A Study of 'Problematic' Opioid Use Among Pain Patients Is Less Alarming Than It Seems

The authors of the meta-analysis misleadingly imply that pain treatment should be blamed for recent increases in drug-related deaths.

A new meta-analysis of 148 studies estimates that 9.3 percent of patients who receive opioid prescriptions for chronic pain experience "dependence" or "opioid use disorder" (D&OUD). The authors, Kyla Thomas and six other researchers at Bristol Medical School in England, also calculated prevalence rates for "signs and symptoms" of D&OUD (29.6 percent), "aberrant behavior" (22 percent), and those deemed "at risk" of D&OUD (12.4 percent). They conclude that "problematic pharmaceutical opioid use appears to be common in chronic pain patients treated with opioid analgesics, with nearly one in 10 experiencing dependence and opioid use disorder, one in three showing signs and symptoms of dependence and opioid use disorder and one in five showing aberrant behaviour."

Thomas et al. think opioids are overprescribed, and they imply that their results reinforce that assessment. But these numbers are not as alarming as they might seem, and they do not support the connection that the authors misleadingly draw between medical use of opioids and escalating drug-related deaths.

The study, published in the journal Addiction on Wednesday, is the largest meta-analysis of research on this subject so far. But Thomas et al. warn that "the findings should be interpreted with caution" given the "high heterogeneity" of the results produced by the underlying studies, most of which (115) were conducted in the United States. That is not the only reason for caution.

Stefan Kertesz, a University of Alabama at Birmingham pain and addiction specialist, thinks the study is "not bad" but says it is "important to understand what it says." He notes that most of the studies in the meta-analysis "focus on 'prevalence,'" as opposed to "new-onset incidence." Prevalence rates "may tend to overestimate the likelihood of new-onset problems" that begin after opioids are prescribed, he says. Incidence studies, by contrast, "carefully rule out anyone with a prior diagnosis," which is "incredibly hard to do."

Another issue is whether the authors of the underlying studies properly applied diagnostic criteria. "Whenever someone drags through large data records for the existence of a diagnostic code, when the folks applying those codes are mostly not trained in addiction, we can't take the code's presence at face value," Kertesz says. And even if the diagnoses are accurate, they are based on criteria that vary from study to study, depending on the definition used. "The diagnostic codes are irregular," Kertesz says, and "there are many papers that show they are applied with poor fidelity."

The studies covered by the meta-analysis, which did not include cancer patients or people who take opioids for acute pain, defined what Thomas et al. call "D&OUD" in various ways, spanning two versions of the International Classification of Diseases (ICD) and four editions of the American Psychiatric Association's Diagnostic and Statistical Manual of Mental Disorders (DSM). Keeping in mind the potential pitfalls, Kertesz says, "I would not discount or rule out" the estimate that 9.3 percent of chronic pain patients taking opioids fit one of D&OD labels.

That estimate, Kertesz notes, is similar to the results of "carefully collected prevalence studies using high-quality diagnostic interviews," such as this 2015 study reported in Lancet Psychiatry. That study generated various prevalence estimates, including 9.1 percent based on DSM-5 criteria that excluded "tolerance" and "withdrawal" (because those are predictable consequences of regular, long-term opioid use that do not necessarily signify a serious problem), 11.6 percent based on those criteria without the exclusions, and 8.5 percent based on the ICD-10 criteria for "dependence."

The estimate "goes way up" to 18.6 percent, Kertesz notes, "with what they describe as ICD-10 use disorder," which is "not much used." Under DSM-IV, the prevalence estimates were 8.9 percent for "dependence" and 10.1 percent for "abuse," which is "kind of odd," he says, since the latter category describes a much broader range of behavior.

Kertesz also wonders whether prevalence estimates from older studies, conducted at a time when prescribing practices were looser, accurately reflect the current situation. "Today's prescribing is much more carefully regulated," he says, "and actual U.S. opioid prescribing is down…to levels per capita last seen in 1992 or a bit before then. So if we adopt analyses that involve lots of U.S. data from our high-prescribing [period] and assume it also characterizes prevalence after prescribing went down so much, we might just be wrong."

Most of what Thomas et al. describe as "problematic pharmaceutical opioid use" falls short of behavior that would qualify as "dependence" or "opioid use disorder." That wide net is consistent with the concern they express about using opioids to relieve chronic pain.

"Chronic non-cancer pain (CNCP), defined as pain which is not cancer-related and which lasts for longer than 3 months, is one of the most common causes of disability globally," Thomas et al. write. "Population-based studies have reported that almost one in five US adults and as many as one in two UK adults suffer with the condition, which is often managed in primary care settings. Despite limited evidence of long-term clinical benefit of opioids for CNCP and guidelines advising against their use for many pain conditions, opioid analgesics continue to be widely prescribed." And that, as Thomas et al. see it, is a problem that can be counteracted by looking at broader measures of "problematic pharmaceutical opioid use."

This anti-opioid bias is apparent in the authors' alarm at the rates they calculated, which are not very different from the numbers that Nora Volkow, director of the National Institute on Drug Abuse, and A. Thomas McLellan, former deputy director of the Office of National Drug Control Policy, presented in a 2016 New England Journal of Medicine article. "Although published estimates of iatrogenic addiction vary substantially from less than 1% to more than 26% of cases," they wrote, "part of this variability is due to confusion in definition. Rates of carefully diagnosed addiction have averaged less than 8% in published studies, whereas rates of misuse, abuse, and addiction-related aberrant behaviors have ranged from 15 to 26%."

In general, Volkow and McLellan noted, "addiction occurs in only a small percentage of persons who are exposed to opioids—even among those with preexisting vulnerabilities." Yet Thomas et al. describe their similar numbers as evidence of a problem that "appears to be common," which is more a matter of spin than a product of new information revealed by the meta-analysis.

