. With a conventional two-tailed test, these results would not be significant. Given a prevailing confirmation bias, aversion to null findings, and obsession with significance levels, this article probably would not have been published without the one tailed test.

The authors’ stirring overall conclusion from the two studies:

By expanding the boundaries of known environmental influences on amyloid plaques, neurofibrillary tangles, and hippocampal volume, our results suggest a new pathway to identifying mechanisms and potential interventions related to Alzheimer’s disease

pubpeerPubPeer discussion of this paper [https://pubpeer.com/publications/16E68DE9879757585EDD8719338DCD ]

Comments accumulated for a couple of days on PubPeer after I posted some concerns about the first study. All of the comments were quite smart, some directly validated points that I been thinking about, but others took the discussion in new directions either statistically or because the commentators knew more about neuroscience.

Using a mechanism available at PubPeer, I sent emails to the first author of the paper, the statistician, and one of the NIA personnel inviting them to make comments also. None have responded so far.

Tom Johnstone, a commentator who exercise the option of identifying himself noted the reliance on inferential statistics in the absence of reporting basic relationships. He also noted that the criterion used to drop covariates was lax. Apparently familiar with neuroscience, he expressed doubts that the results had any clinical significance or relevance to the functioning of the research participants.

Another commentator complained of the small sample size, use of one tailed statistical tests without justification, the “convoluted list of covariates,” and “taboo” strategy for selecting covariates to be retained in the regression equation. This commentator also noted that the authors had examined the effect of outliers, conducting analyses both with and without the inclusion of the most extreme case. While it didn’t affect the overall results, exclusion dramatically change the significance level, highlighting the susceptibility of such a small sample to chance variation or sampling error.

Who gets the blame for misleading claims in this article?

dr-luigi-ferrucciThere’s a lot of blame to go around. By exaggerating the size and significance of any effects, the first author increases the chance of publication and also further funding to pursue what is seen as a “tantalizing” association. But it’s the job of editors and peer reviewers to protect the readership from such exaggerations and maybe to protect the author from herself. They failed, maybe because exaggerated findings are consistent with the journal‘s agenda of increasing citations by publishing newsworthy rather than trustworthy findings. The study statistician, Martin Slade obviously knew that misleading, less than optimal statistics were used, why didn’t he object? Finally, I think the NIA staff, particularly Luigi Ferrucci, the Scientific Director of NIA  should be singled out for the irresponsibility of attaching their names to such misleading claims. Why they do so? Did they not read the manuscript?  I will regularly present instances of NIH staff endorsing dubious claims, such as here. The mind-over-disease, psychosomatic hypothesis, gets a lot of support not warranted by the evidence. Perhaps NIH officials in general see this as a way of attracting research monies from Congress. Regardless, I think NIH officials have the responsibility to see that consumers are not misled by junk science.

This article at least provided the opportunity for an exercise that should raise skepticism and convince consumers at all levels – other researchers, clinicians, policymakers, and those who suffer from Alzheimer’s disease and those who care from them – we just cannot sit back and let trusted sources do our thinking for us.

 

What we can learn from a PLOS Medicine study of antidepressants and violent crime

Update October 1 7:58 PM: I corrected an inaccuracy in response to a comment by DJ Jaffe, for which I am thankful.

An impressively large-scale study published in PLOS Medicine of the association between antidepressants and violent crime is being greeted with strong opinions from those who haven’t read it. But even those who attempt to read the article might miss some of the nuances and the ambiguity that its results provide.

2305701220_0fc3d01183_bIn this issue of Mind the Brain, we will explore some of these nuances, which are fascinating in themselves. But the article also provides excellent opportunities to apply the critical appraisal skills needed for correlational observational studies using administrative data sets.

Any time there is a report of a mass shooting in the media, a motley crew of commentators immediately announces that the shooter is mentally ill and has been taking psychotropic medication. Mental illness and drugs are the problem, not guns, we are told. Sprinkled among the commentators are opponents of gun-control, Scientologists, and psychiatrists seeking to make money serving as expert witnesses. They are paid handsomely to argue for the diminished responsibility for the shooter or for product liability suits against Pharma. Rebuttals will be offered by often equally biased commentators, some of them receiving funds from Pharma.

every major shoorting
This is not from the Onion, but a comment left at a blog that expresses a commonly held view.

guns-health-care-82880109353

What is generally lost is that most shooters are not mentally ill and are not taking psychotropic medication.

