Dennis 2005.

Methods	Randomised clinical trial (stratified for age, sex, and predicted probability of poor outcome), UK
Participants	4023 hospitalised adults with either: 1. admission to a hospital due to a stroke (1st or recurrent stroke) within 7 days of onset OR 2. suffering a stroke whilst already in hospital where the randomising clinician was uncertain about the best feeding policy and with consent or assent obtained from close relatives as well as having passed a shallow screen. The participants were at nutritional risk due having had a stroke. Male:Female: 53% male Mean age = 71 years Exclusion: (a) People with subarachnoid haemorrhage, people who experienced a transient ischaemic attack (TIA) or trivial stroke and were likely to remain in hospital for only a few days (b) people who could swallow but in whom nutritional supplementation was contraindicated (e.g. morbidly obese) (c) those in coma (i.e. unresponsive to pain) or who were very unlikely to survive more than a few days because of some severe non‐stroke illness OR (d) people who had already been entered into the same FOOD Trial
Interventions	Experimental group: oral nutritional supplement (equivalent to 360 mL at 6·27 kJ/mL and 62·5 g/L in protein every day) and regular hospital diet(n = 2016) Control group: regular hospital diet(n = 2007)
Outcomes	Death or poor outcome and overall survival at 6 months, health‐related QoL among survivors, time to hospital discharge, length of stay in hospital, number of days of tube‐feeding, adverse effects of feeding regimens, premature cessation of feeding regimens and reasons
Study dates	Nov 1996 to August 2003
Notes	We contacted the authors on 12th November 2015 by email: martin.dennis@ed.ac.uk. We received data on quality of life.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated
Allocation concealment (selection bias)	Low risk	Locked computer
Blinding of participants and personnel (performance bias) All outcomes	High risk	Not blinded. Participants knew their allocation.
Blinding of outcome assessment (detection bias) All outcomes	High risk	Only a blinded assessment at 6 months follow‐up.
Incomplete outcome data (attrition bias) All outcomes	Low risk	7 dropouts but reasons for the dropouts were clearly stated and the trial used intention‐to‐treat.
Selective reporting (reporting bias)	Low risk	All clinically relevant outcomes were reported, as stated in the protocol.
For‐profit bias	Low risk	FOOD was funded by the NHS R&D Health Technology Assessment Programme (Reference 96/29/01), The Stroke Association (Reference 17/98) and Chest Heart and Stroke Scotland (Reference 97/4). The Singapore Medical Research Council supported the trial in Singapore. The Royal Australasian College of Physicians supported the trial in Hawkes Bay, New Zealand.
Other bias	Low risk	The trial appeared to be free of other components that could put it at risk of bias.