Opioids for the palliation of refractory breathlessness in adults with advanced disease and terminal illness

Hayley Barnes; Julie McDonald; Natasha Smallwood; Renée Manser

doi:10.1002/14651858.CD011008.pub2

. 2016 Mar 31;2016(3):CD011008. doi: 10.1002/14651858.CD011008.pub2

Opioids for the palliation of refractory breathlessness in adults with advanced disease and terminal illness

Hayley Barnes ^1,^✉, Julie McDonald ^2,³, Natasha Smallwood ⁴, Renée Manser ⁵

Editor: Cochrane Pain, Palliative and Supportive Care Group

PMCID: PMC6485401 PMID: 27030166

Abstract

Background

Breathlessness is a common and disabling symptom which affects many people with advanced cardiorespiratory disease and cancer. The most effective treatments are aimed at treating the underlying disease. However, this may not always be possible, and symptomatic treatment is often required in addition to maximal disease‐directed therapy. Opioids are increasingly being used to treat breathlessness, although their mechanism of action is still not completely known. A few good sized, high quality trials have been conducted in this area.

Objectives

To determine the effectiveness of opioid drugs in relieving the symptom of breathlessness in people with advanced disease due to malignancy, respiratory or cardiovascular disease, or receiving palliative care for any other disease.

Search methods

We performed searches on CENTRAL, MEDLINE, EMBASE, CINAHL, and Web of Science up to 19 October 2015. We handsearched review articles, clinical trial registries, and reference lists of retrieved articles.

Selection criteria

We included randomised double‐blind controlled trials that compared the use of any opioid drug against placebo or any other intervention for the relief of breathlessness. The intervention was any opioid, given by any route, in any dose.

Data collection and analysis

We imported studies identified by the search into a reference manager database. We retrieved the full‐text version of relevant studies, and two review authors independently extracted data. The primary outcome measure was breathlessness and secondary outcome measures included exercise tolerance, oxygen saturations, adverse events, and mortality. We analysed all studies together and also performed subgroup analyses, by route of administration, type of opioid administered, and cause of breathlessness. Where appropriate, we performed meta‐analysis. We assessed the evidence using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach and created three 'Summary of findings' tables.

Main results

We included 26 studies with 526 participants. We assessed the studies as being at high or unclear risk of bias overall. We only included randomised controlled trials (RCTs), although the description of randomisation was incomplete in some included studies. We aimed to include double blind RCTs, but two studies were only single blinded. There was inconsistency in the reporting of outcome measures. We analysed the data using a fixed‐effect model, and for some outcomes heterogeneity was high. There was a risk of imprecise results due to the low numbers of participants in the included studies. For these reasons we downgraded the quality of the evidence from high to either low or very low.

For the primary outcome of breathlessness, the standardised mean post‐treatment dyspnoea score was 0.32 points better in the opioid group compared to the placebo group (ranging from a 0.53 point reduction to a 0.10 point reduction) (12 RCTs, 338 participants, low quality evidence). The standardised mean change from baseline dyspnoea score was 0.11 points better in the opioids group compared to the placebo group (ranging from a 0.40 point reduction to a 0.19 increase) (six RCTs, 194 participants, very low quality evidence). A lower score indicates an improvement in breathlessness.

The evidence for the six‐minute walk test (6MWT) was conflicting. The total distance in 6MWT was 28 metres (m) better in the opioids group compared to placebo (ranging from 113 m to 58 m) (one RCT, 11 participants, very low quality evidence). However, the change in baseline was 48 m worse in the opioids group (ranging from 36 m to 60 m) (two RCTs, 26 participants, very low quality evidence).

The adverse effects reported included drowsiness, nausea and vomiting, and constipation. In those studies, participants were 4.73 times more likely to experience nausea and vomiting compared to placebo, three times more likely to experience constipation, and 2.86 times more likely to experience drowsiness (nine studies, 162 participants, very low quality evidence).

Only four studies assessed quality of life, and none demonstrated any significant change.

Authors' conclusions

There is some low quality evidence that shows benefit for the use of oral or parenteral opioids to palliate breathlessness, although the number of included participants was small. We found no evidence to support the use of nebulised opioids. Further research with larger numbers of participants, using standardised protocols and with quality of life measures included, is needed.

Plain language summary

Opioids for treating breathlessness at the end of life

Background

People with lung disease may experience breathlessness. Initial treatments should focus on the underlying causes of breathlessness. However, as the disease progresses, it may be better to focus on treating the symptoms. As well as standard care, opioids (e.g. morphine, given either by mouth, by nebuliser, or injected) may help relieve these symptoms. However, opioids also have side effects, such as drowsiness, constipation, nausea (feeling sick), and vomiting.

Review question

We wanted to know if opioid drugs reduced breathlessness in people with lung disease. We also looked at whether opioids improved their ability to exercise, and what side effects people had. We also wanted to know if opioid drugs improved their quality of life.

Study characteristics

We searched for studies up to 19 October 2015, and we included 26 studies with 526 people. These people had breathlessness from different types of lung disease. Some were given opioid drugs and some were given other drugs or a placebo, and studies compared the reporting of breathlessness to see if there was any difference. Some studies also looked at the amount of time people could exercise to see if there were any differences. Some people came from home, and some came from the hospital setting.

Key findings

There was some low quality evidence that showed a benefit of using oral or injectable opioid drugs for the treatment of the symptoms of breathlessness. There was no evidence for opioids by nebuliser. Some people experienced drowsiness, nausea, and vomiting. More research is needed using more people, and looking at effects on quality of life.

Quality of the evidence

We rated the quality of the evidence using one of the following grades: very low, low, moderate, or high. Very low quality evidence means we are uncertain about the results. High quality evidence means we are very certain about the results. For this Cochrane review, we found that the evidence was of low to very low quality. We included randomised controlled trials which were blinded, which means that participants and those people that assessed the results did not know whether the participants had received the opioid drug or a placebo. However, the trials were of small size, and some studies did not give enough information to allow us to assess whether they were of good quality.

Summary of findings

Background

Description of the condition

Breathlessness may be described as "a subjective experience of breathing discomfort that consists of qualitatively distinct sensations that vary in intensity" (ATS 1999). Breathlessness, also termed dyspnoea, shortness of breath, air hunger, awareness of respiratory distress, or laboured breathing, may be variably perceived by different patients, depending on multiple physiological, psychological, social, environmental, and cultural factors (Guz 1997). It is a common symptom at the advanced stages of illness, and may be as disabling to the patient and their families as pain, nausea and vomiting, delirium, and other end of life symptoms (Neuman 2006).

Respiratory motor activity is regulated by automatic centres in the brainstem and voluntary signals from the cortex, and controls chest wall expansion, lung inflation, and ventilation. Feedback is provided by chemoreceptors, mechanoreceptors, and sensory receptors. Breathlessness may be explained by a mismatch between afferent sensory information processed at the cortex and respiratory motor command from the cortex and brainstem. Alterations in arterial blood pH (acidity), partial pressure of carbon dioxide (pCO₂), and partial pressure of oxygen (pO₂) stimulate central chemoreceptors in the medulla and peripheral chemoreceptors in the carotid and aortic bodies, which transmit impulses to the brainstem respiratory centres, and adjust breathing based on acid base homeostasis (Nattie 1995; Fitzgerald 1986). Mechanoreceptors and stretch receptors located in the lung parenchyma and bronchioles sense changes in the expansion of the lung and become irritated by certain mechanical and chemical stimuli, and affect subsequent levels and patterns of breathing (Nishino 2011). Changes in air flow, smooth muscle tone, and impulses from C fibres located adjacent to the alveoli and pulmonary capillaries respond to changes in pulmonary interstitial and capillary pressures (Widdicombe 1982). Sensory receptors in respiratory muscle and the diaphragm involved in spinal and supraspinal reflexes influences central respiratory activity (Bolsher 1987; Bolsher 1988). Each of these mechanisms may contribute to the mismatch of neural activity and consequent mechanical and ventilatory outputs, and create sensations of dyspnoea, air hunger, and increased desire to breathe, which may cause distress.

Recent neuroimaging studies also suggest that neural structures that involve pain and dyspnoea may be shared, further contributing to the affected person's discomfort and distress associated with an increased sensation of ventilation (Brannan 2001; Liotti 2001; Parsons 2001; Peiffer 2001; Evans 2002; von Leupoldt 2009).

There are many currently incurable and progressive cardiopulmonary, neuromuscular, and malignant conditions in which dyspnoea is a common symptom in the advanced stages of disease. The dominant mechanism that leads to dyspnoea may vary between conditions and in many conditions more than one mechanism may be responsible. Illnesses such as interstitial lung disease, pulmonary hypertension, and congestive heart failure stimulate pulmonary receptors (irritant, mechanical, and vascular) leading to an increased respiratory drive and increased afferent input to the respiratory centre. Chronic conditions that are severe enough might also lead to gas exchange abnormalities through mechanisms such as ventilation‐perfusion (V/Q) mismatching (e.g. pulmonary vascular disease) or diffusion impairment (e.g. interstitial lung disease) leading to stimulation of chemoreceptors and increased respiratory drive. Conditions that reduce the oxygen‐carrying capacity of the blood (e.g. anaemia) or reduce cardiac output (e.g. cardiac failure) also stimulate chemoreceptors. Respiratory muscle weakness in conditions such as motor neurone disease or myopathy, and decreased compliance of the chest wall in conditions such as severe kyphoscoliosis and pleural effusion, impair ventilatory mechanics which reduces the afferent feedback for a given efferent input (Manning 1995). There are multiple potential aetiological factors related to breathlessness in chronic obstructive pulmonary disease (COPD). There is an increased resistive load from narrowing of the airways and increased elastic load from hyperinflation resulting in impaired ventilator mechanics. In addition, hypoxia and or hypercapnia may be present, leading to stimulation of chemoreceptors, and finally dynamic airway compression may stimulate receptors within the airway (Parshall 2012).

Multiple mechanisms for breathlessness have also been described in individuals with advanced cancer (Mazzocato 1999). Cancers that involve the lungs may obstruct airways leading to ventilation perfusion mismatch, and pleural effusions are common. Many people with lung cancer also have COPD. Dudgeon 1998 showed that people with terminal cancer often have abnormal spirometry (most commonly a mixed obstructive/restrictive pattern or a restrictive pattern). They also found that respiratory muscle weakness may be an important contributor to dyspnoea and that co‐morbidities such as anaemia and cardiac disease are common.

Initial approaches should aim to treat the underlying causes of breathlessness. However, as the disease progresses, such treatments may be less appropriate due to decreased effectiveness and discomfort caused to the person, and a more symptom‐based approach may be required.

