Gay 2003.

Methods	Randomised, double‐blind, parallel group trial. Machines had settings: 'Set CPAP'; 'Set NBL' and 'A'. Assessors could alter the settings for safety reasons. Randomisation not reported
Participants	N = 27 participants (22 M/5 F). Age: 44 years; BMI: 35 kg/m2; AHI: 43; ESS: 13.8 Inclusion criteria: > 18 years; AHI > 10 and < 100; ability to follow instructions and provide informed consent; willingness to return for follow‐up visit 30 days after random allocation to CPAP/BiPAP; residence within 200 miles of clinic Exclusion criteria: inability to wear a mask; prior surgical treatment for OSA; prior CPAP usage; other significant comorbidities
Interventions	Bi‐level PAP versus CPAP. Participants also given instruction via educational video on CPAP and OSA. Study duration: 30 days
Outcomes	Machine usage (average hours used) Symptoms (ESS) AHI Tolerability Quality of life (FOSQ)
Funding & conflicts of interest statements	Quote: "Dr. Peter Gay received grant support for this study by Respironics Inc. (noted in manuscript)."
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Information not available
Allocation concealment (selection bias)	Low risk	Quote: "the technician selected the 'A' mode which in a double‐blinded and selective way, locked in the optimal settings for either CPAP or NBL"
Blinding of participants and personnel (performance bias) Machine usage, symptoms, quality of life, withdrawal, adverse effects	Low risk	Quote: "the technician selected the 'A' mode which in a double‐blinded and selective way, locked in the optimal settings for either CPAP or NBL.............true identification of the 'A' mode was not revealed until the end of the trial"
Blinding of participants and personnel (performance bias) AHI, blood pressure, treatment pressure	Low risk	These outcomes unlikely to be affected by study design.
Blinding of outcome assessment (detection bias) Machine usage, symptoms, quality of life, withdrawal, adverse effects	Low risk	Quote: "the technician selected the 'A' mode which in a double‐blinded and selective way, locked in the optimal settings for either CPAP or NBL.............true identification of the 'A' mode was not revealed until the end of the trial"
Blinding of outcome assessment (detection bias) AHI, blood pressure, treatment pressure	Low risk	These outcomes unlikely to be affected by study design.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Randomisation was performed prior to baseline PSG, after which a number of participants became ineligible. All participants recruited to the 2nd phase completed the study. There were no withdrawals from this study, although the sample reflects a selected population.
Selective reporting (reporting bias)	High risk	Data on quality of life (FOSQ) were not reported but described as 'equivalent' between two treatment arms.
Other bias	Low risk	No concerns identified