SOFA 2006

Methods	Study on Omega‐3 Fatty Acids and Ventricular Arrhythmia (SOFA) 2 arm, parallel RCT (n‐3 EPA + DHA vs MUFA), 12 months Summary risk of bias: low
Participants	People with previous ventricular arrhythmias and implantable cardioverter defibrillators N: 273 intervention, 273 control (273 intervention, 273 control analysed) Level of risk for CVD: high Men: 84% intervention, 85 % control Mean age in years (SD): 60.5 (12.8) intervention, 62.4 (11.4) control Age range: unclear (18 years and older) Smokers: 16% intervention, 8% control Hypertension: 53% intervention, 49% control Medications taken by at least 50% of those in the control group: beta‐blockers Medications taken by 20%‐49% of those in the control group: lipid lowering, antiarrythmic medications (combined) Medications taken by some, but less than 20% of the control group: amiodarone, sotalol Location: 8 countries in Europe Ethnicity: not reported
Interventions	Type: supplement (capsule) Comparison: EPA + DHA vs MUFA + omega 6 Intervention: 2 g/d (4 capsules) purified fish oil. 961 mg n‐3 PUFAS (464 mg EPA + 335 mg DHA and 162 mg other n‐3 PUFAs) daily. 3000 ppm vitamin E (Loders Croklann, Wormeveer). Dose: 0.8 g/d EPA + DHA Control: 2 g/d high‐oleic acid sunflower oil. 3000 ppm vitamin E (Loders Croklann, Wormeveer) Compliance: daily diary, checked by research nurses every 4 months. Judging by capsule count, 207 patients in the fish oil group and 218 in the placebo took more than 80% of their capsules. N‐3 fatty acid composition in serum cholesterol levels was measured at baseline and the end of the trial. The EPA concentration in serum cholesterol esters increased in the expected range. No data provided Length of intervention: 12 months
Outcomes	Main study outcome: spontaneous ventricular tachyarrhythmias and all‐cause mortality Dropouts: 33 intervention (23 partial follow‐up), 33 control (14 partial follow‐up) Available outcomes: deaths, MI, new angina, new heart failure, no fatal arrhythmias, cancer, cardiovascular events, side effects Response to contact: yes but no data provided
Notes	Study funding: Wageningen Centre for Food Sciences (alliance of major Dutch food industries and others)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Patients using beta‐blockers were separately randomised in blocks of 2. A computer randomisation programme randomly took the first treatment of a block. The second patient in a block of 2 always received the opposite treatment.
Allocation concealment (selection bias)	Low risk	Treatments (blinded medication numbers) were centrally assigned by a telephone allocation service.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Double blinding. Bottles containing capsules labelled with medication numbers that are unidentifiable for patients as well as investigators. Fish oil and placebo capsules have identical appearance. Difference can't be tasted if swallowed with water (as suggested)
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Quote: "blinded endpoint adjudication committee"
Incomplete outcome data (attrition bias) All outcomes	Low risk	ITT analysis. Did a partial follow‐up on some patients who dropped out due to non‐compliance.
Selective reporting (reporting bias)	Low risk	NCT00110838, trial registered in May 2005, end of trial January 2005, trial results published in 2006. However, rationale and design paper (stating outcomes) published in 2003. Outcomes in the 2006 paper appear to be the same as in Rationale paper.
Attention	Low risk	Unlikely as intervention blinded to investigators and only intervention was capsules
Compliance	Unclear risk	Daily diary, checked by research nurses every 4 months. Judging by capsule count, 207 patients in the fish oil group and 218 in the placebo took more than 80% of their capsules. N‐3 fatty acid composition in serum cholesterol levels was measured at baseline and the end of the trial. The EPA concentration in serum cholesterol esters increased in the expected range. No data provided
Other bias	Low risk	No further bias noted