Article Text
PDF
Abstract
The efficacy of low-dose systemic opioids for chronic breathlessness was questioned by the recent Cochrane review by Barnes et al. We examined the reasons for this conflicting finding and re-evaluated the efficacy of systemic opioids. Compared with previous meta-analyses, Barnes et al reported a smaller effect and lower precision, but did not account for matched data of crossover trials (11/12 included trials) and added a risk-of-bias criterion (sample size). When re-analysed to account for crossover data, opioids decreased breathlessness (standardised mean differences −0.32; −0.18 to −0.47; I2=44.8%) representing a clinically meaningful reduction of 0.8 points (0–10 numerical rating scale), consistent across meta-analyses.
- Palliative Care
- COPD Pharmacology
- Perception of Asthma/Breathlessness
Statistics from Altmetric.com
Introduction
Chronic breathlessness1 is common across a range of advanced diseases and associated with major adverse health outcomes.2 The candidate treatment with best evidence to date is regular, low-dose, non-nebulised (systemic) morphine.2 The efficacy of low-dose systemic opioids was supported by a Cochrane review by Jennings et al,3 ,4 an adequately powered crossover trial in 2003,5 and the meta-analysis in people with severe COPD by Ekström et al.6
A new Cochrane meta-analysis by Barnes et al,7 drawing from a similar evidence base, reported a smaller benefit of opioids than the other reviews, and wider 95% CIs which nearly crossed zero. The risk of bias was rated as ‘high’ for all studies; previous ratings were mainly ‘unclear’ or ‘low’.3 ,4 ,6 Barnes et al7 rated the quality of evidence for opioids for breathlessness as ‘very low’.
We aimed to determine the reasons for the different conclusions and to re-evaluate the efficacy of systemic opioids for chronic breathlessness.
Methods
Data were extracted from the published meta-analyses by Jennings et al,3 ,4 Ekström et al6 and Barnes et al7 (by ME), and cross-validated (DCC and MJJ) regarding study populations, designs, interventions and methods, for the whole study population and in participants with COPD, respectively.
Breathlessness measures were analysed as standardised mean differences (SMD).8 For crossover trials, the SE was estimated using the crossover information, directly from the published report or calculated from significance test statistics as recommended.8 The effect of opioids compared with placebo was analysed using a random effects model. A detailed description of the statistical methods is given in the online supplementary file (appendix 1).
supplementary file
Results
Included studies
All included studies were double-blind, placebo-controlled randomised trials; 13/14 studies were crossover designs (table 1). Jennings et al and Barnes et al included patients with any advanced, life-limiting disease, whereas Ekström et al restricted the analysis to patients with COPD. Research questions, interventions, comparisons and treatment durations were similar between the three meta-analyses (table 1).
Characteristics of meta-analyses of systemic opioids for breathlessness
The study populations overlapped significantly with over half of the studies by Barnes et al also included in studies by Jennings et al and Ekström et al (see online supplementary table S1). For two studies omitted by Barnes et al, the reasons for exclusion were not stated.
Efficacy
In contrast to the other meta-analyses, Barnes et al used a fixed effects model which does not account for variations in the true effect between studies, and analysed all data as if from parallel trials and did not account for matched crossover data (11/12; 92% of included studies).
Opioids were associated with a decrease in breathlessness in both studies by Jennings et al and Ekström et al (table 1). In the primary analysis of Ekström et al, systemic opioids improved breathlessness in COPD outpatients measured at steady state (5 studies, 91 participants), SMD −0.33 (95% CI −0.52 to −0.14).
Barnes et al split the analysis by route of administration and type of outcome measure (table 1). Point estimates of efficacy ranged from SMD −0.27 (oral opioid, post-treatment scores) to mean difference 0.20 (subcutaneous opioid, change scores). Precision was markedly lower across all analyses. The estimate for COPD in the study by Barnes et al included all types of both systemic and nebulised opioids. Estimates for systemic opioids or efficacy at steady state were not reported.
When Barnes et al was re-analysed using a random effects model accounting for crossover data (figure 1), opioids decreased breathlessness, SMD −0.32 (95% CI −0.47 to −0.18; p<0.001; I2=44.8%) compared with placebo, consistent with the studies by Jennings et al and Ekström et al. Using the SD from a large study,5 this effect size corresponds to a reduction of 0.8 points on a 0–10 numerical rating scale. The finding was consistent when excluding the three studies for which the SEs were imputed.
The meta-analysis of Barnes et al7 re-analysed using random effects model and accounting for matched data of crossover trials. In the pooled analysis compared with placebo, systemic opioids reduced breathlessness by a mean 0.32 (95% CI 0.18 to 0.47; p<0.001) SDs.
Risk of bias and quality of evidence
Conclusions regarding risk of bias were similar between the studies by Jennings et al and Ekström et al, with unclear or low risk of bias for most items (table 1). In contrast, Barnes et al categorised all studies as having high risk of bias due to low sample size defined as <50 participants in each treatment arm. This criterion had no stated rationale and resulted in the quality of evidence for systemic opioids being downgraded from moderate (Ekström et al) to low or very low in the study by Barnes et al (table 1).
Discussion
The conflicting findings regarding the efficacy of opioids for chronic breathlessness in the recent Cochrane review are likely due to their use of inappropriate methodology. When re-analysed to account for crossover data, opioids were associated with a statistically and clinically significant reduction in breathlessness,9 consistent across meta-analyses.3 ,4 ,6
Analysing crossover studies as parallel studies can result in selection bias, with spuriously too high or too low effect estimates, as well as reduced precision.10 Recommended methods to account for crossover data are available10 and were used by Jennings et al4 and Ekström et al.3 ,6 In addition, study selection should align to predefined eligibility criteria with reasons for exclusion stated to minimise selection bias.
While any judgement of risk of bias is subjective, the bias criterion related to study size introduced by Barnes et al, which resulted in all studies being rates as high risk of bias, is questionable. It is the power of the study which could lead to bias, and not the sample size per se, which is based on the power calculation. Adequate power can be provided by trials with total sample sizes below 50,5 especially in crossover trials where the participants act as their own control thus increasing power.
We suggest that the analysis by Barnes et al and the relevant guidelines for analysis and review of the Cochrane Collaboration are updated to accommodate these issues.
Conclusion
Moderate level evidence to date supports that regular, low-dose morphine is the first-line pharmacological treatment for the relief of chronic breathlessness in severe illness.
Footnotes
Contributors Concept and design: ME, DCC, JMB, MJJ, JH; first draft: ME, SB; statistical analysis: JMB; interpretation, revision and acceptance of the final version to be published: ME, SB, DCC, JMB, JH, MJJ.
Funding ME was supported by unrestricted grants from The Swedish Society of Medicine, the Swedish Respiratory Society, the Swedish Heart-Lung Foundation, the Scientific Committee of Blekinge County Council and the Wera and Emil Cornell Foundation.
Competing interests DCC has received intellectual property payments and advisory board payments. MJJ has been a clinical consultant for Mayne Pharma. Authors of this paper have longstanding interest in the research of breathlessness and have published several opioid-related trials and meta-analyses (including ME and DCC6). MJJ was an external clinical academic (not statistical) peer reviewer for the original Barnes et al protocol submitted to Cochrane.
Provenance and peer review Not commissioned; externally peer reviewed.