Gravar-mail: Optimal sampling ratios in comparative diagnostic trials