How reliable are randomised controlled trials for studying the relationship between diet and disease? A narrative review