Many pharmacological and non‐pharmacological interventions have been recommended to help alleviate symptoms of breathlessness in advanced disease. Management of symptoms is often multimodal, with varying treatments utilised depending on the person's co‐morbidities, and psychosocial, environmental, and cultural factors.

A Cochrane systematic review on non‐pharmacological interventions demonstrated efficacy for neuro‐electrical muscle stimulation, chest wall vibration, walking aids, breathing training, and use of hand‐held fans (Bausewein 2008). Another Cochrane review demonstrated effectiveness of exertional oxygen therapy in non‐hypoxaemic COPD patients (Uronis 2011), and suggested a slight, but not statistically significant, improvement in adults with heart failure, cancer (not end‐stage disease), and kyphoscoliosis (Cranston 2008).

Some guidelines recommend opioids as the first‐line pharmacological treatment for breathlessness (ATS 1999; Mahler 2010; Parshall 2012; Mahler 2013; Wiseman 2013). A Cochrane review published in 2001 concluded that there was some evidence to support the use of oral and parenteral opioids to palliate breathlessness, but the number of participants studied was small and they recommended that larger trials were needed using standard protocols and incorporating quality of life measures (Jennings 2001).

A Cochrane review (Simon 2010), found no evidence for a beneficial effect of benzodiazepines for the relief of breathlessness in people with advanced cancer and COPD. However, the overall effect size was small and further research is required.

Description of the intervention

Opioids are chemical substances derived from the opium poppy. In the human body they bind to the μ, κ, and δ receptors located in the cerebral cortex, limbic system, midbrain, brainstem, and outside the central nervous system in the bronchioles, alveolar walls, myocardial cells, peripheral sensory nerve fibres, and primary afferent neurons.

How the intervention might work

Exogenous and endogenous opioids specifically bind to the μ receptors to reduce transmission of pain signals (Chahl 1996). Opioids also depress respiratory drive by directly blunting the responsiveness of the brainstem centres, which are affected by hypoxia and hypercapnia. Decreased respiratory output results in a decrease in corollary discharge from the brainstem to perceptual areas in the cerebral cortex and thus reduced the sensation of breathlessness. Corollary discharge describes the hypothesis that a sensory 'copy' of the motor output is sent from the motor cortex to the sensory cortex and imparts a conscious awareness of respiratory effort (Beach 2006).

Opioids may also cause blunting of perceptual sensitivity to sensations of breathlessness. Neuroimaging studies demonstrate that μ opioid receptor agonists can modulate the central processing of breathlessness similar to that of pain relief. Administration of opioids stimulate activity in the anterior cingulate cortex, thalamus, frontal cortex, and brainstem, the same areas which are activated when breathlessness occurs (Banzett 2000; Peiffer 2001; Petrovic 2002; Pattinson 2009).

Peripheral opioid receptors are located in bronchioles and alveolar walls of the respiratory tract (Zebraski 2000). Opioid administration may modulate breathlessness by binding to these opioid receptors. It is theorised that opioid administration could modulate breathlessness by binding to these peripheral opioid receptors. However, to date, studies of nebulised opioids have lacked efficacy compared with systemically administered opioids, and there is a lack of efficacy when nebulised opioids are compared with systemically administered opioids (Polosa 2002; Mahler 2013).

Other effects of opioids include drowsiness, euphoria, confusion, peripheral vasodilation, constipation, nausea and vomiting, and cough suppression.

The choice of preparation and pharmacokinetics of opioids may vary depending on individual needs. Small doses of short‐acting opioids may be commenced in opioid‐naïve people, and once a stable dose has been achieved, may be switched to long acting preparations. Currow 2011 found that 70% of participants derived benefit from 10 mg sustained‐release once‐daily preparations. Transmucosal, transdermal, subcutaneous, or intravenous modes may be more appropriate for people whose swallowing is impaired or who are approaching the final stages of end of life. It is unclear if all opioids and all routes are equal in their ability to relieve breathlessness. Opioids differ significantly in their pharmacodynamic properties, from differences in their absorption, to metabolism and affinity for receptors.

Why it is important to do this review

The use of opioids to treat breathlessness in advanced illness is variably accepted in medical practice, and some health professionals and patients have concerns regarding efficacy and side effects (Oxberry 2012; Rocker 2012). Much of the literature around opioids for breathlessness are narrative reviews and opinion pieces, and a systematic review is required to specifically examine the quality of evidence from randomised controlled trials (RCTs), to evaluate efficacy in terms of symptom control and quality of life, and to assess adverse effects.

This review will build on a previous Cochrane systematic review (Jennings 2001). In more recent years, additional RCTs have been published (Mazzocato 1999; Johnson 2002; Abernethy 2003), mechanisms of action have been further elucidated, and guidelines that examine the risk of bias and assessment of heterogeneity in Cochrane reviews have been updated (Higgins 2011).

Objectives

To determine the effectiveness of opioid drugs in relieving the symptom of breathlessness in patients with advanced disease due to malignancy, respiratory or cardiovascular disease, or receiving palliative care for any other disease.

Methods

Criteria for considering studies for this review

Types of studies

We included parallel‐group randomised controlled trials (RCTs) compared to either placebo or other treatment, as well as crossover studies in which participants were randomised to order of treatment. We defined 'randomised' as studies that were described by the study author as 'randomised'. There was no language restriction. All identified trials, published and unpublished, were eligible for inclusion.

Types of participants

We considered adults with any type of advanced progressive illness with persistent breathlessness despite optimal or appropriate treatment of reversible factors.

We also included participants suffering from breathlessness due to any type of illness, who were considered to be at an advanced stage of illness, or palliative stage, as defined by the study authors.

Types of interventions

Any opioid drug, given by any route in any dose, for the treatment of breathlessness compared to placebo, or any other pharmacological or non‐pharmacological interventions that were directly compared with the opioid treatment.

Types of outcome measures

Primary outcomes

Subjective measurement of breathlessness intensity or severity, including but not limited to Borg and the modified Borg scale, verbal categorical scales of breathlessness, and visual analogue scales (VAS) of breathlessness (O'Donnell 1998).

Secondary outcomes

Quality of life measure by any scale.
Any physiological and functional assessments of breathlessness including but not limited to six‐minute walk tests (6MWT), shuttle tests, and actigraphy.
Performance status.
Pulse oximetry.
Arterial blood gas analysis.
Adverse events including constipation, delirium, and others.
Mortality.

Search methods for identification of studies

Electronic searches

We searched the following electronic databases up to 19 October 2015.

The Cochrane Central Register of Controlled Trials (CENTRAL) (the Cochrane Library) Issue 10 of 12, 2015.
MEDLINE (OVID) 1946 to October week 2 2015.
EMBASE (OVID) 1974 2015 October 16.
CINAHL(EBSCO) 1982 to Octpber 2015.
Web of Science (ISI) to October 2015.

We have presented the search strategies we used in Appendix 1.

Searching other resources

For ongoing studies we searched the following up to 19 October 2015.

The metaRegister of Controlled Trials (mRCT) (http://www.controlled‐trials.com/mrct/).
ClinicalTrials.gov (http://clinicaltrials.gov/).
The World Health Organization (WHO) International Clinical Trials Registry Platform (ICTRP) search portal (http://apps.who.int/trialsearch/).

We handsearched reference lists of included studies, relevant chapters, and review articles. We used Google to search for conference abstracts.

We attempted to contact the trial investigators of two studies to determine the potential for inclusion. However, we did not receive a reply by the time we completed this review.

Data collection and analysis

Selection of studies

Two review authors (RM and HB) independently screened all abstracts to determine whether they met the inclusion criteria. We sought the full‐text publications of articles that definitely met or may have met the inclusion criteria. Two review authors (RM and HB) then reviewed these full‐text articles to determine eligibility. We resolved any disagreement with discussion and consensus. We included a Preferred Reporting Items for Systematic Reviews and Meta‐Analyses (PRISMA) study flow diagram in the review to document the screening process (Liberati 2009), as recommended in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011).

Data extraction and management

Two review authors (RM and HB) independently extracted data from the included studies. Where appropriate, we imported data and pooled them in Cochrane's statistical software, Review Manager (RevMan) (Review Manager 2014), for further analysis.

We used a data collection form for study characteristics and outcome data, which we piloted on one study included in the review.

We extracted the following data.

Methods: study design, duration of the study, study setting, and date of study.
Participants: number, mean age and age range, gender, inclusion and exclusion criteria.
Intervention: intervention, dose, mode of administration, concomitant medications, and exclusions.
Outcomes: primary and secondary outcomes as specified, type of scale used, and time points collected.
Notes: funding for trial and any conflicts of interest for trial authors.
'Risk of bias' summary.

We extracted the mean and standard deviation (SD) values from each study. Where the included studies reported standard error or confidence intervals (CIs) were reported, we converted these to SD values according to the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011).

Assessment of risk of bias in included studies

Two independent authors (HB and RM) assessed the included studies for risk of bias using the Cochrane's 'Risk of bias' assessment tool (Higgins 2011). We assessed the following: allocation (random sequence generation and allocation concealment); blinding of participants and personnel, blinding of outcome assessors; incomplete outcome data; and other bias. We scored each of these domains separately as either low risk of bias, unclear risk of bias (insufficient information to make a judgement), or high risk of bias as outlined below.

Generation of allocation sequence:
- for each included study we described the method used to generate the allocation sequence in sufficient detail to allow an assessment of whether it should produce comparable groups;
- we assessed the method used to generate the allocation sequence as either: low risk of bias (any truly random process such as random number table or computer random number generator); or unclear risk of bias (method used to generate sequence not clearly stated). We excluded studies that used a non‐random process (e.g. odd or even date of birth; hospital or clinic record number).
Allocation concealment:
- for each included study we described the method used to conceal the allocation sequence in sufficient detail and to determine whether intervention allocation could have been foreseen in advance of, or during recruitment, or changed after assignment;
- we assessed the methods as either: low risk of bias (e.g. telephone or central randomisation; consecutively numbered sealed opaque envelopes); or unclear risk of bias (method not clearly stated). We excluded studies that did not conceal allocation (e.g. open list).
Blinding or masking (checking for possible performance bias):
- for each included study we described the methods used, if any, to blind study participants and personnel from knowledge of which intervention a participant received. We judged studies at low risk of bias if they were blinded, or if we judged that the lack of blinding could not have affected the results. We assessed blinding separately for different outcomes or classes of outcomes;
- we assessed the methods used to blind study participants and outcome assessors from knowledge of which intervention a participant received. We assessed the methods as either: low risk of bias (study stated that it was blinded and described the method used to achieve blinding, e.g. identical tablets; matched in appearance and smell); or unclear risk of bias (study stated that it was blinded but did not provide an adequate description of how it was achieved). We intended to exclude studies that were at high risk of bias and were not double‐blinded.
Incomplete outcome data (checking for possible attrition bias through withdrawals, dropouts, protocol deviations). We assessed the methods used to deal with incomplete data as either: low risk (information from all participants were included in the main results, any dropouts are reported, any systematic differences between the two treatment arms are reported); unclear risk of bias (used 'last observation carried forward' analysis); or high risk of bias (used 'completer' analysis).

