Saturday, October 8, 2011

Do Opioids Increase Pneumonia Risk?

Opioid Risks In a large study of elder adults, those taking opioid analgesics were found to have a higher risk of developing pneumonia than among similar persons not taking opioids. At the same time, the researchers found that benzodiazepines, often given for insomnia and anxiety, did not affect pneumonia risk in this population. However, one must question the validity of this study and whether it is merely another “myth-representation” of alleged problems with opioids.

Writing in an advance online edition of the Journal of the American Geriatrics Society (JAGS), researchers at the Group Health Research Institute and the University of Washington report a retrospective case-control study to examine whether use of opioids and/or benzodiazepines is associated with increased risk of acquired pneumonia in elderly persons. Subjects included community-dwelling, immunocompetent adults aged 65 to 94 (median age 77, half male).

Presumptive pneumonia cases were identified from health plan automated data, collected during 2000 to 2003, and validated through medical record reviews. For each case with pneumonia, 2 control subjects without pneumonia were selected for comparison, matched on age, sex, and calendar year. Information about opioid and benzodiazepine use came from computerized pharmacy data. Information on covariates, including comorbid illnesses and functional and cognitive status, was derived from medical record reviews and electronic health data.

The analysis included 1,039 validated cases of pneumonia and 2,022 matched controls. Overall, the use of opioids was found to be more frequently associated with cases of pneumonia (13.9%) than in control subjects (8.0%); the adjusted Odds Ratio (OR) for this relationship was 1.38; 95% Confidence Interval (CI), 1.08 to 1.76 vs nonuse. Additionally reported results included the following…

  • Patients taking long-acting opioids, such as sustained-release morphine, were more than 3 times as likely to get pneumonia as were those not taking opioids (OR = 3.43; 95% CI = 1.44–8.21 vs nonuse). Whereas, the odds were much lower and statistically non-significant for short-acting opioids, OR = 1.27; 95% CI = 0.98–1.64 vs nonuse.

  • Recently starting opioid use was a risk factor. During the first 14 days of use, patients who took opioids were more than 3 times as likely to get pneumonia as were those not taking opioids (OR = 3.24; 95% CI = 1.64–6.39 vs nonuse).

  • Risk of pneumonia was greater in patients using opioids having immunosuppressive qualities (OR = 1.88; 95% CI = 1.26–1.79 vs nonuse) than with non-immunosuppressive opioids (OR = 1.23; 95% CI = 0.89–1.69). Those agents reported by the authors as having immunosuppressive actions included codeine, sustained-release morphine, fentanyl, and methadone.

  • Use of opioids for a longer time period, defined as 3 months or more before getting pneumonia, was not associated with increased respiratory infection.

  • In a separate analysis of the same dataset, taking benzodiazepines with or without opioids did not affect the risk of getting pneumonia.

The authors conclude that this was the first large epidemiological study to look at how opioid use affects the risk of developing pneumonia in a community-based population. The use of opioids but not benzodiazepines appears to be associated with increased risk of developing pneumonia in older persons, and the differences in risk seen for individual opioid regimens warrant further study.

COMMENTARY: This was an interesting study, but primarily because it exemplifies many of the research biases and flaws that have created a mythology of false perceptions regarding opioid analgesic risks. Unfortunately, this study might be cited by future authors as evidence of yet another potential harm of opioids — ie, their influence on pneumonia — that is actually more myth than fact. [Other “myth-representations” of opioid harms were recently discussed in an UPDATE here.]

So, what’s wrong with this particular study?

To begin, study subjects were members of the Group Health Cooperative, a nonprofit healthcare system with extensive computerized pharmacy, laboratory, and medical records that were used for data gathering and analysis. We have previously cautioned about the limitations and biases of such data-mining exercises [eg, here]. Even worse, the data used for this present study were but a subset of information originally gathered for another purpose entirely; ie, to examine pneumonia risk in relation to influenza vaccination. The propriety and validity of reporting “newly discovered” opioid risks today that rely on but a segment of old data from 2000 to 2003 might be questioned.

There are further aspects of this research that could/should be questioned…

  • The authors report outcomes as odds ratios (OR), which can be deceptively larger than risk ratios (RR) that are more easily interpreted by readers [discussed in UPDATE here]. Necessary data is not provided in the report to convert ORs to RRs; however, the primary outcome of an OR=1.38 for opioids being more frequently associated with cases of pneumonia would most likely convert to an RR of less than 1.25, which is generally considered a small effect size. In short, the main finding of this study might have been statistically significant but was of clinically minimal importance, if any — a fact that is overlooked in the report.

    The secondary outcomes — eg, by opioid type, duration of opioid therapy, etc. — should have been similarly converted to risk values. Along with that, however, these repeated analyses of the same data would have been weakened by reduced statistical power and false positive (Type I) errors. In other words, the more one examines and manipulates the same data, the smaller the subgroup sizes become and the more likely it is to find statistically significant differences that do not truly exist.  

  • The data do not indicate why opioid analgesics were prescribed for subjects and how their pain conditions might have influenced development of pneumonia. While the authors state that opioids used in cough suppressants were excluded from analyses, they concede that in some cases opioids might have been prescribed “for early symptoms of pneumonia.” Furthermore, data were adjusted, using multivariate linear regression, to account for a long list of other potentially confounding comorbidities that might have predisposed patients to developing pneumonia, including asthma, COPD, other lung disease and associated treatments/medications, swallowing disorder, cardiovascular disease, dementia, and many others.

