EPE‐A 2014

Methods	EPE‐A RCT, parallel, 3 arms (n‐3 EPA, low dose vs high dose vs unclear placebo), 12 months Summary risk of bias: moderate or high
Participants	People with non‐alcoholic steatohepatitis (NASH) and non‐alcoholic fatty liver disease (NAFLD) N: 86 intervention‐high, 82 int low, 75 control (analysed 64, 55, 55 respectively, ITT analysis for primary outcomes) Level of risk for CVD: low (although 35% had type II diabetes) Men: 33.7% intervention‐high, 41.5% intervention‐low, 42.7% control Mean age in years (SD): 47.8 (11.1) intervention‐high, 47.8 (12.5) intervention‐low, 50.5 (12.5) control Age range: not reported Smokers: not reported Hypertension: not reported Medications taken by at least 50% of those in the control group: not reported Medications taken by 20%‐49% of those in the control group: not reported Medications taken by some, but less than 20% of the control group: not reported Location: USA Ethnicity: white intervention‐low: 94%, intervention‐high: 87%, control: 90.7% African American intervention‐low: 3.7%, intervention‐high: 2.3%, control: 4.0% Others intervention‐low: 2.4%, intervention‐high: 10.5%, control: 5.3%
Interventions	Type: supplement (omega 3 capsule) Comparison 1: high EPA vs low EPA (unclear what replaced EPA) Comparison 2: EPA vs unclear (placebo contents not reported) Intervention‐high: EPA‐E 2.7 g/d, 3 × EPA‐E 300 mg capsules. Dose: 2.7 g/d EPA + DHA Intervention‐low: EPA‐E 1.8 g/d, 2 × EPA‐E 300 mg capsules + 1 placebo capsule Dose: 1.8 g/d EPA + DHA Control: 3 × placebo capsules. The pills were identical with respect to size, colour and smell Compliance: estimated by pill count and measuring the ratio of serum EPA to arachidonic acid. compliance rates for the 3 groups (placebo vs EPA‐E 1800 mg/d vs EPA‐E 2700 mg/d) were 89.5% (6.8%), 90.3% (5.7%) and 89.5% (5.3%), respectively Length of intervention: 12 months
Outcomes	Main study outcome: histological response in standardised scoring of liver biopsies and change in ALT level Dropouts: 22 intervention‐high, 27 intervention‐low, 20 control Available outcomes: cardiac events, deaths (none), angina, adverse events (weight, BMI, lipids, glucose, HbA1c, HOMA, hsCRP all reported as medians so not useable in meta‐analyses) Response to contact: yes (provided methodological details)
Notes	Data combined for the 2 intervention groups for binary outcomes and higher dose data vs control used for continuous outcomes Study funding: supported entirely by Mochida Pharmaceuticals
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Block randomisation using an interactive voice‐response system to assign subjects in a 1:1:1 ratio between the 2 arms for each site separately. Participants were stratified by the presence of type 2 diabetes. The total fraction of such individuals was capped at 40% of the study cohort
Allocation concealment (selection bias)	Low risk	As above (remote computer‐generated randomisation)
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Double‐blind stated, but no further details. Author confirmed researchers and outcome assessors were blinded to treatment allocation and pills were identical with respect to size, colour and smell
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	No details
Incomplete outcome data (attrition bias) All outcomes	High risk	Number and characteristics of participants lost to follow‐up similar across arms, however < 80% provided outcome data relevant to this systematic review
Selective reporting (reporting bias)	Low risk	Registered June 2010, study started June 2010, completed October 2012. All outcomes in trials registry entry were also reported in the trials registry. Secondary outcomes reported were not planned (compared with first version of clinicaltrials.gov entry)
Attention	Low risk	All participants had same follow‐up visits.
Compliance	Low risk	Compliance was estimated by pill count and measuring the ratio of serum EPA to arachidonic acid. Compliance rates for the 3 groups (placebo vs EPA‐E 1800 mg/d vs EPA‐E 2700 mg/d) were 89.5% (6.8%), 90.3% (5.7%) and 89.5% (5.3%) respectively
Other bias	Low risk	None noted