Summary of findings 2. Positional therapy compared to inactive control for obstructive sleep apnoea.

Positional therapy compared to inactive control for obstructive sleep apnoea
Patient or population: adults with obstructive sleep apnoea Setting: intervention at home except in one study where intervention was for one night in the laboratory Intervention: positional therapy Comparison: no positional therapy or sham therapy
Outcomes	Anticipated absolute effects* (95% CI)		Relative effect (95% CI)	No of participants (RCTs)	Certainty of the evidence (GRADE)	Comments
	Risk with control	Risk with positional therapy
Epworth Sleepiness Scale (ESS) Follow‐up: 4 weeks to 2 months	The mean ESS ranged from 9.4‐10.9	MD 1.58 lower (2.89 lower to 0.26 lower)	‐	187 (2 RCTs)	⊕⊕⊕⊝ Moderatea	Laub 2017 was an open‐label study over 2 months with home polygraphy at 2 months. Jackson 2015 studied participants with hospital‐based PSG after 4 weeks of intervention. Lower ESS scores are better. MCID is estimated to be a fall of 2‐3 points
Apnoea‐Hypopnoea Index (AHI) Follow‐up: 1 night to 2 months	The mean AHI ranged from 16.8‐19.9 event/hour	MD 7.38 event/hour lower (10.06 lower to 4.7 lower)	‐	277 (4 RCTs)	⊕⊕⊝⊝ Lowb,c	Laub 2017 was an open‐label study over 2 months with laboratory polygraphy at 2 months. Jackson 2015 studied participants with hospital‐based PSG after 4 weeks of intervention. The other 2 trials were cross‐over design, Bignold 2011 being conducted over 1 week with home PSG and Van Maanen 2012 over 1‐2 weeks with laboratory‐based PSG. Lower AHI is better. MCID for AHI is considered to be 5
Adherence Measured as number of participants who continued to use device at end of 2 months Follow‐up: 2 months	Study population		OR 0.80 (0.33 to 1.94)	101 (1 RCT)	⊕⊕⊝⊝ Lowd	Laub 2017 was an open‐label study over 2 months with laboratory polygraphy at 2 months. They measured device use for minimum 4 h/night over 2 months. Rate of discontinuation of therapy used for this comparison, thus lower OR implies fewer dropouts and improved adherence.
	755 per 1000	712 per 1000 (504 to 857)
Adverse effects Measured as number of participants who discontinued device at end of 2 months Follow‐up: 2 months	255 per 1000	288 per 1000	OR 1.25 (0.52 to 3.03)	101 (1 RCT)	⊕⊕⊝⊝ Lowd	Laub 2017 was an open‐label study over 2 months with laboratory polygraphy at 2 months. They reported that 15 participants dropped out of the SPT arm: 5 due to adverse effects (frequent awakening and poor sleep (2), pain in the thorax and unpleasant feeling (3); 2 for lack of effect; and remaining 8 for other reasons (lost to follow up 5, withdrawal 1, sleep problem improved 1 and did not understand trial 1)
Quality of life Assessed using SF‐36 or FOSQ Follow‐up: 4 weeks	Mean FOSQ: 3.3	MD 0.2 higher (0.02 lower to 0.42 higher)		86 (1 RCT)	⊕⊕⊕⊝ Moderatee	Jackson 2015 studied participants with hospital‐based PSG after 4 weeks of intervention. They reported that FOSQ scores were significantly higher in positional therapy group (P < 0.01). Higher scores on FOSQ indicate improved quality of life.
Sleep quality Assessed by average duration of slow‐wave and REM sleep periods Follow‐up: 1 night	Mean % of REM sleep: 19.2%; mean % of slow wave sleep: 19.7%	MD for % REM sleep 0.9% lower (5.06% lower to 3.26% higher); MD for % slow wave sleep 1.20% lower (5.22% lower to 2.82% higher)		30 (1 RCT)	⊕⊕⊝⊝ Lowe,f	Van Maanen 2012 conducted a cross‐over trial using hospital‐based PSG and reported that percentage of REM sleep and percentage of slow‐wave sleep were not significantly different between the groups.
Cognitive dysfunction Follow‐up: 4 weeks	Mean motor reaction time in seconds is 193.5	MD 12.4 seconds lower (23.10 lower to 1.70 lower)		86 (1 RCT)	⊕⊝⊝⊝ Very lowg	Jackson 2015 studied participants with in‐hospital PSG after 4 weeks of intervention. They reported that results of motor vigilance test was not significantly different between the groups.
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; CPAP: continuous positive airway pressure; FOSQ: Functional Outcomes Sleep Questionnaire; MCID: minimal clinically important difference; MD: mean difference; OR: odds ratio; PSG: polysomnography; RCT: randomised controlled trial; REM: rapid eye movement; RR: risk ratio; SF‐36: short‐form 36; SPT: sleep position trainer
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.

^aLaub 2017 has high risk of bias. It was an open‐label study and had a high attrition rate. It did not mention methods of allocation concealment. Downgraded for risk of bias.
 ^bBignold 2011 and Van Maanen 2012 did not explicitly state the randomisation and allocation concealment procedure. Jackson 2015 had discrepancy in the stated allocation procedure and the actual distribution of the participants in the study. Laub 2017 was an open‐label study. It did not state the procedure of allocation concealment and had significant loss to follow‐up. Downgraded for risk of bias.
 ^cLaub 2017 and Bignold 2011 used home‐based monitors, while the other two studies used laboratory‐based polysomnography with reported kappa for agreement 0.6. Downgraded for imprecision of measurement techniques.
 ^dLaub 2017 did not mention methods of allocation concealment. It was an open‐label study with a high attrition rate. Downgraded for risk of bias.
 ^eThe confidence interval of the estimate is imprecise. Downgraded for imprecision.
 ^fVan Maanen 2012 has moderate risk of bias, has no clear primary outcome and multiple comparisons. Downgraded for risk of bias.
 ^gThis study is at risk of bias, and has multiple comparisons. The study authors reported no difference between the groups for the outcomes on cognition. However, we noticed motor reaction time to be significantly different in favour of positional therapy. We downgraded this parameter for indirectness as its clinical significance in isolation is uncertain. Downgraded twice for risk of bias and once for indirectness.