The need for large-scale randomized evidence without undue emphasis on small trials, meta-analyses, or subgroup analyses.