We suggest not using opioids for the treatment of breathlessness in people with serious respiratory illness (conditional recommendation against the intervention, very low certainty of evidence).

We considered studies of any opioid drug, given by intravenous, subcutaneous or oral routes in any dose, for the treatment of breathlessness or cough. The effects of opioids were considered separately for 1) opioids self-administered regularly at home for four or more consecutive days with outcomes ideally measured in daily life (e.g. breathlessness “now”) at a singular (either morning or evening, combined averages of morning and evening, or an unspecified time) or multiple time points (both morning and evening separately) [89–99], and 2) opioids administered as one or more doses in the laboratory setting with participants completing an exercise test, and the effect of one or more doses measured at varying times [94, 100–104].

The search for systematic reviews identified 2736 records, of which 17 were screened in full text. Five relevant systematic reviews were identified, which included 15 eligible trials. The search for additional trials identified 2332 records, with 14 screened in full text, and two additional trials included. One included trial published after the search date [91] was identified from a qualitative study identified during the search [105]. We included 17 trials (876 participants). 11 trials included only people with COPD (11 trials, 316 participants [91, 96–104]) although the majority of participants were from RCTs with mixed cohorts (four trials, 505 participants [89, 90, 93, 94]). One trial included only people with PAH (19 participants [92]) and another only included people with ILD (36 participants [95]). Seven trials included participants with a modified Medical Research Council (mMRC) dyspnoea score ≥3 [91, 93–95, 99, 100, 104], and two trials included participants with mMRC ≥2 [90, 98].

For our critical outcome of breathlessness, opioids did not reduce breathlessness intensity measured in daily life compared to the comparator (placebo in all studies, except one which used promethazine) measured either in the morning (SMD −0.10, 95% CI −0.64–0.43; 10 trials, 795 participants) or evening (SMD −0.10, 95% CI −0.64–0.44; 10 trials, 795 participants) [89–98]. In a meta-analysis of five trials [94, 100, 102–104] examining breathlessness measured at iso-load during laboratory exercise testing, breathlessness intensity was lower following opioids compared with placebo (SMD −0.50, 95% CI −0.84–−0.16; five trials, 70 participants). This translated to a ∼10 mm mean difference on visual analogue scale, which meets the MID of 10 mm, suggesting that this effect is clinically significant [27].

For the important outcome of HRQoL, opioids administered regularly at home for four or more consecutive days showed no significant effect compared to placebo (SMD −0.42, 95% CI −0.98–0.13; six trials, 703 participants) [90, 91, 93, 95, 96, 98]. For the important outcome of cough, opioids administered regularly at home for four or more consecutive days showed no significant effect on cough scores compared to placebo (SMD −1.42, 95% CI −3.99–1.16; two trials, 147 participants) [95, 98].

Two separate meta-analyses examining arterial blood gas (ABG) parameters (four trials examining arterial carbon dioxide tension (PaCO2) [95, 97–99] and two trials examining arterial oxygen tension (PaO2) [98, 99]), with opioids being administered regularly at home for four or more consecutive days compared to placebo, showed no difference in PaO2 measurements (SMD −0.22, 95% CI −0.56–0.12; two trials, 122 participants), but a statistically significant increase in PaCO2 (SMD 0.86, 95% CI 0.03–1.69; four trials, 165 participants). This translates to a 2.2 mmHg increase in PaCO2, which is not clinically significant. Meta-analyses of ABG parameters (including two trials examining PaCO2 and PaO2 [101, 102]), where opioids were administered as one or more doses in the laboratory setting compared to placebo, showed no significant difference for PaO2 (SMD −0.52, 95% CI −1.14–0.10; two trials, 21 participants) or PaCO2 (SMD 0.63, 95% CI 0.00–1.26; two trials, 21 participants).

The frequency of key adverse events was increased among people receiving opioids compared to the comparator (placebo in all studies, except one which used promethazine) in studies where opioids were administered regularly at home for four or more consecutive days [90–93, 95–99]. This included constipation (OR 3.08, 95% CI 1.69–5.61; nine RCTs, 781 participants), nausea or vomiting (OR 3.32, 95% CI 1.70–6.51; eight trials, 733 participants) and drowsiness (OR 1.37, 95% CI 1.01–1.86; eight trials, 704 participants). In some studies, treatment-emergent adverse events were mild and self-limiting on withdrawal of morphine [89, 90, 92, 93, 95, 97, 99]. However, serious adverse events were reported in two studies [91, 96]; the Breathlessness, Exertion, And Morphine Sulfate study [91] indicated that 33% (46 out of 139) of participants treated with morphine developed serious adverse events, including hospitalisation and death. No differences in frequency of nausea/vomiting were detected between people receiving opioids or placebo in studies where opioids were administered in the laboratory setting [100, 101, 104].

The overall certainty of evidence was very low. Certainty of evidence was affected by reporting bias (lack of prospectively registered protocols), selection bias (unclear random sequence generation and allocation concealment), indirectness (most participants had COPD), lack of sufficient washout periods and small sample sizes. No studies included people at the very end of life, who are usually highly symptomatic and often prescribed opioids for symptom palliation. Studies mainly included people with COPD.