Background: We previously evaluated metabolic markers (alpha-hydroxybutyrate, bradykinin hydroxyproline, mannose, alpha-tocopherol, 10:1 carnitine, and X-12063) for the early prediction of type 2 diabetes (T2D). The present study compared glucose measurements at different time points during an oral glucose tolerance test (OGTT) with the metabolites for predicting future T2D.

Methods: Fasting (FPG), 30, 60, and 120 minute plasma glucose (PG) levels were measured in 543 individuals from the Botnia Prospective Study, 146 of whom progressed to T2D within a 10 year period. Machine learning based predictive models for the future risk of T2D were built based on (1) OGTT-derived glucose measurements, (2) clinical risk factors (age, sex, body mass index, and family history of T2D, (3) metabolites, (4) combinations of PG levels and metabolites, (5) combinations of PG levels and clinical risk factors, and (6) combinations of PG levels, clinical risk factors, and metabolites. The metabolites were also correlated with the PG levels.

Results: The 1-h PG predictive performance (area under the receiver operating characteristic curve, AUC = 0.76, sensitivity = 0.70, specificity = 0.69) was superior to FPG, 2-h PG, or individual metabolites, and was as good as the multivariate model with all six metabolites (AUC = 0.78, DeLong’s p = 0.24). The combined predictive model that included the 1-h PG, clinical risk factors, and six metabolites (AUC = 0.81, sensitivity = 0.74, specificity = 0.75) significantly improved the predictive performance beyond the 1-h PG alone (p = 0.02) and the combined model with clinical risk factors and six metabolites (p = 0.02).

Conclusion: The 1-h PG predicts risk for T2D better than FPG or the 2-h PG. Although several multivariate models containing the metabolites improved the prediction performance beyond the 1-h PG, the 1-h PG alone was on par, and is a more practical biomarker. Shortening the OGTT to 1 hour provides a better and simpler assessment of T2D risk than the current OGTT.


G. Peddinti: None. T. Tuomi: None. M. Bergman: None. L. Groop: None.

Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. More information is available at