Low 10-year reproducibility of glycaemic index and glycaemic load in a prospective cohort study: methodological and statistical issue to avoid misinterpretation.