CHAOS 1996Low.

Methods	Cambridge Heart Antioxidant Study (CHAOS). Randomised, double‐blind, placebo‐controlled trial with parallel group design (two intervention groups).
Participants	Country: United Kingdom. Number of participants randomised: 2002; 1690 men and 312 women, mean age 61.8 years. Inclusion criteria: angiographically proven coronary arthrosclerosis. Exclusion criteria: prior use of vitamin supplements containing vitamin E.
Interventions	Participants were randomly assigned to receive: group 1: vitamin E 800 IU, and then 400 IU (free 2R,4'R,8'R‐alpha‐tocopherol from natural sources in soya oil) (n = 1035): 546 patients received 800 IU daily for a median of 731 days (range 3 to 981), and 489 newly recruited patients received 400 IU daily for a median of 366 days (8 to 961). These two groups are not distinguished in the analysis. group 2: matching placebo (oil only), (n = 967). Median follow‐up was 510 days (range 3 to 981).
Outcomes	The primary outcome measures were: non‐fatal myocardial infarction alone and combination of non‐fatal myocardial infarction and cardiovascular death.
Notes	Compliance with treatment was measured as the ratio of days that study medication was requested to per‐protocol days prescribed. 73.2% of all prescribed alpha‐tocopherol or placebo were requested as follow‐up medications. There was no difference between treatment groups in the proportion of participants who were 100% compliant with the trial medication (48% placebo, 49% alpha‐tocopherol). Complete follow‐up data were available in 98% of participants. There were no differences between the groups in completeness of follow‐up (98% placebo, 97.8% active treatment). Overall 23 participants in the active and 19 participants in the placebo group were lost to follow‐up. Study agents were supplied by Henkel Corporation (La Grange, Illinois, USA) Data were extracted from Mitchinson MJ, Stephens NG, Parsons A, Bligh E, Schofield PM, Brown MJ. Mortality in the CHAOS trial. Lancet 1999;353:381‐2.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Sequence generation was achieved using computer random number generation.
Allocation concealment (selection bias)	Low risk	Allocation was controlled by a central and independent randomisation unit, so that intervention allocations could not have been foreseen in advance of, or during, enrolment.
Blinding (performance bias and detection bias) All outcomes	Low risk	The trial was described as blinded, the parties that were blinded, and the method of blinding was described, so that knowledge of allocation was adequately prevented during the trial.
Incomplete outcome data (attrition bias) All outcomes	Low risk	The numbers and reasons for dropouts and withdrawals in all intervention groups were described.
Selective reporting (reporting bias)	Low risk	Pre‐defined, or clinically relevant and reasonably expected outcomes are reported on.
Other bias	Low risk	The trial appears to be free of other components that could put it at risk of bias.