How the Lancet Cooked the Numbers