Yet such recurring stories in the media have created a strong impression in the public and even professionals that a large scientific literature exists which establishes a tie between antidepressant use and violence.

Even when there has been some exposure to psychotropic medication, its causal role in the shooting cannot be established either from the facts of the case or the scientific literature.

The existing literature is seriously limited in quality and quantity and contradictory in its conclusions. Ecological studies [ 1, 2,]  conclude that the availability of antidepressants may reduce violence on a community level. An “expert review” and a review of reports of adverse events conclude there is a link between antidepressants and violence. However, reports of adverse events being submitted to regulatory agencies can be strongly biased, including by recent claims in the media. Reviews of adverse events do not distinguish between correlates of a condition like depression and effects of the drug being used to treat it. Moreover, authors of these particular reviews were serving as expert witnesses in legal proceedings. Authorship adds to their credibility and publicizes their services.

The recent study in PLOS Medicine should command the attention of anyone interested in the link between antidepressants and violent crime. Already there have been many tweets and at least one media story claiming vindication of the Scientologists as being right all along  I expected the release of the study and its reaction in the media would give me another opportunity to call attention to the entrenched opposing sides in the antidepressant wars  who only claim to be driven by strength of evidence and dismiss any evidence contrary to their beliefs, as well as the gullibility of journalists. But the article and its coverage in the media are developing a very different story.

At the outset, I should say I don’t know if evidence can be assembled for an unambiguous case that antidepressants are strongly linked to violent crime. Give up on us ever been able to rely on a randomized trial in which we examine whether participants randomized to receiving an antidepressant rather than a placebo are convicted more often for violent crimes. Most persons receiving antidepressant will not be convicted for a violent crime. The overall base rate of convictions is too low to monitor as an outcome a randomized trial. We are left having to sort through correlational observational, clinical epidemiological data typically collected for other purposes.

I’m skeptical about there being a link strong enough to send a clear signal through all the noise in the data sets that we can assemble to look for it. But the PLOS Medicine article represents a step forward.

stop Association does not equal causation
From Health News Review

Correlation does not equal causality.

Any conceivable data set in which we can search will pose the challenges of competing explanations from other variables that might explain the association.

  • Most obviously, persons prescribed antidepressants suffer from conditions that may themselves increase the likelihood of violence.
  • The timing of persons seeking treatment with antidepressants may be influenced by circumstances that increase their likelihood of violence.
  • Violent persons are more likely to be under the influence of alcohol and other drugs and to have histories of use of these substances.
  • Persons taking antidepressants and consuming alcohol and other drugs may be prone to adverse effects of the combination.
  • Violent persons have characteristics and may be in circumstances with a host of other influences that may explain their behavior.
  • Violent persons may themselves be facing victimization that increases the likelihood of their committing violence and having a condition warranting treatment with antidepressants.

Etc, etc.

The PLOS Medicine article introduces a number of other interesting possibilities for such confounding.

Statistical controls are never perfect

Studies will always incompletely specify of confounds and imperfectly measure them. Keep in mind that completeness of statistical control requires that all possible confounding factors be identified and measured without error. These ideal conditions are not attainable. Yet any application of statistics to “control” confounds that do not meet these ideal conditions risks producing less accurate estimate of effects than simply examining basic associations. Yet, we already know that these simple associations are not sufficient to indicate causality.

The PLOS Medicine article doesn’t provide definitive answers, but it presents data with greater sophistication than has previously been available. The article’s careful writing should make misinterpretation or missing of its main points less likely. And one of the authors – Professor Seena Fazel of the Department of Psychiatry, Oxford University – did an exemplary job of delivering careful messages to any journalist who would listen.

Professor Seena Fazel
Professor Seena Fazel

Professor Fazel can be found explaining his study in the media at 8:45 in a downloadable BBC World New Health Check News mp3.

Delving into the details of the article

The PLOS Medicine article is of course open access and freely available.

Molero, Y., Lichtenstein, P., Zetterqvist, J., Gumpert, C. H., & Fazel, S. (2015). Selective Serotonin Reuptake Inhibitors and Violent Crime: A Cohort Study. PLoS Med, 12(9), e1001875.

Supplementary material are also available from the web [1, 2, 3] for the study including a completed standardized STROBE checklist of items  that should be included in reports of observational studies, additional tables, and details of the variables and how they were obtained.