Selective reporting bias (checking for within study reporting bias, checking for possible attrition bias due to the amount, nature, and handling of incomplete outcome data). We assessed the methods as either low risk of bias (whether the study fully reported all prespecified outcomes); unclear risk of bias (it appeared not all pre‐outcomes were fully reported); or high risk of bias (the study highlighted not all prespecified outcomes were reported).

Size (checking for possible biases confounded by small size). Small studies have been shown to overestimate treatment effects, probably because the conduct of small studies is more likely to be less rigorous, allowing critical criteria to be compromised (Kjaergard 2001; Nüesch 2010; Dechartres 2013). We considered studies to be at low risk of bias if they had 200 participants or more in each treatment arm; at unclear risk of bias if they had 50 to 200 participants per treatment arm; or at high risk of bias if they had fewer than 50 participants per treatment arm.

Free of other bias (bias due to problems not covered elsewhere in the table):
- for each included study we described any important concerns we had about other possible sources of bias (e.g. baseline imbalance, bias of the presentation data, representation of gender, etc.).

We resolved any disagreement by discussion and consensus.

We performed funnel plot analysis and compared fixed‐effect versus random‐effects magnitude of effect to determine if there was any suggestion of bias.

Measures of treatment effect

We presented results from continuous variables, such as the breathlessness scales, using a fixed‐effect model and calculated standardised mean differences (SMDs) where scales were combined, such as when pooling VAS and Borg scale, with the corresponding 95% CIs. Where scales were not combined, and to assess effect across subgroups, we used the mean difference (MD). Where studies reported results based on a variable range of doses, we used the higher dose.

For dichotomous data, including adverse events, we reported relative risk ratios (RRs) where we could pool the data. Where we were unable to pool these data, we included these results in a descriptive analysis.

Unit of analysis issues

Our unit of analysis was the participant. We did not identify any cluster RCTs. We took measurements from the intervention and control group, and analysed the data as if it was a parallel trial, due to lack of paired data available. We attempted to contact the study authors to obtain paired data. However, due to no response we were unable to obtain original data. Most included studies were crossover trials, and therefore we included the data available in data reports, and acknowledged the limitations of this approach.

Dealing with missing data

Where possible we attempted to contact the principal investigator of the included studies to obtain missing data.

Assessment of heterogeneity

For pooled analyses, we quantified statistical heterogeneity using the I² statistic, which describes the percentage of the total variation across trials due to heterogeneity rather than sampling error. We considered significant statistical heterogeneity to be present if the I² statistic value was greater than 50%.

Where we identified significant heterogeneity, we further assessed this using predetermined subgroups.

Assessment of reporting biases

We attempted to contact the principal investigator of the included study for missing data where reporting bias appeared possible.

Data synthesis

A priori, we decided to analyse continuous data according to a fixed‐effect model, due to the concerns around the small‐study effects on the results of the meta‐analysis for all continuous outcomes (Higgins 2011). However, we provided random‐effects model data in the sensitivity analysis to compare the sensitivity of the results to different statistical methods. We calculated SMDs where we combined scales, such as when we pooled the VAS and Borg scale data, with the corresponding 95% CIs. Where scales were not combined, and to assess effect across subgroups, we used the MD. Where studies reported results based on a variable range of doses, we used the higher dose.

We used RevMan (Review Manager 2014) to perform meta‐analyses and presented our primary outcomes in a 'Summary of findings' table, as recommended in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011) using the GRADEPro Guideline Development Tool (GDT) software (GradePro 2015).

We assessed the overall quality of the evidence for each outcome using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) system (GradePro 2015) and presented the main findings of the review in a transparent and simple tabular format in the 'Summary of findings' tables. In particular, we included key information concerning the quality of evidence, the magnitude of effect of the interventions examined, and the sum of available data on the main outcomes. We chose to present the outcomes stipulated a priori and that which would be clinically meaningful.

The GRADE system uses the following criteria for assigning the grade of evidence based on RCTs.

High: further research is very unlikely to change our confidence in the estimate of effect.
Moderate: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate.
Low: further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate.
Very low: any estimate of effect is very uncertain.

We decreased the grade of evidence if the following occurred.

Serious (−1) or very serious (−2) limitation to study quality.
Important inconsistency (−1).
Some (−1) or major (−2) uncertainty about directness.
Imprecise or sparse data (−1).
High probability of reporting bias (−1).

Subgroup analysis and investigation of heterogeneity

We performed the following subgroup analyses.

Type of illness (e.g. chronic obstructive pulmonary disease (COPD), heart failure, malignancy, and neuromuscular disorders).
Mode of delivery of opioid drug (e.g. oral, subcutaneous, intravenous, nebulised, intra‐nasal, sublingual, buccal, transdermal, and other modes).
Dose.
Type of opioid (e.g. morphine, dihydrocodeine, fentanyl).

In the protocol we indicated that we would perform meta‐analyses according to the subgroups of dose and 'Risk of bias' assessment. Due to the wide variation and heterogeneity of reported doses we chose to analyse this in a descriptive analysis. We compared the 'Risk of bias' difference in a sensitivity analysis.

Post‐hoc we chose to include the type of opioid as a subgroup analysis as we felt this would be an important assessment for clinicians and policy makers.

Sensitivity analysis

We performed sensitivity analyses by systematically excluding studies from the overall analysis based on the potential sources of heterogeneity outlined above, and if homogeneous subgroups have not already been identified and analysed separately. We also compared data from fixed‐effect and random‐effects models to assess for heterogeneity.

Results

Description of studies

Results of the search

We identified 376 citations by using the search strategy, and selected 50 articles for full‐text review after screening the abstracts of the initial search results. See Figure 1 for further details.

We included 26 studies with 526 participants in the review.

Included studies

See the 'Characteristics of included studies' table.

Study characteristics

Eighteen studies with 276 participants provided data for the primary outcome of breathlessness and were included in the meta‐analysis (Abernethy 2003; Bar‐Or 1982; Bruera 1993; Charles 2008; Chua 1997; Eiser 1991; Harris‐Eze 1995; Hui 2014; Jankleson 1997; Jensen 2012; Johnson 1983; Leung 1996; Light 1996; Mazzocato 1999; Noseda 1997; Oxberry 2011; Poole 1998; Woodcock 1981). Four additional studies examined the primary outcome of breathlessness but we were unable to extrapolate data for meta‐analysis (Davis 1996; Grimbert 2004; Johnson 2002; Masood 1995). Two additional studies did not report the primary outcome, but reported secondary outcomes (Williams 2003; Young 1989). Two additional studies compared opioids to an intervention other than placebo; Navigante 2010 compared morphine to midazolam, Rice 1987 compared codeine to promethazine. One study (Oxberry 2011) compared morphine with oxycodone and placebo.

Twenty‐four included studies were crossover trials. Hui 2014 was a parallel group RCT that compared subcutaneous fentanyl with placebo, and Navigante 2010 was a parallel RCT that compared morphine to midazolam.

Most included studies were performed over a fixed period during one day or on two consecutive days, with a washout period of only on day. Six studies involved more chronic administration of the drug or placebo, continuing for study periods between four days and six weeks, with a washout period of between three days and two weeks (Woodcock 1981; Johnson 1983; Eiser 1991; Poole 1998; Abernethy 2003; Navigante 2010).

Participants

All included studies were small, with fewer than 50 participants per treatment arm. The number ranged from six to 25 participants, with an average of 19 participants per study. Fourteen studies recruited ambulatory care participants (Abernethy 2003; Eiser 1991; Hui 2014; Masood 1995; Harris‐Eze 1995; Rice 1987; Poole 1998; Oxberry 2011; Navigante 2010; Leung 1996; Johnson 1983; Johnson 2002; Young 1989; Woodcock 1981), two studies recruited inpatients (Mazzocato 1999; Noseda 1997), one study had a mix of inpatients and outpatients (Charles 2008), and nine studies did not specify the participant setting (Light 1996; Williams 2003; Jensen 2012; Jankleson 1997; Grimbert 2004; Chua 1997; Bruera 1993; Bar‐Or 1982; Davis 1996).

Fourteen studies involved primarily or exclusively participants with chronic obstructive pulmonary disease (COPD) (Abernethy 2003; Bar‐Or 1982; Light 1996; Poole 1998; Eiser 1991; Jankleson 1997; Jensen 2012; Johnson 1983; Leung 1996; Masood 1995; Noseda 1997; Rice 1987; Woodcock 1981; Young 1989). Six studies included only participants with malignant disease (Bruera 1993; Charles 2008; Davis 1996; Grimbert 2004; Hui 2014; Mazzocato 1999), four studies were comprised primarily of cardiac failure participants (Chua 1997; Johnson 2002; Oxberry 2011; Williams 2003), and one study was comprised of participants with interstitial lung disease (Harris‐Eze 1995). One study involved end stage disease from all causes (Navigante 2010).

Intervention

Eight studies specifically recruited participants not currently on opioids (Abernethy 2003; Harris‐Eze 1995; Jensen 2012; Johnson 1983; Masood 1995; Navigante 2010; Poole 1998; Rice 1987), who were thus opioid naive. Twelve studies did not specify whether opioid use was part of the exclusion criteria, or whether it formed part of the co‐interventions (Bar‐Or 1982; Light 1996; Eiser 1991; Jankleson 1997; Noseda 1997; Woodcock 1981; Young 1989; Davis 1996; Chua 1997; Oxberry 2011; Leung 1996; Williams 2003). Three studies examined participants already on opioids (Bruera 1993; Grimbert 2004; Hui 2014). Bruera 1993 used 50% more of the participant's usual dose in a PRN (pro re nata, or as required) manner. Hui 2014 used a sliding scale of 30 mcg to 350 mcg fentanyl for all interventional participants, and included regular opioids in both the interventional and control arm. Three studies (Charles 2008; Mazzocato 1999; Grimbert 2004) used a predefined dose of opioids, regardless of the participant's current opioid use.

