Objective: Fibroblast growth factor 21 (FGF21) improves glucose and lipid metabolism, but high circulating levels are found in type 2 diabetes, suggesting FGF21 resistance. Serum FGF21 predicts incident diabetes, but its performance compared to established and emerging predictors is not known. We aimed to study the performance of FGF21 in diabetes prediction, relative to other adipokines and established risk factors including 2-h plasma glucose (2hG) during the oral glucose tolerance test (OGTT).
Design/participants/measurements: We studied 1380 nondiabetic subjects from the Hong Kong Cardiovascular Risk Factor Prevalence Study using the second visit (2000-2004) as baseline when serum levels of FGF21 and other adipokines were measured. Glycaemic status was assessed by OGTT. Incident diabetes was defined as fasting glucose level (FG) ≥ 7 mmol/l or 2hG ≥ 11·1 mmol/l or use of antidiabetic agents, at subsequent visits.
Results: A total of 123 participants developed diabetes over 9·0 years (median). On multivariable logistic regression analysis, FGF21 (P = 0·003), adipocyte fatty acid-binding protein (P = 0·003) and adiponectin (P = 0·035) were independent predictors of incident diabetes. FGF21 had the best change in log likelihood when added to a diabetes prediction model (DP) based on age, family history, smoking, hypertension, BMI, dyslipidaemia and FG. It also improved the area under ROC curve (AUROC) of diabetes prediction (DP) from 0·797 to 0·819 (P = 0·0072), rendering its performance comparable to the 'DP + 2hG' model (AUROC=0·838, P = 0·19).
Conclusions: As a biomarker for diabetes prediction, serum FGF21 appeared to be superior to other adipokines and, on its own, could be considered as an alternative to the OGTT.
© 2016 John Wiley & Sons Ltd.