An incredible sample

Out of Sweden’s total population of 7,917,854 aged 15 and older in 2006, the researchers identified 856,493 individuals who were prescribed a selective serotonin reuptake inhibitor (SSRI) antidepressant from 2006-2009 and compared them to the 7,061,361 Swedish individuals who were not been prescribed this medication in that four year period.

SSRIs  were chosen for study because they represent the bulk of antidepressants being prescribed and also because SSRIs are the class of antidepressants to which the question of an association with violence of the most often raised. Primary hypotheses were about the SSRIs as a group, but secondary analyses focused on individual SSRIs – fluoxetine, citalopram, paroxetine, sertraline, fluvoxamine, and escitalopram. It was not expected that the analyses at the level of individual SSRI drugs have sufficient statistical power to explore associations with violent crimes. Data were also collected on non-SSRI antidepressants and other psychotropic medication, and these data were used to adjust for medications taken concurrently with SSRIs.

With these individuals’ unique identification number, the researchers collected information on the particular medications and dates of prescription from the Swedish Prescribed Drug Register. The register provides complete data on all prescribed and dispensed medical drugs from all pharmacies in Sweden since July 2005. The unique identification number also allowed obtaining information concerning hospitalizations and outpatient visits and reasons for visit and diagnoses.

crime sceneThese data were then matched against information on convictions for violent crimes for the same period from the Swedish national crime register.

These individuals were followed from January 1, 2006, to December 31, 2009.

During this period 1% of individuals prescribed an SSRI were convicted of a violent crime versus .6% of those not being prescribed an SSRI. The article focused on the extent to which prescription of an SSRI affected the likelihood of committing a violent crime and considered other possibilities for any association that was found.

A clever analytic strategy

Epidemiologic studies most commonly make comparisons between individuals differing in their exposures to particular conditions in terms of whether they have particular outcomes. Detecting bona fide causal associations can be derailed by other characteristics being associated with both antidepressants and violent crimes. An example of a spurious relationship is one between coffee drinking and cardiovascular disease. Exposure to coffee may be associated with lung cancer, but the association is spurious, due to smokers smoking Confoundinglighting up when they have coffee breaks. Taking smoking into account eliminates the association of coffee and cardiovascular disease. In practice, it can be difficult to identify such confounds, particularly when they are left unmeasured or imperfectly measured.

So, such Between-individual analyses of people taking antidepressants and those who are not are subject to a full range of unmeasured, but potentially confounding background variables.

For instance, in an earlier study in the same population, some of these authors found that individuals with a full (adjusted OR 1.5, 95% CI 1.3-1.6) or half (adjusted OR 1.2, 95% CI 1.1-1.4) sibling with depression were themselves more likely to be convicted of violent crime, after controlling for age, sex, low family income and being born abroad. The influence of such familial risk can be misconstrued in a standard between-individual analysis.

This article supplemented between-individual analyses with within-individual stratified Cox proportional hazards regressions. Each individual exposed to antidepressants was considered separately and served as his/her own control. Thus, these within-individual analyses examined differences in violent crimes in the same individuals over time periods differing in whether they had exposure to an antidepressant prescription. Periods of exposure became the unit of analysis, not just individuals.

The linked Swedish data sets that were used are unusually rich. It would not be feasible to obtain such data in other countries, and certainly not the United States.

The results as summarized in the abstract

Using within-individual models, there was an overall association between SSRIs and violent crime convictions (hazard ratio [HR] = 1.19, 95% CI 1.08–1.32, p < 0.001, absolute risk = 1.0%). With age stratification, there was a significant association between SSRIs and violent crime convictions for individuals aged 15 to 24 y (HR = 1.43, 95% CI 1.19–1.73, p < 0.001, absolute risk = 3.0%). However, there were no significant associations in those aged 25–34 y (HR = 1.20, 95% CI 0.95–1.52, p = 0.125, absolute risk = 1.6%), in those aged 35–44 y (HR = 1.06, 95% CI 0.83–1.35, p = 0.666, absolute risk = 1.2%), or in those aged 45 y or older (HR = 1.07, 95% CI 0.84–1.35, p = 0.594, absolute risk = 0.3%). Associations in those aged 15 to 24 y were also found for violent crime arrests with preliminary investigations (HR = 1.28, 95% CI 1.16–1.41, p < 0.001), non-violent crime convictions (HR = 1.22, 95% CI 1.10–1.34, p < 0.001), non-violent crime arrests (HR = 1.13, 95% CI 1.07–1.20, p < 0.001), non-fatal injuries from accidents (HR = 1.29, 95% CI 1.22–1.36, p < 0.001), and emergency inpatient or outpatient treatment for alcohol intoxication or misuse (HR = 1.98, 95% CI 1.76–2.21, p < 0.001). With age and sex stratification, there was a significant association between SSRIs and violent crime convictions for males aged 15 to 24 y (HR = 1.40, 95% CI 1.13–1.73, p = 0.002) and females aged 15 to 24 y (HR = 1.75, 95% CI 1.08–2.84, p = 0.023). However, there were no significant associations in those aged 25 y or older. One important limitation is that we were unable to fully account for time-varying factors.