The included studies used the following opioids: oral dihydrocodeine (Bar‐Or 1982; Chua 1997; Johnson 1983; Rice 1987), oral diamorphine (Eiser 1991), intravenous diamorphine (Williams 2003), oral morphine (Light 1996; Mazzocato 1999; Poole 1998; Abernethy 2003; Woodcock 1981), nebulised morphine (Davis 1996; Charles 2008; Grimbert 2004; Harris‐Eze 1995; Leung 1996; Jankleson 1997; Masood 1995; Noseda 1997; Young 1989), subcutaneous fentanyl (Navigante 2010; Hui 2014), subcutaneous morphine (Bruera 1993), nebulised fentanyl (Jensen 2012), oral oxycodone (Oxberry 2011), and hydromorphone (Charles 2008).

The doses of dihydrocodeine ranged from 15 mg three times a day to 60 mg three times a day in 1 mg/1 kg doses. The diamorphine dose ranged from 2.5 to 5 mg four times a day. Sustained release morphine was used in 10 to 20 mg doses. Oxycodone was administered in 2.5 mg doses four times a day. Subcutaneous morphine doses ranged from 2.5 to 10 mg. There was a wide range of nebulised morphine doses used, from 1 mg to 50 mg.

Nine studies delivered the opioids by the oral route (Bar‐Or 1982; Eiser 1991; Johnson 1983; Woodcock 1981; Abernethy 2003; Chua 1997; Light 1996; Oxberry 2011; Poole 1998), two studies used parenteral opioids (Bruera 1993; Hui 2014), and ten studies gave the drugs via nebulisation (Davis 1996; Harris‐Eze 1995; Masood 1995; Young 1989; Grimbert 2004; Jankleson 1997; Leung 1996; Noseda 1997; Charles 2008; Jensen 2012). Some studies compared different routes of administration.

Eight studies continued regular use of co‐interventions including steroids and bronchodilators (Bruera 1993; Masood 1995; Woodcock 1981; Young 1989; Charles 2008; Hui 2014; Mazzocato 1999; Rice 1987). Two studies involved the use of oxygen inhalation (Leung 1996; Noseda 1997). In both cases the measures were applied to the use of the drug and placebo arm and we felt this did not bias the study results.

Outcomes

Twelve studies performed some form of exercise testing (Bar‐Or 1982; Hui 2014; Poole 1998; Chua 1997; Harris‐Eze 1995; Leung 1996; Light 1996; Eiser 1991; Jensen 2012; Johnson 1983; Williams 2003; Woodcock 1981). They used a variety of different exercise tests, including incremental treadmill tests, incremental cycle ergometer tests, non incremental treadmill or endurance treadmill tests, and six‐minute walk tests (6MWT).

There was significant variety in the reporting of breathlessness outcome measure, but all studies used well‐validated scales, including the visual analogue scale (VAS), Borg Scale, and oxygen cost diagram (McGavin 1978; O'Donnell 1998). Several studies did not report breathlessness at a fixed point during exercise (Beauford 1993; Masood 1995). Some studies did not report the primary outcome of breathlessness, did not include sufficient data, did not report standard deviations (SDs) or error, or reported data in such a way that the relevant numbers could not be extrapolated (Young 1989; Davis 1996; Jankleson 1997; Masood 1995; Williams 2003; Grimbert 2004).

In most cases, the studies asked their participants to assess their own levels of breathlessness, by VAS or Borg scale. Some studies asked participants to guess which substance contained the opioid or placebo drug, and other studies offered participants the opportunity to continue on opioid therapy. One study, Poole 1998, used the Chronic Respiratory Disease Questionnaire (CRQ) dyspnoea scale.

Excluded studies

See the 'Characteristics of excluded studies' section.

We excluded eight studies for the following reasons: participants were not randomised (Beauford 1993; Peterson 1996; Shorati 2012; Smith 2009), there was no comparison to a placebo or other intervention (Allard 1999; Bruera 2005; Navigante 2003), or it was a review (Thomas 2010).

Ongoing studies

We identified two ongoing studies (Cuervo Pinna 2012; Daubert 2014).

Risk of bias in included studies

We assessed the risk of bias in the included studies using the Cochrane 'Risk of bias' assessment tool (Higgins 2011), and included the domains of allocation, blinding, incomplete outcome data, and other bias. We judged eight studies to be at an overall low risk of bias. We considered 18 studies to be at an overall unclear risk of bias, that is we had insufficient information to make a judgement, usually due to inadequate descriptions of the methods of randomisation or blinding.

Please see Figure 2 and Figure 3 for a summary of the 'Risk of bias' findings.

We assessed the overall quality of the evidence for each outcome using the GRADE system (GradePro 2015) and presented these results in the 'Summary of findings' tables, which shows the main findings of the review in a transparent and simple tabular format. In particular, we included key information concerning the quality of evidence, the magnitude of effect of the interventions examined, and the sum of available data on the main outcomes.

Allocation

We assessed random sequence generation as adequate (low risk) in eight out of 26 studies (Poole 1998; Johnson 2002; Abernethy 2003; Charles 2008; Navigante 2010; Oxberry 2011; Jensen 2012; Hui 2014). Most studies (18 studies) did not describe the methods of sequence generation (unclear risk of bias).

We judged allocation concealment as adequate (low risk) in eight out of 26 studies (Poole 1998; Abernethy 2003; Grimbert 2004; Charles 2008; Navigante 2010; Oxberry 2011; Jensen 2012; Hui 2014), which suggests that information from most of the studies presented an unclear risk of bias. Many studies did not state the method of allocation concealment, though many of the studies reported their design as randomised.

We did not judge any studies as at high risk of allocation or random sequence generation bias.

Blinding

Blinding of participants and personnel with respect to the intervention was adequate, though blinding of the outcome assessment was overall poor.

We judged the blinding of participants and personnel to be adequate in 10 out of 26 studies, indicating low risk of bias (Johnson 1983; Harris‐Eze 1995; Light 1996; Poole 1998; Abernethy 2003;Grimbert 2004; Charles 2008; Oxberry 2011; Jensen 2012; Hui 2014). These studies used placebo interventions, which were reported to have been designed to appear the same as the opioid intervention. We judged blinding of participants and personnel to be at high risk of bias in Navigante 2010 because only the participants were blinded, not the investigators or those that performed the outcome assessment. Of the 15 studies that we assessed as being at an unclear risk of bias for this domain, the studies did not specifically or adequately describe the details to which the intervention and control were blinded, though many studies reported themselves as blinded.

Overall, blinding of the outcome assessment was poor. We assessed only four out of 26 studies as at low risk of bias (Abernethy 2003; Charles 2008; Oxberry 2011; Hui 2014). We judged Navigante 2010 to be at high risk as those performing the outcome assessment were not blinded, and we judged the remaining 21 studies to be at an unclear risk of bias. Most studies did not clearly describe the methods by which the outcome assessment was blinded, though some described themselves as double blinded. This may be in part due to the primary outcome of breathlessness requiring the participant to score their own symptoms.

Incomplete outcome data

The included studies generally reported data completely, with 21 out of 26 studies adequately described. We therefore judged them to be at low risk of bias. We judged the remaining five studies to be at an unclear risk of bias (Bruera 1993; Chua 1997; Davis 1996; Light 1996; Young 1989). The included studies usually recorded adverse events, but generally these did not cause participants to drop out of the study. Most studies were conducted on consecutive days, so loss to follow‐up was less likely to occur.

Selective reporting

We judged the risk of selective reporting to be low in all studies. We did not detect any evidence of selective reporting bias.

Other potential sources of bias

We judged 20 studies to be at low risk for this domain (Woodcock 1981; Bar‐Or 1982; Rice 1987; Young 1989; Masood 1995; Leung 1996; Light 1996; Noseda 1997; Poole 1998; Mazzocato 1999; Johnson 2002; Abernethy 2003; Williams 2003; Grimbert 2004; Charles 2008; Navigante 2010; Oxberry 2011; Jensen 2012; Hui 2014), and one at high risk of bias because it did not state that it systematically studied adverse events (Eiser 1991). We judged the remaining five studies to be at unclear risk of other bias because insufficient information was available.

Size

The studies were of small sample size, with a mean of 19 participants per study, and with fewer than 50 participants per treatment arm. Thus we judged all 26 studies to be at overall high risk of bias for this domain.

Effects of interventions

See: Table 1; Table 2; Table 3

for the main comparison.

Opioids compared with placebo in people with breathlessness in advanced disease or terminal illness
Patient or population: adults with refractory breathlessness Setting: inpatient and outpatient setting Intervention: opioids Comparison: placebo
Outcomes	*Illustrative comparative risks (95% CI)**		Relative effect  (95% CI)	Number of participants  (studies)	Quality of the evidence  (GRADE)	Comments
	Assumed risk	Corresponding risk
	Placebo	Opioids
Breathlessness: change from baseline¹	The standardised mean change from baseline ranged from −2 to 0.60 in the control group	The standardised mean change from baseline was 0.11 points better in the opioids group compared to the placebo group (ranging from a 0.40 point reduction to a 0.19 point increase in breathlessness)	—	194  (6 RCTs)	⊕⊝⊝⊝ very low^2,3,4	A lower score indicates an improvement in breathlessness
Breathlessness: post‐treatment score¹	The standardised mean post‐treatment score ranged from −43 to 49 in the control group	The standardised mean post‐treatment score was 0.32 points better in the opioid group compared to the placebo group (ranging from a 0.53 point reduction to a 0.10 point reduction in breathlessness)	—	338  (12 RCTs)	⊕⊕⊝⊝ low^2,3	A lower score indicates an improvement in breathlessness
Exercise tolerance: 6MWT⁵ ‐ total distance	The total distance in 6MWT was 368m in the placebo group	The total distance in 6MWT was 28 m better in the opioids group compared to placebo (ranging from 113 m to 58 m)	—	11 participants  (1 RCT)	⊕⊝⊝⊝ very low^2,3,4,6	—
Exercise tolerance: 6MWT⁵‐ change from baseline	The change from baseline was from ‐21m to 37m in the placebo group	The change in baseline was 48 m worse in the opioids group (ranging from 36 m to 60 m)	—	26  (2 RCTs)	⊕⊝⊝⊝ very low^1,2,3	—
Adverse events: constipation	55 per 1000	179 per 1000	RR 3 (95% CI 1.63 to 5.51)	162 (9 RCTs)	⊕⊝⊝⊝ very low^2,3	—
Adverse events: nausea and vomiting	67 per 1000	201 per 1000	RR 4.73 (95% CI 1.73 to 12.97)	104 (7 RCTs)	⊕⊝⊝⊝ very low^2,3	—
Adverse events: drowsiness	58 per 1000	128 per 1000	RR 2.86 (95% CI 1.17 to 7.02)	156 (9 RCTs)	⊕⊝⊝⊝ very low^2,3	—
Quality of life⁷	The change from baseline score in the control group was 2.94	The quality of life change from baseline score in the opioid group was 0.86 points lower (ranging from 9.90 points lower to 8.18 points higher)	—	16 (1 RCT)	⊕⊝⊝⊝ very low^2,3,4	—
The basis for the assumed risk* (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).
GRADE Working Group grades of evidence  High quality: further research is very unlikely to change our confidence in the estimate of effect.  Moderate quality: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate.  Low quality: further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate.  Very low quality: we are very uncertain about the estimate.

