Large-scale proteomics improve risk prediction for type 2 diabetes

figure

posted on 2025-04-03, 15:41 authored by Ruijie Xie, Tomislav Vlaski, Kira Trares, Christian Herder, Bernd Holleczek, Hermann Brenner, Ben Schöttker

OBJECTIVE

This study evaluated the incremental predictive value of proteomic biomarkers in assessing 10-year type 2 diabetes risk when added to the clinical Cambridge Diabetes Risk Score (CDRS).

RESEARCH DESIGN AND METHODS

Data from 21,898 UK Biobank participants were used for model derivation and internal validation, and 4,454 ESTHER cohort (Germany) participants for external validation. Proteomic profiling included the OLINK-Explore (2,085 proteins) and OLINK-Target-96-Inflammation panel (73 proteins).

RESULTS

Adding 15 proteins from OLINK-Explore or 6 proteins from the OLINK-Inflammation panel improved the C-index of the CDRS by 0.029 or 0.016 in internal validation with net reclassification of 23.0% and 29.0%, respectively. External validation was only conducted for the 6-protein-extended model and the C-index improved by 0.014.

CONCLUSIONS

The OLINK-Explore-based 15-protein-model enhanced the CDRS model performance most and this promising prediction model should be externally validated. Our successful external validation of the OLINK-Inflammation-panel-based 6-protein-model shows that this is a promising endeavor.