Should subgroup analysis of randomized clinical trials have a direct impact on clinical practice?