Eiser 1991.

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.