Study supports lower cut-off point for defining prediabetes