Burden 2017.

Study characteristics
Methods	RCT. Parallel design with 2 arms. Duration: oral nutritional supplements were administered from diagnosis to the day preceding surgery for a minimum of 5 days. Location: multicentre study in the UK.
Participants	Inclusion criteria: diagnosis of colorectal cancer, scheduled for surgery with pre‐operative weight loss > 1 kg/3 – 6 months. Exclusion criteria: pregnant, had a pacemaker, already taking a similar ONS, with insulin dependent diabetes Number randomised: 101 participants (intervention group n = 55; control group n = 46). Gender split: intervention 64% male, control 70% male. Age: mean (SD) intervention group 70.5 (11.7) years; control group 68.9 (11.5) years. Nutritional status: median (IQR) % weight loss, intervention 4.9 kg (2.2 ‐ 8.8); control 6.8 kg (3.4 ‐ 12.1)
Interventions	Intervention: participants received dietary advice and ONS in the form of 250 mL/day oral nutritional supplements (10.1 KJ and protein 0.096 g/mL) and dietary advice. Control: participants received dietary advice alone in the form of dietary advice.
Outcomes	Primary outcome: 1 or more surgical site infection or chest infection. Secondary outcomes: % weight loss, total complications, and body composition measurements.
Publication details	Language: English. Funding: Macmillan Cancer Support and British Dietetic Association. Publication status: peer‐reviewed journal.
Notes	Non‐normally distributed data are displayed as median and IQRs. In this format, the data are not in a format sufficient for meta‐analyses. Data on handgrip strength are reported as mean (SD) and thus included in meta‐analyses. On request, change scores of weight, kcal intake, and protein intake were obtained from the author, and included in the meta‐analyses. Outcomes are reported at 2 time‐points, preoperatively and post‐operatively. As the intervention was given preoperatively only, we have chosen to only report the preoperative outcome measures in this review.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	The participants were randomly allocated on a 1:1 ratio by using blocks of two ensuring equal numbers in each group. Allocation was stratified according to tumour site (rectal versus colon) and surgical approaches (open versus laparoscopic). 4 lists of random numbers were produced by a statistician, and an independent researcher set up the randomization procedure for each of the strata.
Allocation concealment (selection bias)	Low risk	Sequentially‐numbered opaque sealed envelopes were used, which allowed block randomization sequence allocation to be implemented and ensured sequence allocation concealment.
Blinding (performance bias and detection bias) Clinical outcomes	Low risk	Described as single blind. For the purposes of blinding, control participants were given sealed cardboard boxes of identical weight and appearance as the supplement group at the time of group allocation. These boxes contained bottled water in 125 mL bottles. Thus research team was blind to the intervention, but the participants were not.
Blinding (performance bias and detection bias) Functional outcomes	Low risk	Described as single blind. For the purposes of blinding, control participants were given sealed cardboard boxes of identical weight and appearance as the supplement group at the time of group allocation. These boxes contained bottled water in 125 mL bottles. Thus the research team was blind to the intervention, but the participants were not.
Blinding (performance bias and detection bias) Nutritional outcomes	Low risk	Described as single blind. For the purposes of blinding, control participants were given sealed cardboard boxes of identical weight and appearance as the supplement group at the time of group allocation. These boxes contained bottled water in 125 mL bottles. Thus research team was blind to the intervention, but the participants were not.
Blinding of participants and personnel (performance bias) All outcomes	High risk	The research team was blind to the intervention, but the participants were not. It is possible that assessment of some outcomes was influenced by lack of blinding
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	The research team was blind to the intervention, but the participants were not.
Incomplete outcome data (attrition bias) All outcomes	High risk	1 of 55 (2%) in the intervention group died; 6 of 46 (13%) in the control group failed to complete (4 died, 1 participant withdrew from the control group and 1 further participant from the control group was excluded from the analysis). For the outcome measures: handgrip strength, % weight loss, PG‐SGA, energy intake and protein intake data are provided for fewer participants than originally included. The reasons why data were missing for these outcomes is not reported.
Selective reporting (reporting bias)	Unclear risk	Study protocol identified. All outcomes reported apart from hospital readmissions and quality of life (secondary outcomes). Many outcomes reported as median and IQR, thus not in a format sufficient for meta‐analysis.
Other bias	Unclear risk	Baseline variables reported. Overall, there were more participants in the intervention group (n = 55) than in the control group (n = 46). At the point of recruiting participants, if it was undecided if surgery was open or laparoscopic, the default used was open‐surgery stratum for randomization. The two arms of the trial were well‐matched with similar proportions of participants within each stratum, site of cancer, and type of operation (laparoscopic or open).