Morphine compared with midazolam in people with breathlessness in advanced disease or terminal illness
Patient or population: adults with refractory breathlessness Setting: outpatient setting Intervention: morphine Comparison: midazolam
Outcomes	*Illustrative comparative risks (95% CI)**		Relative effect  (95% CI)	Number of participants  (studies)	Quality of the evidence  (GRADE)	Comments
	Assumed risk	Corresponding risk
	Midazolam	Morphine
Breathlessness: post‐treatment score¹	The mean dyspnoea score in the midazolam group was 4	The mean post‐treatment score was 2 points higher in the opioids group (ranging from 1.07 to 2.93)	—	63  (1 RCTs)	⊕⊝⊝⊝  very low^2,3,4	A lower score indicates an improvement in breathlessness
The basis for the assumed risk* (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).
GRADE Working Group grades of evidence  High quality: further research is very unlikely to change our confidence in the estimate of effect.  Moderate quality: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate.  Low quality: further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate.  Very low quality: we are very uncertain about the estimate.

Codeine compared with promethazine in people with breathlessness in advanced disease or terminal illness
Patient or population: adults with refractory breathlessness Setting: outpatient setting Intervention: codeine Comparison: promethazine
Outcomes	*Illustrative comparative risks (95% CI)**		Relative effect  (95% CI)	Number of Participants  (studies)	Quality of the evidence  (GRADE)	Comments
	Assumed risk	Corresponding risk
	Promethazine	Codeine
Breathlessness: post‐treatment score¹	The mean dyspnoea score in the promethazine group was 6	The mean post‐treatment score was 0.30 points lower in the codeine group (ranging from 0.83 points lower to 0.23 points higher)	—	7  (1 RCT)	⊕⊝⊝⊝  very low^2,3,4	A lower score indicates an improvement in breathlessness
The basis for the assumed risk* (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).
GRADE Working Group grades of evidence  High quality: Further research is very unlikely to change our confidence in the estimate of effect.  Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate.  Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate.  Very low quality: We are very uncertain about the estimate.

Meta‐analysis	Number of studies	Pooled SMD	Confidence interval	P value for SMD	Heterogeneity test
All studies, fixed‐effect Change from baseline	6	‐0.11	−0.40 to 0.19	0.47	I² statistic = 76%, P = 0.0009
All studies, random‐effects Change from baseline	6	‐0.40	−1.04 to 0.24	0.22	I² statistic = 76%, P = 0.0009
All studies, fixed‐effect Post‐treatment score	12	−0.32	−0.53 to −0.10	0.004	I² statistic = 0%, P = 0.88
All studies, random‐effects Post‐treatment score	12	−0.32	−0.53 to −0.10	0.004	I² statistic = 0%, P = 0.88
Studies excluded with unclear bias Change from baseline	5	0.01	−0.29 to 0.31	0.95	I² statistic = 80%, P = 0.0006
Studies excluded with unclear bias Post‐treatment score	5	0.20	−0.50 to 0.10	0.20	I² statistic = 0%, P = 0.76

Date	Event	Description
14 May 2019	Amended	Final amendments to analyses.
1 May 2019	Amended	We addressed errors in data extraction and updated our meta‐analyses accordingly.
8 February 2018	Review declared as stable	See Published notes
23 October 2017	Amended	Minor change to Feedback
21 August 2017	Feedback has been incorporated	See Feedback.
1 August 2016	Amended	Minor changes to text, e.g. in Abstract, Results, and Summary of findings table.
8 June 2016	Amended	Abstract results amended

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Breathlessness	18		Std. Mean Difference (IV, Fixed, 95% CI)	Subtotals only
1.1 Change from baseline	6	194	Std. Mean Difference (IV, Fixed, 95% CI)	‐0.11 [‐0.40, 0.19]
1.2 Post‐treatment score	12	338	Std. Mean Difference (IV, Fixed, 95% CI)	‐0.32 [‐0.53, ‐0.10]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Breathlessness	3		Mean Difference (IV, Fixed, 95% CI)	Subtotals only
1.1 Hydromorphone: change from baseline	1	40	Mean Difference (IV, Fixed, 95% CI)	‐0.26 [‐1.17, 0.65]
1.2 Oral diamorphine: post‐treatment score	1	20	Mean Difference (IV, Fixed, 95% CI)	0.5 [‐1.44, 2.44]
1.3 Oxycodone: change from baseline	1	70	Mean Difference (IV, Fixed, 95% CI)	0.08 [‐0.87, 1.03]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Exercise tolerance	12		Mean Difference (IV, Fixed, 95% CI)	Subtotals only
1.1 6MWT: total distance	1	22	Mean Difference (IV, Fixed, 95% CI)	‐28.0 [‐113.58, 57.58]
1.2 6MWT: change from baseline	2	52	Mean Difference (IV, Fixed, 95% CI)	47.78 [35.88, 59.67]
1.3 Workload (watts)	4	70	Mean Difference (IV, Fixed, 95% CI)	‐1.70 [‐3.29, ‐0.10]
1.4 Treadmill distance	2	60	Mean Difference (IV, Fixed, 95% CI)	‐37.64 [‐96.17, 20.88]
1.5 Time on treadmill	3	76	Mean Difference (IV, Fixed, 95% CI)	‐13.42 [‐42.55, 15.70]
1.6 Maximum work cycle	1	24	Mean Difference (IV, Fixed, 95% CI)	‐1.30 [‐1.81, ‐0.79]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Adverse events: constipation	9	324	Risk Ratio (M‐H, Fixed, 95% CI)	3.0 [1.63, 5.51]
2 Adverse events: nausea and vomiting	7	208	Odds Ratio (M‐H, Fixed, 95% CI)	4.73 [1.73, 12.97]
3 Adverse events: drowsiness	9	312	Odds Ratio (M‐H, Fixed, 95% CI)	2.86 [1.17, 7.02]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Breathlessness	15	460	Std. Mean Difference (Fixed, 95% CI)	‐0.42 [‐0.58, ‐0.26]
1.1 VAS scale	7	244	Std. Mean Difference (Fixed, 95% CI)	‐0.45 [‐0.65, ‐0.24]
1.2 BORG scale	5	146	Std. Mean Difference (Fixed, 95% CI)	‐0.36 [‐0.69, ‐0.02]
1.3 Oxygen cost diagram	1	22	Std. Mean Difference (Fixed, 95% CI)	‐0.48 [‐1.33, 0.37]
1.4 NRS scale	1	20	Std. Mean Difference (Fixed, 95% CI)	‐0.07 [‐0.95, 0.80]
1.5 CRQ scale	1	28	Std. Mean Difference (Fixed, 95% CI)	‐0.56 [‐1.32, 0.20]

Methods	Randomised, double blind crossover study
Participants	Opioid naive adults with dyspnoea despite treatment for reversible factors for end stage respiratory, cardiac or palliative conditions 48 randomised participants Mean age: 76 years 35/48 male Exclusion criteria were recent use of opioids, confusion, obtundation, adverse reactions to opioids, and history of substance misuse
Interventions	20 mg oral sustained release morphine sulphate or placebo tablet Co‐interventions included coloxyl and senna
Outcomes	VAS dyspnoea scale Respiratory rate Sedation/obtundation Conspitation
Notes	Outcome measurement at day 4 and day 8 The study authors concluded that sustained release oral morphine at low dosage provides significant symptomatic improvement in refractory dyspnoea in the community setting Dose is equivalent to 20 mg oral morphine
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomisation and blinding were co‐ordinated through the hospital pharmacy’s centralised service. This included computerised generation of the allocation sequence in random permuted blocks and blinded disbursement of medication.
Allocation concealment (selection bias)	Low risk	Randomisation and blinding were coordinated through the hospital pharmacy’s centralised service. This included computerised generation of the allocation sequence in random permuted blocks and blinded disbursement of medication.
Blinding of participants and personnel (performance bias)   All outcomes	Low risk	Double blinded. The placebo medication was identical in appearance and taste to the active medication; the bottle indicated which medication to take each day. Participants were unblinded to the investigators for serious adverse events only. There was no blinding for constipation. To accommodate this, the only investigator aware of the constipation was the study nurse (AM), who was not involved in the analysis.
Blinding of outcome assessment (detection bias)   All outcomes	Low risk	The placebo medication was identical in appearance and taste to the active medication; the bottle indicated which medication to take each day. Participants were unblinded to the investigators for serious adverse events only.
Incomplete outcome data (attrition bias)   All outcomes	Low risk	The study assessed the primary outcome dyspnoea at the end of 4^th day of the treatment period only. There were 10 withdrawals during treatment, with some due to adverse events.
Selective reporting (reporting bias)	Low risk	The study collected statistical details about some adverse events and some outcomes, such as overall well being and MRC Exercise tolerance scores, but did not report these in detail in the study report except to say there were no significant differences.
Other bias	Low risk	We judged that this trial appeared to be free of other sources of bias.
Size bias	High risk	There were < 50 participants per treatment arm.

Methods	Randomised crossover study
Participants	Participants with chronic obstructive pulmonary disease (COPD), otherwise not specifically stated 11 participants
Interventions	30 mg three times daily (TDS) or 60 mg TDS oral dihydrocodeine Compared to oral placebo
Outcomes	Used oxygen cost diagram to measure breathlessness Six‐minute walking test (6MWT) Oxygen consumption Adverse events
Notes	Abstract, so limited information provided They concluded that there was a marked improvement in subjective disability when 30 mg TDS was given, but not 60 mg TDS Dose is equivalent to 18 mg oral morphine
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	The study authors did not clearly state the method of random sequence generation.
Allocation concealment (selection bias)	Unclear risk	The study authors did not clearly state the methods of allocation concealment.
Blinding of participants and personnel (performance bias)   All outcomes	Unclear risk	The study authors did not clearly state the methods of blinding of participants and personnel.
Blinding of outcome assessment (detection bias)   All outcomes	Unclear risk	The study authors did not clearly state the methods of outcome assessment.
Incomplete outcome data (attrition bias)   All outcomes	Low risk	There were 16 randomised participants and 5 withdrawals due to adverse effects, all discussed but not likely related to the primary outcome measure.
Selective reporting (reporting bias)	Low risk	We did not detect any selective reporting bias.
Other bias	Low risk	We judged that this trial appeared to be free of other sources of bias.
Size bias	High risk	There was a small sample size, and this study is likely to be at high risk of bias; there were < 50 participants per arm.

