Proteomic CKD Prediction in Type 2 Diabetes

The five-plasma protein model integrating eGFR with a five-protein–based CKD predictor, developed by Xu et al. and published in JASN,1 represents an important advance in proteomic risk stratification for CKD in type 2 diabetes. However, several methodological and interpretive concerns merit consideration before clinical translation. Although the model achieved high sensitivity (82%) and negative predictive value (96%), the positive predictive value was only 22%,1 indicating that most individuals classified as high risk would not develop CKD. This low positive predictive value partly reflects the modest prevalence of incident CKD (about 11%) in the validation cohort. Expressed in prevalence-independent terms, the estimated positive likelihood ratio (approximately 3.3) offers limited rule-in strength, whereas the negative likelihood ratio (around 0.24) demonstrates strong rule-out performance. Such characteristics make the model useful for population screening, but insufficient for stand-alone clinical decision making. Inspection of model coefficients shows that α-1-microglobulin/bikunin precursor and placental growth factor exert disproportionately large influences relative to eGFR and kidney injury molecule-1.1 This suggests that the algorithm primarily captures systemic inflammatory and endothelial injury rather than kidney-specific damage. Comparable proteomic and metabolomic studies show that angiogenic and inflammatory signatures often reflect generalized cardiometabolic dysfunction rather than isolated kidney injury.2,3 Nevertheless, systemic inflammation and endothelial dysfunction are integral to CKD pathophysiology, and their inclusion may still reflect relevant upstream processes. The presence of partially redundant proteins, such as α-1-microglobulin/bikunin precursor and kidney injury molecule-1,1 further suggests potential for model simplification to enhance assay practicality and cost-efficiency. Both discovery and validation cohorts were recruited before the widespread use of sodium-glucose cotransporter 2 inhibitors and glucagon-like peptide-1 receptor agonists, therapies known to reduce CKD incidence by approximately one third.4 The model may therefore overestimate current risk and requires recalibration to contemporary contexts. External validation in therapy-era cohorts could allow researchers to adjust the coefficients for current risk distributions and drug exposures, aligning predicted risks with modern treatment patterns. In addition, the study populations were ethnically homogeneous, limiting the assessment of ancestry-related proteomic variability that can affect the calibration of biomarkers. In summary, the eGFR+five-protein–based CKD predictor model provides strong rule-out performance and valuable mechanistic insight. While confirmatory precision and generalizability require refinement, it offers a promising foundation for future validation in diverse, therapy-exposed populations and for advancing kidney-specific proteomic profiling.