Stow 2015.

Study characteristics
Methods	RCT (feasibility study). Cluster randomised. Duration: 6 months. Location: UK.
Participants	Inclusion criteria: MUST score > 1, able to eat and drink, registered with a GP and eligible for treatment by the UK National Health Service, living in residential care home. Exclusion criteria: receiving enteral or parenteral nutrition, receiving nutritional support (advice or oral nutritional supplements), known eating disorder or condition not compatible with receiving the intervention, non‐native English speaker, lacking capacity. Number randomised: 93 participants (intervention group 1, n = 29; intervention group 2, n = 32; control group, n = 32). Attrition: 63 of 93 residents (36%). Gender split: 20 males, 73 females. Age: mean (SD) years not reported. Nutritional status: 50/91 (55%) high nutritional risk, 41/91 (45%) medium nutritional risk (MUST)*.
Interventions	Intervention (intervention group 1): particpants received ONS in the form of standard care* plus 2x liquid oral nutritional supplements providing 600 kcal, 24 g protein daily. Intervention (intervention group 2): participants received dietary advice in the form of face‐to‐face instruction from the primary researcher to increase intake by approximately 600 kcal and 20 ‐ 25g protein daily. Control: participants received no dietary advice and no ONS in the form of standard care* plus instructions to care and catering staff on increasing residents' daily intake by 600 kcal, 20 ‐ 25g protein per day. *provision of a calorie‐dense diet, small frequent, energy‐enriched meals in a family‐style dining room with prompting and assistance.
Outcomes	Nutritional intake, weight, BMI, handgrip strength, MUAC, TSF, VAS, QoL (EQ‐5D, CO‐OP).
Publication details	Language: English. Funding: undertaken by an MRes student funded by NIHR Clinical Academic Training Programme for AHPs. Publication status: peer‐reviewed journal.
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: sequence was generated using a computer‐generated random number list. Homes were randomised once eligible residents had been identified.
Allocation concealment (selection bias)	Low risk	Quote: sequence was generated using a computer‐generated random number list. Homes were randomised once eligible residents had been identified.
Blinding (performance bias and detection bias) Clinical outcomes	Low risk	No blinding, but unlikely that assessment of clinical outcomes would be influenced by lack of blinding.
Blinding (performance bias and detection bias) Functional outcomes	High risk	No blinding, and likely that assessment of functional outcomes would be influenced by lack of blinding. Quote: 1 primary researcher communicated intervention allocation … and conducted outcome assessments. It was therefore impossible to blind the researcher to intervention.
Blinding (performance bias and detection bias) Nutritional outcomes	High risk	No blinding, and likely that assessment of nutritional outcomes would be influenced by lack of blinding. Quote: 1 primary researcher communicated intervention allocation…and conducting outcome assessments. It was therefore impossible to blind the researcher to intervention.
Blinding of participants and personnel (performance bias) All outcomes	High risk	Quote: residents were recruited prior to random allocation of care homes to the treatment/control arms. Individual residents were not told of the care home intervention assignment. …due to the nature of the interventions, ... it was not possible to blind the staff delivering them. Judgement, incomplete blinding and lack of blinding likely to have influenced outcome assessments.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Researcher collected all data; it was impossible to blind the researcher to group. No blinding and lack of blinding might have influenced the assessment of some outcomes.
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Attrition amount fully reported but group allocation unclear. 63/93 (36%) participants completed the study. 23 deaths, 6 in the control (dietary advice) group, 6 in the intervention (supplement) group and 11 in the routine care group (information provided by author). Numbers and reasons for withdrawal reported but not according to group allocation.
Selective reporting (reporting bias)	Low risk	Protocol published on Current Controlled Clinical Trials. All outcomes specified reported or commented on, e.g. qualitative outcomes to be reported in a separate paper and QoL not analysed because of completion by too few residents.
Other bias	High risk	Baseline characteristics presented, the group receiving the control (food‐based intervention) were heavier than the intervention (supplement) group and had a higher energy, protein and fluid intake at baseline as well as a higher EQ5D VAS score. Judged to be likely that these differences would influence outcome assessment. Assessment of risk of bias in cluster‐randomised trials (1) Recruitment bias: yes (care home residents recruited after recruitment of care homes) (2) Baseline imbalance: weight, energy, protein & fluid intake, QoL (detail above) (3) Loss of clusters: no (4) Incorrect analysis: ? (feasibility trial and so not carried out for analyses in this paper, but planned for subsequent trial) (5) Comparability with individually randomised trials / different types of clusters: inclusion in meta‐analyses results in heterogeneity for some outcomes