Hazard ratios (HRs) are explained here and are not to be confused with odds ratios (ORs) explained here. Absolute risk (AR) is the most intuitive and easy to understand measure of risk and is explained here, along with reasons that hazard ratios don’t tell you anything about absolute risk.

Principal findings

  • There was an association between receiving a prescription for antidepressants and violent crime.
  • When age differences were examined, the 15-24 age range was the only one from which the association was significant.
  • No association was found for other age groups.
  • The association held for both males and females analyze separately in the 15 – 24 age range. But…

Things not to be missed in the details

Only a small minority of persons prescribed an antidepressant were convicted of a violent crime, but the likelihood of a conviction in persons exposed to antidepressants was increased in this 15 to 24 age range.

There isn’t a dose-response association between SSRI use and convictions for violent crimes. Even in the 15 to 24 age range, periods of moderate or high exposure to SSRIs were not associated with violent crimes any more than no exposure. Rather, the association occurred only in those individuals with low exposure.

A dose response association would be reflected in the more exposure to antidepressants an individual had, the greater the level of violent crimes. A dose response association is a formal criterion for a causal association adequate evidence of a causal relationship between an incidence and a possible consequence.

In the age bracket for which this association between antidepressant use and conviction of a violent crime was significant, antidepressant use was also associated with an increased risk of violent crime arrests, non-violent crime convictions, and non-violent crime arrests, using emergency inpatient and or outpatient treatment for alcohol intoxication or misuse.

Major caveats

The use of linked administrative data sets concerning both antidepressant prescription and violent crimes is a special strength of this study. It allows a nuanced look at an important question with evidence that could not otherwise be assembled. But administrative data have well-known limitations.

The data were not originally captured with the research questions in mind and so key variables, including data concerning potential confounds were not necessarily collected. The quality control for the administrative purposes for which these data were collected, may differ greatly from what is needed in their use as research data. There may be systematic errors and incomplete data and inaccurate coding, including of the timing of these administrative events.

Administrative data do not always mesh well with the concepts with which we may be most concerned. This study does not directly assess violent behavior, only arrest and convictions. Most violent behavior does not result in an arrest or conviction and so this is a biased proxy for behavior.

This study does not directly assess diagnosis of depression, only diagnosis by specialists. We know from other studies that in primary and specialty medical settings, there may be no systematic effort to assess clinical depression by interview. The diagnoses that are recorded may simply be only serve to justify a clinical decision made on the basis other than a patient meeting research criteria for depression. Table 1 in the article suggests that only about a quarter of the patients exposed to antidepressants actually had a diagnosis of depression. And throughout this article, there was no distinction made between unipolar depression and the depressed phase of a bipolar disorder. This distinction may be important, given the small minority of individuals who were convicted of a violent crime while exposed to a SSRI.

Alcohol-and-Anti-DepressantsPerhaps one of the greatest weaknesses of this data set is its limited assessment of alcohol and substance use and abuse. For alcohol, we are limited to emergency inpatient or outpatient treatment for alcohol intoxication or misuse. For substance abuse, we have only convictions designated as substance-related. These are poor proxies for more common actual alcohol and substance use, which for a variety of reasons may not show up in these administrative data. Substance-related convictions are simply too infrequent to serve as a suitable control variable or even proxy for substance. It is telling that in the 15-24 age range, alcohol intoxication or misuse is associated with convictions for violent crimes with a strength (HR = 1.98, 95% CI 1.76–2.21, p < 0.001) greater than that found for SSRIs.

