Oslo Diet‐Heart 1966.

Methods	RCT
Participants	Men with previous MI (Norway) CVD risk: high Control: randomised 206, analysed 148 (at 5 years) Intervention: randomised 206, analysed 152 (at 5 years) Mean years in trial: control 4.3, intervention 4.3 % male: 100 age: mean control 56.3, intervention 56.2 (all 30‐67)
Interventions	Modified fat diet vs control Control aims: no dietary advice but direct questions answered, supplement = 1 vitamin tablet daily Intervention aims: reduce meat & dairy fats, increase fish, vegetables, supplement ‐ 1 vitamin tablet daily, 0.5L soy bean oil per week (free to 25% of participants), sardines in cod liver oil (free at certain times to encourage compliance) Control methods: usual diet Intervention methods: continuous instruction and supervision by dietitian, including home visits, letters and phone calls Total fat intake: mod fat unclear, cont unclear Saturated fat intake: mod fat unclear, cont unclear Style: diet advice & supplement (food) Setting: community
Outcomes	Stated trial outcomes: coronary heart disease morbidity and mortality Data available on total mortality? yes Cardiovascular mortality? yes Events available for combined cardiovascular events: total MI, sudden death, stroke, angina Secondary outcomes: non‐fatal and total MI, stroke Tertiary outcomes: weight, total cholesterol, systolic and diastolic BP (but no variance information is provided)
Notes	Weight change from baseline was ‐0.5kg in the control group (n˜155), ‐2.5kg in the intervention group (n˜160) to 51 months Total cholesterol change from baseline was ‐0.46mmol/L in the control group and ‐1.53mmol/L in the intervention group at 51 months. Systolic BP at baseline was 153.8mmHg in control and 159.0 in intervention, and mean sBP through trial was 154.3mmHg in control and 158.2mmHg in the intervention group. Diastolic BP at baseline was 93.5mmHg in control and 97.1mmHg in intervention, through trial mean dBP was 95.5mmHg in control and 98.6mmHg in intervention participants.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"table of random numbers used", by Prof Knut Westlund
Allocation concealment (selection bias)	Low risk	Randomisation appears to have occurred before medical examination within the study
Blinding (performance bias and detection bias) All outcomes	High risk	Participants were aware of their allocation as was the main trialist. Outcomes were categorised by a diagnostic board, but their blinded status was unclear.
Incomplete outcome data (attrition bias) All outcomes	Low risk	The participants who could not be directly followed up for the 5 years were followed until death or study end through personal interviews, or contact with their physicians or relatives.
Selective reporting (reporting bias)	Low risk	Not relevant for primary and secondary outcomes as all trialists asked for data
Other bias	Low risk
Free of systematic difference in care?	High risk	Dietetic input level very different, although medical care appeared similar. See Control and Intervention Methods in Interventions section of the Table of Characteristics of Included Studies
Free of dietary differences other than fat?	High risk	Differences in fruit and vegetables, fish etc. as above.