Methods	Double blind, randomised crossover study
Participants	Those with cancer experiencing dyspnoea English speaking, over 18 years, expected prognosis of 7 days, Minimental state exam (MMSE) > 24/30, experiencing dyspnoea with no reversible cause 20 participants Mean age: 69 years 11 males, 9 females
Interventions	5 mg nebulised hydromorphone compared to 3 mL nebulised saline Co‐interventions were recorded but not stated
Outcomes	VAS dyspnoea scale Respiratory rate Pulse rate Oxygen saturation
Notes	The study measured outcomes at 10, 20, 30 and 60 minutes post‐dose The study authors concluded that there were no significant differences between treatments
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	The study randomised treatment order using a random number generator. To ensure double blinding, a non‐clinical research doctor prepared treatments composed of medications (commercially obtained) and blinding agents and checked with a non‐clinical nurse, neither of whom was involved with the care of the participant. A research nurse, who was unaware of the order sequence and who was not involved with the clinical care of the participant, subsequently administered pre‐prepared and randomised treatments. The non‐clinical research doctor held a master plan of the randomizations so that treatments could be unblinded in an emergency. There were no such emergencies over the course of the study.
Allocation concealment (selection bias)	Low risk	This was probably ok as the doctor responsible for randomisation was not involved in the care of the participants.
Blinding of participants and personnel (performance bias)   All outcomes	Low risk	Double blinded. Oral medications were administered in orange juice. Blinding as above seems reasonable.
Blinding of outcome assessment (detection bias)   All outcomes	Low risk	As above. The VAS was self administered.
Incomplete outcome data (attrition bias)   All outcomes	Low risk	There were 5 withdrawals. The study authors did not describe the reasons per initial group allocation but for the whole group. They stated that the reasons were not systematically related to treatments or order of treatments.
Selective reporting (reporting bias)	Low risk	There was no evidence of reporting bias.
Other bias	Low risk	We judged that this trial appeared to be free of other sources of bias.
Size bias	High risk	There was size bias as there were only 20 participants in total, and < 50 per treatment arm.

Methods	Randomised crossover study Exercise study Incremental treadmill test (modified Bruce protocol)
Participants	Stable chronic heart failure 12 male participants Mean age 65.5 years, (range 58 to 75 years) Average left ventricular ejection fraction (LVEF) 21.3% (range 8% to 39%) New York Heart Association (NYHA) functional class 2 and 3 No chest pain or inducible ischaemia during previous exercise testing No history pulmonary disease All patients on diuretics and angiotensin‐converting‐enzyme (ACE) inhibitors
Interventions	Dihydrocodeine 1 mg/kg vs placebo Tests took place on separate days
Outcomes	Modified Borg score (dyspnoea) Pulse Systolic blood pressure (BP) End‐tidal CO₂ concentration % Partial pressure of oxygen (PaO₂) Modified Borg score (fatigue) Hypoxic chemosensitivity Hypercapnic chemosensitivity Peak O₂ consumption Ventilation to carbon dioxide output (VE‐VCO₂ slope) Exercise duration
Notes	The study measured outcomes at 3 minutes, 6 minutes, and at peak exercise The study authors concluded that dihydrocodeine was associated with a reduction of exercise ventilation, an improvement in exercise tolerance and a decrease in breathlessness The dose is equivalent to an average of 6 mg oral morphine
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	The study assessed hypoxic and hypercapnic chemosensitivities in these participants 1 hour after the participants received placebo or dihydrocodeine (1 mg/kg body weight) in a randomised, double‐blind design on 2 separate days, followed by treadmill exercise testing on each occasion. The study gave placebo and dihydrocodeine in the form of a drink made up to 200 mL with bitter lemon, which was prepared by the Department of Pharmacy, Royal Brompton Hospital.
Allocation concealment (selection bias)	Unclear risk	There were no details other than the above.
Blinding of participants and personnel (performance bias)   All outcomes	Unclear risk	It was unclear whether or not the participants could tell the difference between the 2 drinks from description above.
Blinding of outcome assessment (detection bias)   All outcomes	Unclear risk	The study authors did not state the methods of blinding of outcome assessment.
Incomplete outcome data (attrition bias)   All outcomes	Unclear risk	The study authors did not state the number of withdrawals, dropouts, or protocol deviations.
Selective reporting (reporting bias)	Low risk	We did not detect any evidence of selective reporting.
Other bias	Unclear risk	It is unclear whether or not there was other risk due to the lack of detail in the methods.
Size bias	High risk	There was size bias as there were only 12 participants in total, and < 50 per treatment arm.

Methods	Randomised crossover study Three 2‐week periods followed by exercise tests with no wash‐out interval Exercise testing (6MWT and treadmill test) at end of each study period 4 withdrawals (1 due to chest infection, 1 because of itching on diamorphine, 1 due to constipation on diamorphine, and 1 due to headache due to cerebral metastases)
Participants	Those diagnosed with severe, stable COPD 14 participants 8 men; 6 women Mean age: 65 years Mean forced expiratory volume in one second (FEV1): 32% predicted Mean partial pressure of oxygen (paO₂) 9.0 range 7.1 to 10.9 kPa Mean partial pressure of carbon dioxide (paCO₂) 5.1 range 3.4 to 6.5 kPa
Interventions	Diamorphine 2.5 mg four times daily (QDS), or diamorphine 5 mg QDS or placebo
Outcomes	Daily diary cards with 10 cm VAS for dyspnoea, feeling of well‐being, drowsiness, number of bronchodilator puffs At the end of each 2 week period: FEV1 PaO₂ PaCO₂ A‐aPO₂ (alveolar‐arterial oxygen tension difference) 6MWT (six minute walk test) VAS dyspnoea for 6MWT Time on treadmill VAS dyspnoea for treadmill O₂ saturation End‐tidal PCO₂ Morphine levels
Notes	Dyspnoea assessed "at completion of each type of exercise" The study authors detected no significant effect of opioid compared with placebo The dose is equivalent to 15 mg oral morphine
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	The study was double‐blind, randomised, and cross‐over with no wash‐out intervals. The study authors did not provide any further details for the methods.
Allocation concealment (selection bias)	Unclear risk	The study authors did not state the method of allocation concealment.
Blinding of participants and personnel (performance bias)   All outcomes	Unclear risk	The study authors did not clearly state the methods of blinding of participants.
Blinding of outcome assessment (detection bias)   All outcomes	Unclear risk	The study authors did not clearly state the methods of blinding of outcome assessment.
Incomplete outcome data (attrition bias)   All outcomes	Low risk	In the first study there were 4 drop outs, 2 due to unrelated reasons and 2 due to side effects in the diamorphine group. However, this is unlikely to influence the results of the assessment of breathlessness.
Selective reporting (reporting bias)	Low risk	We did not detect any evidence of selective reporting.
Other bias	High risk	The study authors did not state that adverse events were systematically studied (apart from drowsiness).
Size bias	High risk	There were only 14 participants in total, and < 50 per treatment arm.

Methods	Double blinded, randomised parallel placebo controlled study 6MWT
Participants	People diagnosed with cancer, aged ≥ 18 years, with breakthrough dyspnoea > 3/10, ambulatory, Karnovsky score > 50%, on a stable opioid dose 20 participants Mean age 50 years (range 30 to 75 years) 11 female 70% Caucasian, 15% black, 15% Hispanic Exclusion criteria: dyspnoea > 7/10, supplemental oxygen > 6L, delirium, allergy to fentanyl, substance abuse, recent coronary disease
Interventions	Subcutaneous fentanyl, dosed on a sliding scale, dose ranging from 30 mcg to 350 mcg compared to subcutaneous saline Co‐interventions included regular opioids, bronchodilators, steroids
Outcomes	NRS dyspnoea scale Borg fatigue scale Heart rate Respiratory rate Oxygen saturation Blood pressure Adverse events
Notes	The study measured outcomes before and after the 6MWT The study authors concluded that prophylactic fentanyl was safe and improved dyspnoea, fatigue, walk distance, and respiratory rate, and that there was also a large placebo effect The dose is equivalent to 1.5 mg to 9 mg oral morphine
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer randomised sequence generation.
Allocation concealment (selection bias)	Low risk	Allocation was concealed by using a secured Web site that was only accessible to the study pharmacist after participant enrolment.
Blinding of participants and personnel (performance bias)   All outcomes	Low risk	Double blinded. Both the participants and research staff conducting the study assessments were blinded to the study intervention and the randomisation sequence.
Blinding of outcome assessment (detection bias)   All outcomes	Low risk	Both the participants and research staff conducting the study assessments were blinded to the study intervention and the randomisation sequence.
Incomplete outcome data (attrition bias)   All outcomes	Low risk	All participants completed the study with no loss of follow up
Selective reporting (reporting bias)	Low risk	There was no evidence of selective reporting.
Other bias	Low risk	We judged that this trial appeared to be free of other sources of bias.
Size bias	High risk	This was a parallel trial with 20 participants in each arm (< 50 participants per treatment arm).

