- Design
- Multicentre retrospective development and external validation study
- Population
- 443 patients with biopsy-confirmed lupus nephritis, five Chinese centres
- Primary outcome
- Discrimination for proliferative lupus nephritis
- Effect
- AUC 0.852 (development); 0.895 (external validation)
A multicentre retrospective study from China developed and externally validated a model to identify proliferative lupus nephritis using routine tests. It included 443 patients with biopsy-confirmed nephritis: 337 from one centre for development and 106 from four others for validation. Published in Rheumatology on 14 October 2026.
The final model used seven variables: lymphocyte percentage, haematuria, leukocyturia, eGFR, uric acid, anti-dsDNA positivity and C3. Its area under the ROC curve was 0.852 in development and 0.895 in external validation, with acceptable calibration.
Biopsy remains the standard, and nothing here replaces it where it is possible. But the model may help prioritise urgent biopsy, guide early treatment when biopsy is delayed or contraindicated, and support decisions in settings with limited access to renal pathology, as in parts of India. The validation cohort was small and from one country, so performance elsewhere is not yet known.
- Keep kidney biopsy as the standard for classifying lupus nephritis wherever it is feasible
- Consider a validated clinical model to prioritise urgent biopsy
- Use urine sediment, eGFR, anti-dsDNA and C3 together rather than any single marker
- Consider early treatment decisions informed by the model when biopsy is contraindicated or delayed, with specialist input
Why it matters
Access to timely renal biopsy is uneven, and treatment intensity depends on the histological class.
Don't overread it
Retrospective development with a small external cohort from one country; performance elsewhere is untested.
The statistics, in plain English
An AUC of 0.895 means the model ranks a patient with proliferative nephritis above one without it about 90% of the time. External validation in other centres is what makes this figure more believable than a development-only result.
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.
Scan to keep reading on your phone. No account needed to start.
Tomorrow morning, before your first patient
One edition a day for rheumatology, written by the desk, every claim tied to its paper. Six minutes.
Get the app — free