There may be important cultural differences between Sweden and other countries to which we want to generalize in terms of the determinants of arrest and conviction, but also treatment seeking for depression and the pathways for obtaining antidepressant medication. There may also be differences in institutional response to drug and alcohol use and misuse, including individuals’ willingness and ability to access services.

An unusual strength of this study is its use of within-individual analyses to escape some of the problems of more typical between-individual analyses not being able to adequately control for stable sources of differences. But, we can’t rely on these analyses to faithfully capture crucial sequences of events that happen quickly in terms of which events occurred first. The authors note that they

cannot fully account for time-varying risk factors, such as increased drug or alcohol use during periods of SSRI medication, worsening of symptoms, or a general psychosocial decline.

Findings examining non-fatal injuries from accidents as well as emergency inpatient or outpatient treatment for alcohol intoxication or misuse as time-varying confounders are tantalizing, but we reached the limits of the administrative data in trying to pursue them.

What can we learn from this study?

Readers seeking a definitive answer from the study to the question of whether antidepressants cause violent behavior or even violent crime will be frustrated.

There does not seem to be a risk of violent crime in individuals over 25 taking antidepressants.

The risk confined to individuals aged between 15 and 25 is, according to the authors, modest, but not insignificant. It represents a 20 to 40% increase in the low likelihood of being convicted of a violent crime. But it is not necessarily causal. The provocative data suggesting that low exposure, rather than no exposure or moderate or high exposure to antidepressants should give pause and suggest something more complex than simple causality may be going on.

This is an ambiguous but important point. Low exposure could represent non-adherence, inconsistent adherence, or periods in which there was a sudden stopping of medication, the effects of which might generate an association between the exposure and violent crimes. It could also represent the influence of time-dependent variables such as use of alcohol or substances that escaped control in the within-individual analyses.

There are parallels between results of the present study what is observed in other data sets. Most importantly, the data have some consistency with reports of suicidal ideation and deliberate self-harm among children and adolescents exposed to antidepressants. The common factor may be increased sensitivity of younger persons to antidepressants and particularly to their initiation and withdrawal or sudden stopping, the sensitivity reflected in impulsive and risk-taking behavior.

The take away message

Data concerning links between SSRIs and violent crime invite premature and exaggerated declarations of implications for public health and public policy.

At another blog, I’ve suggested that the British Medical Journal requirement that that observational studies have a demarcated section addressing these issues encourages authors to go beyond their data in order to increase the likelihood of publication – authors have to make public health and public policy recommendations to show that their data are newsworthy enough for publication. It’s interesting thata media watch group  criticized BMJ for using too strong causal language in covering this observational PLOS Medicine article.

I’m sure that the authors of this article felt pressure to address whether a black box warning inserted into the packaging of SSRIs was warranted by these data. I agree with them not recommending this at this time because of the strength of evidence and ambiguity in the interpretation of these administrative data. But I agree that the issue of young people being prescribed SSRIs needs more research and specifically elucidation of why low dose increases the likelihood of violence versus no or medium to high dose.

The authors do make some clinical recommendations, and their spokesperson Professor Fazel is particularly clear but careful in his interview with BBC World New Health Check News. My summary of what is said in the interview and in other media contacts is

  • Adolescents and young adults should be prescribed SSRIs should be on the basis of careful clinical interviews to ascertain a diagnosis consistent with practice guidelines for prescribing these drugs and that the drug be prescribed at therapeutic level.
  • These patients should be educated about the necessity of taking these medications consistently and advised against withdrawal or stopping the medication quickly without consultation and supervision of a professional.
  • These patients should be advised against taking these medications with alcohol or other drugs, with the explanation that there could be serious adverse reactions.

In general, young persons may be more sensitive to SSRIs, particularly when starting or stopping, and particularly when taken in the presence of alcohol or other drugs.

The importance of more research concerning nature of the sensitivity is highlighted by the findings of the PLOS Medicine article and the issues these findings point to but do not resolve.

Molero Y, Lichtenstein P, Zetterqvist J, Gumpert CH, Fazel S (2015) Selective Serotonin Reuptake Inhibitors and Violent Crime: A Cohort Study. PLoS Med 12(9): e1001875. doi:10.1371/journal.pmed.1001875

The views expressed in this post represent solely those of its author, and not necessarily those of PLOS or PLOS Medicine.