Methods	Randomised, double blind, placebo controlled crossover study
Participants	Diagnosed with COPD, aged 40 years or older, cigarette smokers FEV1 < 0.7 12 participants Mean age 70 years (range ± 2.3 years) 7 males, 5 females Exclusion criteria: those with significant diseases other than COPD, those with sleep disordered breathing, those who has used opioids in the last 2 days
Interventions	Nebulised 50 mcg fentanyl citrate compared to nebulised saline Participants withdrew from beta‐agonists, anticholinergics, caffeine, and theophylline prior to the trial
Outcomes	Dyspnoea on the Borg scale Exercise time Dyspnoea unpleasantness Leg discomfort Heart rate Oxygen saturation VO₂ VCO₂ VE VT FR
Notes	The study authors measured outcomes at pre‐exercise, isotime (highest equivalent of exercise time achieved), and peak exercise They concluded that single‐dose inhalation of fentanyl citrate was associated with significant and potentially clinically important improvements in exercise tolerance in people with COPD
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	The hospital pharmacist performed randomisation, blinding, and dispensing of study medications. This person was an unblinded third party who was not affiliated with either subject recruitment or data collection and analysis.
Allocation concealment (selection bias)	Low risk	Double blinded. The hospital pharmacist performed randomisation, blinding, and dispensing of study medications. This person was an unblinded third party who was not affiliated with either subject recruitment or data collection and analysis.
Blinding of participants and personnel (performance bias)   All outcomes	Low risk	This is likely to have been adequate given both placebo in fentanyl solutions were the same volume and dispensing was independent of those involved in conducting the study. However, it is not entirely clear whether the placebo and drug had the same appearance/taste.
Blinding of outcome assessment (detection bias)   All outcomes	Unclear risk	The study authors did not clearly state the methods of blinding of outcome assessment.
Incomplete outcome data (attrition bias)   All outcomes	Low risk	Clearly reported. There were 3 drop outs in the placebo group: 2 were due to side effects (before exercise test) and 1 due to protocol violation. There was 1 drop out in intervention group due to side effects prior to exercise test. Balanced and unlikely to affect the results
Selective reporting (reporting bias)	Low risk	We did not find any evidence of selective reporting bias.
Other bias	Low risk	We judged that this trial appeared to be free of other sources of bias.
Size bias	High risk	There was a small sample size, and high risk of bias; there were < 50 participants per study arm.

Methods	Randomised, double blind, crossover study 2 consecutive 1 week periods followed by exercise test Incremental treadmill test 1 drop‐out: developed chest infection and right heart failure on dihydrocodeine
Participants	Those with stable COPD with severe breathlessness, and severe airflow obstruction 19 participants 15 men, 3 women Mean age 64.9 years, SD 9.1 years FEV1 830, SD 260 mL PaO₂ 9.3 SD 0.8 kPa; PaCO₂ 4.8 SD 0.5 kPa At least Grade 3 breathlessness (MRC scale) No recent hospital admissions No sedative drugs Continued usual bronchodilators and steroids
Interventions	Dihydrocodeine 15 mg or placebo over 2 consecutive 1 week periods. Drug to be taken 30 minutes before exercise up to 3 times daily Tests at end of each week period
Outcomes	Pedometer distance for 1 week Daily VAS for breathlessness PEFR FEV1 FVC Incremental treadmill test Distance walked VAS for breathlessness at 75% distance walked on placebo day
Notes	The study authors concluded that dihydrocodeine 15 mg 30 minutes before exercise offers appreciable benefit to participants with severe breathlessness due to chronic airflow obstruction Dose is equivalent to 4.5 oral morphine, up to 3 times daily
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	The study authors did not clearly state the methods of random sequence generation.
Allocation concealment (selection bias)	Unclear risk	The study authors reported allocation concealment but did not clearly state the methods.
Blinding of participants and personnel (performance bias)   All outcomes	Low risk	Participants were given similar solutions of tablets with opioid or placebo.
Blinding of outcome assessment (detection bias)   All outcomes	Unclear risk	The study authors did not clearly state the methods of blinding of outcome assessment.
Incomplete outcome data (attrition bias)   All outcomes	Low risk	The study authors clearly stated the methods of withdrawals, drop outs, and reasons, which did not appear to change the study outcomes.
Selective reporting (reporting bias)	Low risk	We did not detect any evidence of selective reporting bias.
Other bias	Low risk	We did not detect any other bias.
Size bias	High risk	There was a small sample size, and high risk of bias; there were < 50 participants per arm. Measured interquartile range so could not be included in the meta‐analysis

Methods	Randomised, crossover study Exercise study Incremental cycle ergometer
Participants	COPD 7 male participants Mean age 66.4 years SD 3.25 years FEV1 0.99 SD 0.3 FEV1/FVC 0.35 SD 0.07 Exercise limited by breathlessness Stable disease Exclusion criteria: PaCO₂ > 45 mmHg, FEV1 > 1.39 L, long‐term oxygen supplementation, cardiac disease, history of narcotic abuse, other significant disease affecting exercise performance, use of tranquillisers, hypnotics, mood altering drugs or opioids in week prior to study, alcoholism in past 5 years
Interventions	Morphine 30 mg or placebo once orally 60 minutes before exercise test, compared to 30 mg morphine plus 10 mg prochlorperazine, or compared to 30 mg morphine plus 25 mg promethazine Tests on separate days
Outcomes	Modified Borg score each minute of exercise Workload Exercise duration VO₂ VCO₂ VE PETO₂ PETCO₂ Heart rate SaO₂
Notes	The study authors concluded that the administration of 30 mg morphine plus promethazine significantly improved the exercise tolerance of participants with COPD, without significantly impairing the mental capabilities of the participants The dose is equivalent to 30 mg oral morphine
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	The study authors did not state the methods of random sequence generation.
Allocation concealment (selection bias)	Unclear risk	The study authors did not state the methods of allocation concealment.
Blinding of participants and personnel (performance bias)   All outcomes	Low risk	Placebos were identical in appearance; the participants were blinded to the intervention and the placebo.
Blinding of outcome assessment (detection bias)   All outcomes	Unclear risk	The study authors did not clearly state the methods of blinding of outcome assessment.
Incomplete outcome data (attrition bias)   All outcomes	Unclear risk	The study authors did not state the number of participants enrolled or recruited, and only provided a statement about the number that completed the study. It is unclear whether or not this had impact on the trial outcomes.
Selective reporting (reporting bias)	Low risk	We did not detect any evidence of selective reporting bias.
Other bias	Low risk	We judged that this trial appeared to be free of other sources of bias.
Size bias	High risk	This study had a small sample size; there were < 50 participants per arm.

Methods	Double blind, randomised, crossover study Exercise study Incremental cycle ergometer
Participants	Stable severe COPD with disabling breathlessness 12 men ADLs limited by breathlessness FEV1 < 1.51 Exclusion criteria: exacerbations needing antibiotics, change in oral steroid dose or hospital admission within 2 months, overt cardiac disease, contra‐indication to exercise testing, pCO₂ > 7.0, use of opioids, benzodiazepines, or other sedative agent within 1 month
Interventions	Morphine 10 mg nebulised or morphine 25 mg nebulised or morphine 1 mg intravenous or morphine 2.5 mg intravenous or placebo nebulised or placebo intravenous 15 minutes before exercise tests Each test was separated by at least 48 hours
Outcomes	Heart rate Respiratory rate VO₂ RER SaO₂ VAS for breathlessness Exercise duration Plasma morphine levels Ventilation
Notes	Nebulised Beta 2 agonist given before exercise tests Authors conclude no significant effect of opioid compared with placebo on exercise tolerance or breathlessness The primary outcome did not include standard deviation therefore could not be included in the meta‐analysis
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	The study authors did not state the methods of random sequence generation.
Allocation concealment (selection bias)	Unclear risk	The study authors did not state the methods of allocation concealment.
Blinding of participants and personnel (performance bias)   All outcomes	Unclear risk	The study authors did not state whether or not they blinded participants and personnel.
Blinding of outcome assessment (detection bias)   All outcomes	Unclear risk	The study authors did not state the methods of blinding of outcome assessment.
Incomplete outcome data (attrition bias)   All outcomes	Low risk	All recruited participants remained in the study.
Selective reporting (reporting bias)	Low risk	The study authors presented all primary and secondary outcomes described in methods.
Other bias	Low risk	We judged that this trial appeared to be free of other sources of bias.
Size bias	High risk	This study had a small sample size, and was at high risk of bias; there were < 50 participants per arm

Methods	Randomised parallel study
Participants	Ambulatory participants with moderate to severe dyspnoea at rest Mean age 55 years 31 participants in the morphine arm and 32 participants in the midazolam arm
Interventions	Oral morphine: 3 mg then incremental steps at 30 minutes until 50% reduction in dyspnoea Compared to oral midazolam: 2 mg up titrated 25% until 50% reduction in dyspnoea
Outcomes	Dyspnea intensity Adverse events
Notes	Compared morphine to midazolam
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Participants were randomly assigned (using a random number generator in 1:1 ratio in blocks of 6) to 1 of the 2 treatment groups. Numbered envelopes that were used to implement the randomisation were concealed until interventions were assigned.
Allocation concealment (selection bias)	Low risk	Participants were randomly assigned (using a random number generator in 1:1 ratio in blocks of 6) to 1 of the 2 treatment groups. Numbered envelopes that were used to implement the randomisation were concealed until interventions were assigned.
Blinding of participants and personnel (performance bias)   All outcomes	High risk	This study was single blinded, that is only participants were blinded. However, adequacy was unclear – similarity of preparation – morphine group given laxatives.
Blinding of outcome assessment (detection bias)   All outcomes	High risk	This study was single blinded, and only participants were blinded. The investigators were aware of the treatment allocation.
Incomplete outcome data (attrition bias)   All outcomes	Low risk	There was only 1 withdrawal in each group, which was unrelated to the intervention.
Selective reporting (reporting bias)	Low risk	The study authors appear to have reported all outcome data.
Other bias	Low risk	We judged that this trial appeared to be free of other sources of bias.
Size bias	High risk	There were 30 participants per treatment arm and the study was designed as a parallel trial; there were < 50 participants per treatment arm.

PERMALINK

Opioids for the palliation of refractory breathlessness in adults with advanced disease and terminal illness

Hayley Barnes

Julie McDonald

Natasha Smallwood

Renée Manser

Abstract

Background

Objectives

Search methods

Selection criteria

Data collection and analysis

Main results

Authors' conclusions

Plain language summary

Summary of findings

Background

Description of the condition

Description of the intervention

How the intervention might work

Why it is important to do this review

Objectives

Methods

Criteria for considering studies for this review

Types of studies

Types of participants

Types of interventions

Types of outcome measures

Primary outcomes

Secondary outcomes

Search methods for identification of studies

Electronic searches

Searching other resources

Data collection and analysis

Selection of studies

Data extraction and management

Assessment of risk of bias in included studies

Measures of treatment effect

Unit of analysis issues

Dealing with missing data

Assessment of heterogeneity

Assessment of reporting biases

Data synthesis

Subgroup analysis and investigation of heterogeneity

Sensitivity analysis

Results

Description of studies

Results of the search

1.

Included studies

Study characteristics

Participants

Intervention

Outcomes

Excluded studies

Ongoing studies

Risk of bias in included studies

2.

3.

Allocation

Blinding

Incomplete outcome data

Selective reporting

Other potential sources of bias

Size

Effects of interventions

for the main comparison.

2.

3.

Primary outcome: breathlessness

Opioids versus placebo

All studies

4.

1.1. Analysis.

Additional sensitivity analyses

1. Sensitivity analysis: breathlessness: impact of fixed versus random effects model and unclear bias.

5.

