Almost everyone has a personal connection to the national overdose crisis that claimed over 107,000 lives last year. Many have also been touched by the rising toll of suicides in the U.S, which took over 47,000 lives in 2021. Given the pain of those losses, debates over causes and solutions are contentious, especially when they involve the real or speculative role played by prescribed opioids in suicides. In the quest for solutions, researchers and advocates sometimes make recommendations that are not supported by data.
Recently, an article in the American Journal of Psychiatry (AJP) that was widely covered in the lay press, suggested that reductions in the chronic use of prescribed opioids for patients with pain slowed an otherwise discouraging national 20-year rise in suicides. Further, the paper suggests its findings should alleviate concerns about dose reductions in pain patients who have relied on these medicines longterm. However, these conclusions were not supported by the data in the study or in any other available data.
Given rising national concern about a burgeoning opioid crisis, many doctors forcibly reduced doses in long-term recipients of prescribed opioids, often under pressure from regulators and boards. Unfortunately, suicides started happening among patients who were taken off their medication. A series of state- and national-database analyses have documented, retrospectively, elevated rates of mental health crises, suicidal actions or death by suicide among those persons whose doses were reduced, compared to persons not subject to reduction. Because the risk for suicide or suicide attempt remains elevated 1-2 years after the reduction, it has not seemed likely that these tragedies are entirely due to acute withdrawal from opioids. Various authorities have called for caution, as have clinician-researchers like me.
The new article in AJP, from investigators at Columbia University, pushes back. It reports that areas of the country with the biggest declines in opioid prescriptions partly bucked a national trend toward rising suicides.
A few details make it easier to think about what such regional studies can and cannot show. The study relies on suicide data from 2009 to 2017, across 882 “commuting zones” (as devised by the Bureau of Agriculture, areas typically bigger than a county). As mentioned: suicides rose nationally over the study years 2009-2017. Opioid prescriptions dropped from a 2012 peak to the present.
Statistical models tested whether suicide rates rose less in those regions with the greater prescription declines. Of course, regions differ. The study’s models tried to control for that by assigning a single statistical term for each region called a “fixed effect.” The assumption behind a “fixed effect” is the following: As long as the regions differ from each other in “fixed” ways that did not change from 2009-2017, then the models controlled for such differences
This is not a reasonable assumption. We know that regions change in many ways likely to contribute to reductions in opioid prescribing and suicide, without one causing the other. One town might enjoy some economic development, and influx of young families, and new doctors who prescribe less; a decline in suicide might result from all these good things, without the opioid prescriptions having much to do with it. Statistical and graphical analysis could have helped readers learn whether the model’s assumptions were acceptable; but they weren’t presented.
Still, this paper does show that regions with a greater decline in opioid prescriptions (compared to all others) had a smaller rise in total suicides and in opioid-related suicide rates (compared to all others). That regional pattern merits investigation. But there were also exceptions (this pattern wasn’t evident in adults 65 or older). And, for younger persons, unintentional deaths involving opioids seemed to rise most in regions with the greatest prescription declines.[1]
There would be nothing problematic about this article if all it did was present its data and invite further discussion. The article, however, proposed to address the matter of tapering long-term opioid recipients. The introduction cited two of the studies where suicide risk rose after reduction (including one I co-authored), and attempted to refute them -- an unusual approach for an introduction.
Then, the paper’s discussion argues that it is the patient who receives the prescription who is put at personal risk for suicide, citing a VA paper where opioid doses did correlate with suicide risk. But that discussion omits a much richer Australian study that looked for, and could not find, any association between opioid receipt, or opioid dose, and suicidality. Buried in the discussion, a caveat was offered, that the results “do not shed light on the clinical pathways connecting local opioid prescribing to individual opioid overdose suicide deaths.” However, this went unnoticed by many readers. (It was an academic article version of churnalist’s fifth sin: “disclaim and pivot.”)
