A new histologic class separates the fastest progressors in diabetic nephropathy

Design

reanalysis of a multicentre prospective observational biopsy cohort using unsupervised k-means clustering, with external validation

Population

176 patients with diabetes undergoing clinically indicated kidney biopsy (TRIDENT); external validation cohort of 101

Primary outcome

composite of death, dialysis initiation or 40% or greater decline in eGFR

Effect

added Class 5 defined by visceral epithelial hyperplasia with the most rapid progression; prediction AUC 0.68 vs 0.65 at 1 year and 0.75 vs 0.71 at 2 years, with higher net benefit on decision-curve analysis

The Renal Pathology Society classification of diabetic nephropathy is the standard for reporting biopsies and has a known weakness: it discriminates poorly at the severe end, where it matters most. In the TRIDENT cohort of 176 patients with diabetes undergoing clinically indicated kidney biopsy, classes 3 and 4 showed no significant difference in outcome at all.

Unsupervised clustering of light microscopy features produced a modified classification adding a Class 5, defined by visceral epithelial hyperplasia — a podocyte change rather than a mesangial or sclerotic one. That class separated risk groups more clearly and had the most rapid progression to kidney failure, in a cohort where outcomes were death, dialysis initiation or a 40% or greater fall in eGFR. It was validated in a separate cohort of 101 patients.

The gain in prediction is honest about its size: area under the curve rose from 0.65 to 0.68 at one year and 0.71 to 0.75 at two years. That is an improvement, not a transformation, and the authors present decision-curve analysis rather than resting on the AUC alone — which is the right way to argue that a modest statistical gain translates into net clinical benefit.

What it offers a clinician is a name for a group they have probably already noticed: the patient whose biopsy does not look worse than the last one but whose function falls away far faster than the classification predicts.

• Ask the renal pathologist specifically about visceral epithelial hyperplasia when reporting a diabetic nephropathy biopsy
• Do not read classes 3 and 4 as a meaningful prognostic distinction; they were not one in this cohort
• Use the histologic class alongside eGFR trajectory and albuminuria rather than instead of them
• Where Class 5 features are present, plan for kidney replacement therapy earlier — access, transplant workup, patient education
• Note that this derives from patients biopsied for a clinical indication, who are not all patients with diabetic kidney disease

The statistics, in plain English

An area under the curve of 0.75 at two years against 0.71 for the existing classification is a real but modest gain in discrimination — the modified system ranks pairs of patients correctly 75% of the time rather than 71%. The more persuasive evidence is the decision-curve analysis, which asks whether using the new classification to make decisions produces more net benefit than the old one across a range of risk thresholds; a small AUC gain can still be worth having if it separates a genuinely distinct group, and a Class 5 defined by a specific morphological feature is more interpretable than a statistical risk score.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.