Type of opioid

2.1. Analysis.

3.1. Analysis.

Methods	Placebo‐controlled, double blind, randomised crossover study There were 3 drop outs: 3 participants died over the study period, during the night: the study authors did not consider their deaths to be related to the treatment
Participants	Hospital inpatients with severe lung disease, experiencing distressing dyspnoea not relieved by medical therapy Mean age 69 years (± 11 years) Mean FEV1 0.92 SD 0.18 Normal cognitive function 17 participants
Interventions	Nebulised morphine 10 mg + O₂ or morphine 20 mg + O₂, or morphine 10 mg or placebo + O₂ at 2L/min Tests took place over consecutive days
Outcomes	VAS for breathlessness SaO₂% Respiratory rate
Notes	Results using morphine alone were not analysed The study authors reported no benefit of morphine compared with placebo
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	The study authors did not clearly describe the methods of random sequence generation.
Allocation concealment (selection bias)	Unclear risk	The study authors did not clearly describe the methods of allocation concealment.
Blinding of participants and personnel (performance bias)   All outcomes	Unclear risk	All morphine and saline solutions were prepared and coded independently in the hospital pharmacy. However, the study authors did not provide any further description. For the 10 mg/no oxygen group, prongs were applied but no oxygen and it is not possible to determine whether or not the participants or investigators were truly blinded to the intervention based on the description above.
Blinding of outcome assessment (detection bias)   All outcomes	Unclear risk	The study authors did not state the methods of blinding of outcome assessment.
Incomplete outcome data (attrition bias)   All outcomes	Low risk	This study had incomplete data due to deaths during the trial, which were clearly reported and would be expected in this study type. The deaths occurred overnight and were unlikely to be related to the intervention, which took place during the day.
Selective reporting (reporting bias)	Low risk	We did not detect any selective reporting bias.
Other bias	Low risk	We judged that this trial appeared to be free of other sources of bias.
Size bias	High risk	This study had a small sample size, and was at high risk of bias; there were < 50 participants per arm.

Methods	Controlled double blind crossover study
Participants	Those with heart failure LVEF < 45%, on standard medical therapy 35 participants Mean age 70 years (± 11 years) Male 8% Exclusion criteria: co‐existing respiratory illness, peak expiratory flow (PEF) < 150 L/min, opioid sensitivities, renal impairment i.e. glomerular filtration rate (eGFR) < 30 mL/min
Interventions	Oral morphine 5 mg four times daily (QID) compared to oral oxycodone 2.5 mg four times daily (QID) compared to oral placebo Treatment arm 4 days with 3 day washout period
Outcomes	NRS dyspnoea scale NRS coping NRS satisfaction
Notes	Outcomes assessed daily The study authors concluded that there was no benefit over placebo for the relief of breathlessness with short‐term low‐dose oral opioids for congestive heart failure The dose is equivalent to 20 mg oral morphine
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	The study drug manufacturer (Calderdale and Huddersfield NHS Pharmacy Manufacturing Unit, Huddersfield, UK) randomised the order of interventions for each participant using a random number generation programme. It did not use block design and there were 6 possible sequence combinations. The pharmacy dispensed all 3 medications for use in the required sequence with identical labels except for the treatment order. Hence the investigators and participants remained blinded to the treatment sequence and allocation was conducted distant to the research team.
Allocation concealment (selection bias)	Low risk	The placebo was designed to have very similar characteristics to the active medications (a clear, colourless liquid with the same viscosity and similar taste).
Blinding of participants and personnel (performance bias)   All outcomes	Low risk	The pharmacy dispensed all 3 medications for use in the required sequence with identical labels except for the treatment order. Hence the investigators and participants remained blinded to the treatment sequence and allocation was conducted distant to the research team.
Blinding of outcome assessment (detection bias)   All outcomes	Low risk	Investigators were blinded to outcome assessment.
Incomplete outcome data (attrition bias)   All outcomes	Low risk	There was a small number of drop outs, which were all completely reported and unlikely to influence results and reasons stated.
Selective reporting (reporting bias)	Low risk	The study was registered prior to recruitment and audited study. The study reported all prespecified outcomes.
Other bias	Low risk	We judged that this trial appeared to be free of other sources of bias.
Size bias	High risk	There were < 50 participants per treatment arm.

Methods	Double blind crossover study
Participants	Clinically stable males with COPD FEV < 60% 7 participants 3 withdrawals (1 due to side effects of codeine, 1 due to worsening respiratory function) Age range 59 to 79 years All male Exclusion criteria: PCO₂ > 55, history of chemical dependence
Interventions	30 mg codeine four times daily (QID) compared to 25 mg promethazine QID Co‐interventions: beta agonists, theophylline, prednisolone Outcomes measured at baseline and at 1 month
Outcomes	VAS dyspnoea scale Breathlessness rating
Notes	The study authors concluded that the benefits of codeine or promethazine are uncertain
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Subjects were randomised in sets of 4 in a double‐blind fashion to begin oral treatment with identical pills that contained either 30 mg codeine or 25 mg promethazine, both taken 4 times daily. There were limited details in the study report.
Allocation concealment (selection bias)	Unclear risk	The study authors did not clearly report the methods of allocation concealment.
Blinding of participants and personnel (performance bias)   All outcomes	Unclear risk	The study authors did not clearly state whether they blinded participants and personnel.
Blinding of outcome assessment (detection bias)   All outcomes	Unclear risk	The study authors did not clearly state whether there was blinding of outcome assessment.
Incomplete outcome data (attrition bias)   All outcomes	Low risk	One participant dropped out on the third day of the first study arm after developing acute urinary retention while taking codeine. Three participants had marked worsening of their respiratory symptoms and required hospitalisation. Two participants were receiving codeine, 1 on the first study arm and 1 on the second arm. The third subject was receiving promethazine during the first arm. There was a small number of drop outs but also a small total number of participants. There were roughly similar reasons per group.
Selective reporting (reporting bias)	Low risk	We did not detect any selective reporting bias.
Other bias	Low risk	We did not detect any other bias.
Size bias	High risk	This study had a small sample size despite it being a parallel study; there were < 50 participants per arm

Methods	Cross‐over study Exercise study Cycle endurance test Nebulised opioid
Participants	COPD (9) or idiopathic pulmonary fibrosis (2) Mean age 58.4 years (39 to 74) Exercise tolerance limited by dyspnoea FEV1 0.4 to 1
Interventions	Morphine 5 mg or placebo neb 15 mins before exercise test Tests on separate days 100% O₂ inhaled during exercise test
Outcomes	Cycle ergometer exercise test at 80% of pre‐determined E_max Endurance time Ventilation during last minute of exercise FEV1
Notes	The study authors did not use dyspnoea as an outcome measure, although they selected participants because exercise was limited by dyspnoea. They initially studied 18 participants and 7 excluded on run‐in day because exercise was limited by other factors: the participants were excluded from the study authors’ and from this review analysis Mean endurance time increased by 64 seconds: 1 participant had an increase of 400 seconds. If the study authors had excluded that participant from their analysis, the mean increase would have been approximately 25 seconds and not statistically significant. The study authors concluded that morphine had a significant effect on exercise endurance time vs placebo
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	The study authors did not state the methods of random sequence generation.
Allocation concealment (selection bias)	Unclear risk	The hospital pharmacist performed allocation, but the study authors did not further describe the method used to conceal allocation.
Blinding of participants and personnel (performance bias)   All outcomes	Unclear risk	The study authors did not state whether they performed blinding of participants and personnel.
Blinding of outcome assessment (detection bias)   All outcomes	Unclear risk	The study authors did not state whether they performed blinding of outcome assessment.
Incomplete outcome data (attrition bias)   All outcomes	Unclear risk	The study authors did not clearly report complete participant data and dropouts.
Selective reporting (reporting bias)	Low risk	We did not detect any evidence of selective reporting bias.
Other bias	Low risk	We judged that this trial appeared to be free of other sources of bias.
Size bias	High risk	This study had a small sample size, and was at high risk of bias; there were < 50 participants per arm.

Study	Reason for exclusion
Allard 1999	This study did not compare the intervention to a placebo or any other intervention.
Beauford 1993	This study doesn't appear to be randomised, and we were unable to contact the study authors.
Bruera 2005	This study did not compare the intervention to a placebo.
Navigante 2003	This study compared morphine plus midazolam to oxygen therapy.
Peterson 1996	This study doesn't appear to be randomised.
Shorati 2012	This study doesn't appear to be randomised. We were unable to contact the study authors.
Smith 2009	This was a pilot trial of 2 participants only, and it doesn't appear to have randomised participants to treatment.
Thomas 2010	This was a review of 2 randomised controlled trials.

Trial name or title	Cuervo Pinna 2012
Methods	This is a crossover clinical trial in which the study population will do the 6MWT with the study drug and placebo.
Participants	Patients with advanced cancer. Patients must have dyspnoea at rest or dyspnoea moderate effort with an intensity of at least 3 on a scale from 0 to 10.
Interventions	Oral Transmucosal Fentanyl Citrate versus placebo
Outcomes	Primary variable: VAS scale change from baseline. Improvement of the severity of dyspnoea after completion of the 6MWT in patients with advanced cancer. Response to treatment was considered an improvement greater than or equal to two points on the previous level of dyspnea.The evaluation and determination of changes in the level of severity of dyspnoea is done through Visual Analogue Scale (VAS) included in the Edmonton Symptom Assessment System (ESAS). Secondary variables:
Starting date	2012
Contact information	Cuervo Pinna M. Extremadura, Palliative Care Support Team, Badajoz, Spain
Notes

Trial name or title	Daubert 2014
Methods	Prospective, randomised, double blind controlled trial, randomly assigned to either receive lorazepam or morphine. Symptom relief will be evaluated using the Edmonton Symptom Assessment Scale. Patients will receive treatment for 14 days.
Participants	People greater than 18 years of age, enrolled in a hospice service, diagnosed with dyspnoea, able to take oral medications
Interventions	opioids compared to benzodiazepines
Outcomes	The primary outcome will be the change in patients' perception of their quality of life. This will be assessed using the Functional Assessment of Chronic Illness Therapy‐Palliative Care scale, which patients will complete prior to initiation of therapy and after 7 and 14 days of treatment. An intention to‐ treat analysis will be performed and the change in quality of life observed will be compared between groups using a multivariable logistic regression analysis to adjust for confounding variables.
Starting date	Research is in progress from 2014
Contact information	Daubert E, Bolesta S, Wilkes University, E‐mail: eliza.daubert@wilkes.edu
Notes