As a physician-researcher, I care about whether studies are used in ways that misrepresent the risks of curtailing prescriptions in patients with pain. Today, I lead a federally-funded study to examine 110-120 suicides through interviews and record review. We seek to examine them in depth, the way crash-site investigators assess airplane crashes. That’s because we don’t see suicide as simple one-cause affairs.
However, simple stories appeal to reporters. The new paper ignited a storm of inaccurate press coverage, that was seen by many as endorsing the safety of forced opioid reductions, notwithstanding CDC and FDA declarations to the contrary. Speaking to US News and World Report, for example, an associate professor of surgery announced, “for those who have wondered whether curtailing opioid prescriptions could be associated with an increased risk of suicide, this study is reassuring.”
Within days, a widely-quoted, highly-paid expert witness for the plaintiffs in our nation’s ongoing opioid litigation tweeted that the new data “debunk” a “hoax that opioid reduction caused an ‘epidemic’ of suicides” (designation of a suicide “hoax” is not a one-time affair for this expert). Of course, no experts had claimed an “epidemic” of suicides. Rather experts and patients observed that tapering can increase the risk for suicide, and were appropriately concerned to avoid that.
Ultimately, reporters repeated the fallacy that regional data can tell us how to care for individual patients. But it was the way the paper was written -- most notably its introduction and discussion -- rather than its data, that drew that interpretation. I suspect that public mockery of suicide risk was not the outcome desired by the authors, given that the final lines of the paper urged caution with opioid dose changes. Speaking with senior author Dr. Mark Olfson confirmed that sense. He readily offered the following condemnation:
The results of our recent study indicate that regions of the country that experienced the greatest declines in opioid prescribing also tended to have the greatest declines in regional suicide rates. It would be a mistake, however, to assume that this ecological observation informs the daily clinical management of individuals receiving opioids for chronic pain or refutes clinical research demonstrating risks attending forced opioid tapers.
Commonsense care of individual patients requires nuance. The best practice is not to force opioid reductions without consent, save when the justification is exceptionally compelling. Even then, the clinician must have a plan to protect the patient from harm, and reverse course if their dose reduction harms the patient. Individual clinical decisions should not be based on findings from geographic analysis of populations. Clinicians and researchers alike might best avert suicide by seeking a careful understanding of the particular patient, their history, and their context.
Stefan G. Kertesz is a Professor of Medicine and Public Health at the University of Alabama at Birmingham School of Medicine, and physician-investigator at the Birmingham Alabama Veterans Healthcare System. His views are his own, and do not represent positions of his employers. On Twitter he is at @StefanKertesz. His team’s study is at https://sites.uab.edu/csiopioids.
[1] Perhaps this was because young persons with opioid use disorder involving diverted prescriptions switched to more dangerous illicit opioids. However, this is pure speculation, we can’t tell from this study.
I am going to share this article with my GP. I started seeing them (private practice) after the hospital affiliated clinic I had been at for years suddenly tried to taper my opioids, this change was introduced to me by a doctor who had never seen me as a patient before and who had not exchanged more than 10 words with me. She mumbled something and handled me a pamphlet. She also incorrectly calculated my daily morphine equivalence and when I corrected her she told je I was wrong (I am not as my dose is under the amount stated in tbe suggested government opioid guidelines)I have several serious incurable conditions that cause severe chronic pain that affects my daily ability to function normally. Opioids have without a doubt improved my functioning and quality of life with almost zero side effects. I have NOT needed ever increasing doses and have been on the same dose for many years. On the other hand every supposedly safer “opioid alternative” that has been given to me has caused me significant & sometimes serious side effects and also not improved my pain & functioning.
Thanks for reviewing this study for the rest of us. To me, a retired pediatric pulmonologist, the idea that a single change in physician prescription practice would have a measurable impact on societal suicide incidence is absurdly simplistic and naive. There are so many factors, including the impact of the recent pandemic, the political turmoil of our time, the divisions surrounding our adolescent and young adult population, and the decreasing attendance in formal religious settings are just some of the multitude of factors that might contribute to suicide. I hope that the journal included an editorial which urged the readers to maintain a healthy skepticism of this study and others.