Abstract
The potential for respiratory-related harm needs to be considered when using opioid drugs in non-palliative COPD http://ow.ly/i5W03095MKV
From the authors:
We thank M. Ekström and co-workers for their interest in our manuscript [1] and for their contributions to the important topic of opioid drug use in chronic obstructive pulmonary disease (COPD). M. Ekström and co-workers stress the efficacy of selective use of opioids for refractory breathlessness in COPD. We acknowledged in our manuscript [1] that several clinical studies demonstrate that opioids are efficacious in selected COPD patients for relieving refractory breathlessness. However, we have previously outlined reasons why such trials of drug efficacy have limited ability to comprehensively evaluate for possible drug harms [1–3]. Furthermore, chronic musculoskeletal pain, and not respiratory symptoms, is likely the more common reason why individuals with COPD use opioid drugs [3, 4]. Indeed, M. Ekström and co-workers recently published data showing that 97% of opioid prescriptions among Swedes with advanced COPD were for pain [5]. Several Cochrane reviews have suggested there is insufficient evidence to support the use of opioids for chronic musculoskeletal pain [6–8].
We would like to clarify and correct several points that M. Ekström and co-workers made regarding our study. First, the authors appear to focus on one result, that the absolute risk of all-cause mortality was only 0.8% higher among community-dwelling incident opioid users versus non-users. Although this specific risk difference is small, it may be clinically important at a population level. Furthermore, the absolute risk difference for all-cause mortality between the exposed and control groups was larger when one considered use of opioid-only formulations in the community (3.6%; number needed to harm (NNH) 28) or opioid use among long-term care residents (9.3%; NNH 11) [1].
Second, the authors wrote that “stronger opioids were associated with lower risk than weaker opioids”. We reported the opposite. Use of more potent opioid-only agents was associated with greater respiratory-related morbidity and mortality than use of the less potent combination opioid/non-opioid agents (see table 4 in [1]).
Third, Ekstrom et al. wrote “the risk estimates varied markedly between sensitivity analyses with an almost 50% decrease in the excess risk in some models”. This statement is not accurate. Our sensitivity analyses showed an overall pattern of increased risk for adverse events among opioid users, regardless of opioid half-life duration, opioid dose and COPD severity.
Fourth, the authors propose that the increased adverse respiratory event risk associated with incident opioid use may be due to unmeasured factors not included in our propensity score model, like respiratory symptoms, general health status, and opioid indication and dose. While we do not have data on respiratory symptoms in our databases, we incorporated other markers of COPD severity in our propensity score model, the most important of which was COPD exacerbation frequency, which is known to be associated with quality of life [9]. We also included multiple markers of general health status in our propensity score model, including nine different comorbidities. Opioid indication and dose could not be included in the propensity score model, because these variables apply only to the exposed and not the control group. Instead, we conducted a sensitivity analysis by COPD severity and found increased mortality associated with opioid use among the healthiest subgroup of individuals and this subgroup would be the least likely to receive opioids for respiratory symptoms (see table 7 in [1]). We also conducted an opioid dosing sensitivity analysis (see table 6 in [1]).
Fifth, we acknowledge that we did not incorporate nonsteroidal anti-inflammatory drug (NSAID) use in our propensity score model and this could have resulted in residual confounding. We included the presence of musculoskeletal and connective tissue diseases in our propensity score model (individuals with these comorbidities would be the ones more likely to use NSAIDs) and the exposed and control groups were well balanced on this variable. Furthermore, while concomitant NSAID use among opioid recipients may have contributed to the finding of increased all-cause mortality via cardiac ischaemia or gastrointestinal tract bleeding, we fail to see how NSAID use would explain the higher risks of outpatient respiratory exacerbations and COPD or pneumonia-related emergency room visits and hospitalisations among opioid users.
Finally, M. Ekström and co-workers point out that cancer deaths were higher in the exposed versus control group in our study. While this is correct, we mentioned in our manuscript that we repeated our analysis, excluding individuals with lung cancer and any cancer, and we still observed increased respiratory-related morbidity and mortality with incident opioid use (see online supplement 1 in [1]).
We agree with a concluding comment by M. Ekström and co-workers that “pooled analyses of randomised and pharmacovigilence data” on opioids are needed to optimally inform clinical practice, as this approach combines the best data on drug efficacy with the best data on drug harms [10].
Disclosures
Supplementary Material
O. Donnell ERJ-02529-2016_ODonnell
Footnotes
Conflict of interest: Disclosures can be found alongside this article at erj.ersjournals.com
- Received December 23, 2016.
- Accepted December 23, 2016.
- Copyright ©ERS 2017