Mortality benefit in the COMPASS trial: is it related to superior statistical power or better efficacy and safety?