Evidence of systematic and proportional error in a widely used glucose oxidase analyser: Impact for clinical research?