    It is critical to note that, with few exceptions, these comorbidities that might have independently influenced development of pneumonia apart from opioid use, were more prevalent in subjects with pneumonia than in control subjects. As just 3 examples: (a) congestive heart failure was 2.6 times more prevalent in pneumonia cases than in control subjects, (b) chronic obstructive pulmonary disease (COPD) was 3 times more prevalent in cases than in controls, (c) swallowing disorder leading to aspiration was more than 3 times as prevalent in cases compared with controls.

    Once this gross imbalance of comorbidity data in cases versus controls became apparent, abandonment of the research study might have been warranted. And, the validity of trying to statistically adjust the data for so many interacting factors in order to separate out the influence of opioids from this mix requires much more explanation and justification than the authors provide in their published report.

  • The researchers add a novel twist to their analysis by declaring that certain opioids — codeine, morphine, fentanyl, and methadone — suppress the immune system and thereby could predispose patients to infection (pneumonia in this case). However, with rare exceptions, the research exploring these effects has been done in animal models or cell cultures; immunosuppressive actions of select opioids have not as yet been adequately validated and confirmed via clinical investigations involving humans.

    In fact, just the opposite has been reported in the case of methadone; research from the addiction treatment field suggests that methadone may actually preserve [abstract here] or restore [article here] normal immune function. Therefore, the authors assumptions in this current study are somewhat theoretical and speculative; hopefully, future writers will not claim such immunosuppressive effects of opioids as fact, based on this study, until valid clinical research is conducted.

  • The top 4 opioids dispensed in the study population were hydrocodone (34.2%), codeine (24.6%), short-acting oxycodone (20.5%), and propoxyphene (9.3%) accounting for nearly 90% of 3,266 opioid prescriptions. Only one of these, codeine, was characterized as being immunosuppressive. The authors note that, in separate analyses of the top 3 agents, only codeine had a statistically significant effect on pneumonia risks (OR=1.93; 95% CI = 1.22-3.06), while oxycodone or hydrocodone use conferred no significant risk as compared with non-use. At that, however, there is no information regarding why codeine was prescribed or the dosage regimen, and the OR of 1.93 would convert to a Risk Ratio that is probably of only moderate clinical importance at most. Furthermore, research released in late 2010 found that codeine is unexpectedly associated with increased cardiovascular adverse events in the elderly [see UPDATE here], and this might contribute to its role in pneumonia independent of any alleged immunosuppressive effect. 

  • The researchers claim that long-acting opioids conferred greater pneumonia risk than short-acting formulations. However, only 7% of all prescriptions were for long-acting agents: eg, morphine SR, oxycodone SR, fentanyl transdermal, methadone, and levorphanol. While the authors do not report a statistical power analysis, they rather vaguely concede that there was “low power for some comparisons” due to too few patients in certain (unspecified) subgroups. Also, the precision of the Odds Ratio point estimate for long-acting opioids is dubious due to a wide confidence interval resulting from small group size (OR = 3.43; 95% CI = 1.44–8.21 vs nonuse). At the lower range of the CI the effect would be clinically unimportant.

Other criticisms of this research could be discussed; however, readers should by now realize that this study was a failed attempt to discredit the relative safety of opioids, particularly in elderly patients. Additionally, the authors were fully aware of the American Geriatrics Society recommendations favoring opioids over nonsteroidal anti-inflammatory drugs (NSAIDs) in older adults, so they also were trying to raise doubts about that guidelines while ignoring potentially greater risks of NSAIDs compared with opioids in this population.

Another concern is that outcomes of this study, using data nearly a decade old, may have little if any relevance for current clinical practices. The 4th most frequently prescribed opioid at that time, propoxyphene, has since been removed from the market in the U.S. and many other countries. And, in a recent evidence review [UPDATE here], codeine was ranked worst among a long list of analgesics in terms of efficacy and duration of action; so, this, combined with possible cardiotoxic effects of codeine (noted above), suggest that there is little justification for prescribing this analgesic as a sole agent today. And, as described above, individual analyses of oxycodone and hydrocodone — the other most frequently prescribed opioids — demonstrated no significant effect on development of pneumonia. Therefore, in actuality, this study provides essentially no valid evidence of practical clinical value regarding opioids and pneumonia in the elderly.

Finally, it should be noted that this is not the first time data from the Group Health Cooperative have been used to portray questionable opioid risks. A research team from this same organization reported in 2010 a data-mining study seeking to demonstrate that higher opioid doses increase overdose risks [Dunn et al. discussed in UPDATE here]. This flawed research uncovered only 51 opioid-overdose cases during 9 years of data collection (a 0.06% annual incidence rate), higher opioid doses (>100 mg/day morphine equivalent dose) accounted for only about one-fifth of the cases, and in nearly three-quarters of patients there were potentially harmful interacting drugs (including benzodiazepines) prescribed along with opioids that might have influenced overdose.

The greatest potential danger is that the two investigations described above claim significant risks of opioids that, in actuality, are not clearly supported by the research evidence. Unfortunately, many persons — professionals, patients, and news correspondents — only read the abstracts, without more critical appraisal of study details, and new myths about opioid harms are spawned.

We recommend that readers consult our ongoing series on “Making Sense of Pain Research” [listings here] to learn more about how to become critical consumers of research evidence.

REFERENCE: Dublin S, Walker RL, Jackson ML, et al. Use of Opioids or Benzodiazepines and Risk of Pneumonia in Older Adults: A Population-Based Case–Control Study. JAGS. 2011(Sep 13); online ahead of print [abstract here].