CAPRICORN: a story of alpha allocation and beta-blockers in left ventricular dysfunction post-MI

Beta-blocker therapy is beneficial both after myocardial infarction and in mild, moderate and severe chronic heart failure. Recent sub-group analysis of the Goteborg Metoprolol Trial and the AIRE study confirm that patients receiving beta-blockers in the setting of post-MI heart failure fare better than patients not receiving this therapy. For all these reasons the CAPRICORN trial of carvedilol in post-MI LV dysfunction was an important and eagerly awaited trial. The results were presented for the first time at The American College of Cardiology on March 20 2001. CAPRICORN randomised 984 patients to placebo and 975 to carvedilol between 3 and 21 days (mean 10) after a confirmed MI. Patients had to have evidence of a left ventricular ejection fraction 40% or less. All patients had received ACEI therapy for at least 48 hours prior to randomisation. The mean ejection fraction of the patients recruited was 32.7% in the placebo group and 32.9% in the carvedilol group. Follow-up was for a mean of 15 months (maximum 2.7 years). All cause mortality was 15.3% (151 deaths) in the placebo group and 11.9% (116 deaths) in the carvedilol group, giving a hazard ratio of 0.77 (0.60-0.98) and a significance of p = 0.031. And yet this agent will probably not be given a licence for this indication in the European Union and the USA. The reason is one of trial design and statistical declarations. Some way into the trial the Steering Committee decided to change the primary end-point form all-cause mortality to two co-primary end-points and to allocate their alpha power of 0.05 unevenly between the combined end-points of all cause mortality and cardiovascular hospitalisation (alpha 0.045) and all cause mortality (alpha 0.005). In the end neither was achieved, one because of the large number of non-specific hospitalisations for chest pain and the mortality effect because it was allocated a punitive and unachievable target of p < 0.005. The trial is thus officially neutral despite showing convincing clinical benefit. Clearly arcane matters of statistical plans do matter and steering committees should think very carefully before changing the primary end-points of major trials.