A multicenter prospective observational cohort (TRIDENT, n=176) of adults with diabetes undergoing clinically indicated kidney biopsy; the study tested whether unsupervised k-means clustering of light-microscopy features could improve on the standard Renal Pathology Society (RPS) classification for predicting kidney outcomes (death, dialysis initiation, or ≥40% eGFR decline) over 1–2 years, with external validation in 101 individuals.
The existing RPS system showed no significant survival difference between Classes 3 and 4. The new RPS-TRIDENT Modification (RPS-TM) added a Class 5—defined by visceral epithelial hyperplasia—that more clearly separated risk groups and modestly improved 1-year AUC (0.68 vs. 0.65) and 2-year AUC (0.75 vs. 0.71), with higher net benefit on decision-curve analysis; Class 5 had the fastest progression to kidney failure.
- Modest sample sizes (176 derivation, 101 validation), which limits statistical power and generalizability. - AUC improvements are incremental (~3–4 points), so added clinical value at the individual patient level remains uncertain. - Observational design with clinically indicated biopsies introduces selection bias toward more advanced or atypical disease.
When reviewing kidney biopsies in patients with diabetes, look for visceral epithelial hyperplasia as a marker of very high risk—these patients may warrant closer follow-up and earlier consideration of kidney replacement therapy planning. The RPS-TM framework could also inform enrollment criteria for clinical trials targeting advanced diabetic nephropathy.
Explore related topics
