Mortality in the CHAOS trial

CHAOS was a clinical trial of a-tocopherol, designed to test the possibility that a-tocopherol would stabilise advanced coronary atherosclerotic lesions. That report was based on a census at June 18, 1995, and showed 75% decrease in nonfatal myocardial infarction in the a-tocopherol group, but apparently no effect on mortality. The trial was concluded on March 21, 1996, when all patients were offered atocopherol. The disbanding of the trial organisation at this point prevented further census of non-fatal myocardial infarction, but more accurate study of deaths and of compliance has now been possible. Certificates obtained on all deaths in CHAOS provided the certified cause and the name of the doctors to be contacted for further details. The study was conducted blind to treatment groups and included necropsy or case reports of patients who died in hospital and personal contact with family doctors who had certified deaths occurring at home. Compliance was assessed from pharmacy records. The drugs were dispensed at recruitment (168 capsules of 400 IU atocopherol or placebo; 84 days supply if on 800 IU, 168 days, if on 400 IU). The patients themselves ordered new supplies; all the dates were individually recorded in the pharmacy. Non-compliance was always due to cessation rather than irregular medication. The date of the last repeat supply indicated that the latest date the capsules could have been taken regularly was 84 days (or 168 days) later. There is thus no record of compliance for deaths in the first 84 (or 168) days of the trial. An arbitrary cut-off point for compliance of 30 days further deprivation was added to the 84 or 168 days of availability of capsules, based upon a report that, after 3 weeks’ supplementation of 1600 IU daily, plasma concentrations returned to normal within 5 days. An intention-to-treat analysis of deaths was done as before, using Kaplan-Meier, log-rank, and Cox hazards regression analyses. There were 120 deaths from all causes, 68 in the 1035 atocopherol group and 52 in the placebo group of 967 (cardiovascular deaths 53 vs 44, p=0·48). The major predictor of mortality, by Cox hazards regression analysis, was NYHA class of left ventricular function at recruitment (hazard ratio 1·76 [SE 0·24], p=0·0001). 25 out of 28 deaths certified by family doctors as due to myocardial infarction proved after discussion to be sudden unexpected deaths, when the doctor was called after death, and were therefore recategorised as sudden deaths. Together with myocardial infarction, heart failure, and dysrhythmias these were considered to be deaths due to ischaemic heart disease. Ischaemic heart disease deaths in patients who died early and those known to be compliant or non-compliant are shown in the table. The 21 a-tocopherol-non-compliant deaths occurred after deprivation of 31–1027 days (average 365 days). Compliance with a-tocopherol had declined to 42·6% by March, 1996 (placebo 43·3%). 22% of the