Carnaby 2006i.

Methods	Computerised randomisation Blinded outcome assessments by SLT ITT (Control vs low‐intensity data set) Baseline prognostic factors balanced between treatment groups
Participants	1 centre in Australia 306 participants; baseline characteristics similar Enrolment within 2 weeks of stroke onset: mean/median 2 days, range 0 to 12 days Clinical and videofluoroscopic evidence of dysphagia
Interventions	Rx 1: standardised high‐intensity swallowing therapy (n = 102) Rx 2: standardised low‐intensity swallowing therapy (n = 102); split into (n = 51) for each data set C: usual care (n = 102) Treatment for up to 1 month
Outcomes	Outcomes: time to return to normal diet; aspiration pneumonia; dysphagia (PHAD score < 85)
Notes	Trial completed and published 2006 Exclusions: previous swallowing therapy, head and neck surgery, inability to consent Follow‐up: 6 months
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Treatment allocation based on a computer‐generated random numbers list generated via the SPSS statistical package
Allocation concealment (selection bias)	Low risk	Randomisation schedule held at the trial office, remote from the study environment; assignment to 1 of 3 treatment options by a telephone call to the trial office made by the study speech pathologist
Blinding (performance bias and detection bias) All outcomes	High risk	All people involved in the study unaware of treatment allocation, apart from participants and the study speech pathologist who treated participants Assigned to high‐intensity and low‐intensity groups
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participants and speech pathologist aware of treatment allocation
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Outcome assessed by an independent speech pathologist, who was unaware of treatment allocation, every month for 6 months after randomisation
Incomplete outcome data (attrition bias) All outcomes	Low risk	3 participants lost to follow‐up before 6‐month analysis
Selective reporting (reporting bias)	Low risk	All outcomes reported
Other bias	Low risk	None identified