The "confusion in definition" that Volkow and McLellan mentioned remains a challenge. DSM-5, for example, combined what were previously two labels, "substance dependence" and the less serious "substance abuse," into the broad category of "substance use disorder" (SUD). SUD, which is one of the diagnoses that figures in Thomas et al.'s calculation of D&OUD prevalence, spans a wide range of cases, including many that would not fit the old definition of "substance dependence" or what Volkow and McClellan consider "carefully diagnosed addiction."

Moving beyond D&OUD magnifies the difficulty of assessing the seriousness of the problems reflected in Thomas et al.'s numbers. They concede, for instance, that the "aberrant behavior" measure, which 79 of the studies included, is iffy.

"Although aberrant behaviours may indicate a problematic relationship with prescription opioids, other contextual factors may influence the likelihood of these behaviours," Thomas et al. write. "For example, missed clinic visits, no-shows or no follow-up may be associated with other social issues or disability, and not reflect POU….Other aberrant behaviours, such as early refills or multiple prescriptions, may be indicative of inadequate pain control. Despite these shortcomings, we felt that it was still important to include a prevalence estimate of aberrant drug-related behaviours to aid understanding of the extent of the problem of POU in this cohort of patients."

Forty-four of the studies reported results for "signs and symptoms" of D&OUD. This is a broader category that encompasses patients who did not qualify for any drug-related diagnosis—not even the commodious "opioid use disorder" label. Thomas et al. say it requires "the clear presence of behaviours such as craving, tolerance, withdrawal or a loss of control over use (for example continued use despite psychological or physical harm or use which takes priority over usual obligations)." But some of these things are not like the others. While "continued use despite psychological or physical harm" is bad by definition, for example, "tolerance" and "withdrawal" are suspect when applied to chronic pain patients, which is why the Lancet Psychiatry study included a DSM-5 estimate without them.

The "at risk of D&OUD" category, which was based on eight studies, includes patients who "exhibit characteristics which may increase their risk of developing opioid dependence or opioid use disorder in the future" but "do not show aberrant behaviour or meet criteria for dependence or opioid use disorder." Thomas et al. say "studies which used a high score (≥ 8) on the Opioid Risk Tool (ORT) to define [problematic pharmaceutical opioid use], and which were originally classified as misuse by study authors, were included in this category." But they note that "it is likely to be a very weak indicator of actual [problematic pharmaceutical opioid use] or D&OUD."

Thomas et al. complain that opioids are "widely prescribed" for chronic pain, a practice they suggest is not supported by scientific evidence or by "guidelines advising against their use for many pain conditions." That is worrisome, they say, because "long-term opioid prescribing has been associated with many harms, including accidental and fatal prescription opioid overdose." They add that the United States is experiencing "a public health emergency" due to "overdose deaths caused by prescription opioid analgesics, illicit opioids such as heroin and synthetic opioids such as fentanyl."

Thomas et al. cite a 2022 editorial in The Lancet about "the opioid epidemic," which it described as "one of the worst public health disasters affecting the USA and Canada." That epidemic, the editorial said, "involves both prescribed opioids such as oxycodone and illicit drugs such as heroin."

These descriptions are highly misleading. According to the U.S. Centers for Disease Control and Prevention (CDC), 90 percent of opioid-related deaths in 2022 involved "synthetic opioids other than methadone," the category that includes illicit fentanyl. Less than 15 percent involved "natural and semisynthetic opioids," the category that includes "prescribed opioids such as oxycodone," and some of these deaths also involved black-market drugs. The hazard that Thomas et al. cite as a reason to worry about the use of opioids to treat chronic pain, in other words, has little to do with medications prescribed for that purpose.

The CDC's data do not tell us how often the opioids that worry Thomas et al. were prescribed for the people who died after using them. But a 2019 Massachusetts study of opioid-related deaths found that decedents had prescriptions for the opioids that showed up in toxicology tests just 1.3 percent of the time. The risk of a fatal overdose for bona fide pain patients seems to be very small: A 2016 study of opioid-related deaths in North Carolina found just 478 fatalities among 2.2 million residents who were prescribed opioids in 2010, an annual rate of 0.022 percent.

A 2021 European Psychiatry study of "19 European countries and the United States," Thomas et al. note, "found that the United Kingdom had the highest consumption of prescription opioids." Yet while Scotland had the highest opioid-related death rate, England and Wales ranked seventh, far below the United States. "Scotland was a clear outlier, with more opioid-related overdose deaths than in the US during the entire observation period," the authors reported. "Heroin and methadone were the main drugs involved in opioid-related mortality in Scotland." Again, this does not look like evidence that prescribed pain medications like oxycodone and hydrocodone are playing an important role in opioid-related deaths.

In the United States, those deaths overwhelmingly involve illicit fentanyl, and they rose at an accelerated rate even as opioid prescriptions declined substantially. That was entirely predictable, since the crackdown on prescriptions drove nonmedical users toward black-market substitutes that are much more dangerous because their composition is highly variable and unpredictable. Meanwhile, pain patients suffered needlessly as the medications that made their lives bearable became increasingly difficult to obtain. Yet Thomas et al. invoke "the increase in opioid-related fatalities" as a reason for doctors to be even warier of those medications.

The "big picture," Kertesz says, "is that if you prescribe chronic opioids for pain, you are inducing dependency with the hope it will be therapeutic. There is always risk of some kind involving misuse, and you have to inform the patient." Although estimates vary, he says, "the likely risk of inducing a new-onset addiction or new-onset problematic behavior is non-trivial." But exaggerating that risk, as Thomas et al. do by linking pain treatment to "the increase in opioid-related fatalities," poses a different sort of danger, as patients across the country have discovered during the last decade.