Amminger‐Austria.

Methods	Allocation: randomised Blinding: double‐blind (participant, care provider, investigator, outcomes assessor) Setting: Vienna, Austria; the major referral source was the outpatient service (52, 64.2%). Also derived from psychiatrists and psychologists from the department, other youth services or adult mental health services and private mental health professionals Duration: 12 months (12 weeks intervention + 36 weeks monitoring). Thereafter 7‐year follow‐up
Participants	Diagnosis: people at high risk of developing psychosis N = 81 Sex: men and women Age: 13‐25 years
Interventions	1. omega‐3 fatty acids: dose 4 capsules daily – each containing 700 mg of eicosapentaenoic acid, 500 mg of docosahexaenoic acid and 10 mg of Vitamin E. N = 41 2. Placebo (coconut oil capsules matched with appearance and taste). N = 40 Concomitant medication use after randomisation was allowed: antidepressants and benzodiazepines. Existing medication was re‐evaluated at baseline and continued in case of clinical indication. Psychological and psychosocial interventions as well as additional appointments for crisis management were provided.
Outcomes	Transition to PANSS‐defined first episode psychosis Leaving the study early Mental state: PANSS, MADRS Functioning: GAF Adverse effects: UKU Global state: prescription of antipsychotic medication (assumed to represent the severity of psychotic phenomena) Additional outcomes: physiological: neuroinflammation biomarkers, EEG activity, phospholipid metabolism, erythrocyte membrane fatty acid composition and intracellular phospholipase A2 activity
Notes	Cut‐off points on PANSS subscales, (≥ 4 hallucinations, ≥ 4 delusions, and ≥ 5 conceptual disorganisation). Funding: Grant 03T‐315 from the Stanley Medical Research Institute. Power, sample size calculation: "The study was powered to detect a 50% reduction in the expected transition rate, corresponding to a transition rate of 20% in the ‐3 group and an anticipated rate of 40% in the placebo group. Power analysis indicated that 75 subjects would provide a 70% chance of detecting such an effect (2‐sided level of.05). Allowing for a 5%to 10% dropout rate, we sought to recruit at least 80 participants." Adherence: see Table 14
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated random sequence based on a block‐randomised design. Stratified according to MADRS. Two strata with block size of 4 within each stratum.
Allocation concealment (selection bias)	Unclear risk	Quote: "Randomisation kept in a remote secure location and administered by an independent third party until all study data were collected and verified." Comment: precise method of allocation concealment not described
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Participants, parents, and those involved in administering interventions, assessing outcomes, data entry, and/or data analyses were blind to group assignments.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Participants, parents, and those involved in administering interventions, assessing outcomes, data entry, and/or data analyses were blind to group assignments.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Out of 81 randomised participants, 5 discontinued the study before 12 months' follow‐up (attrition 6%). Attrition rate was 8% in the omega‐3 fatty acids group (N = 3: 1 participant/parent decision, 1 physician decision, participant moved out of the country) and 7% in the placebo group (N = 2: 2 participant/parent decisions)
Selective reporting (reporting bias)	Low risk	All outcomes mentioned in registered protocol (NCT00396643) and publications' methods reported
Other bias	Low risk	We did not identify any other sources of bias