Abstract: SA-PO271

The Derivation and Validation of an International Multi-Ethnic Risk Prediction Model in IgA Nephropathy

Session Information

Category: Glomerular

  • 1005 Clinical Glomerular Disorders

Authors

  • Barbour, Sean, University of BC, Vancouver, British Columbia, Canada
  • Coppo, Rosanna, N/a, Torino, Italy
  • Suzuki, Yusuke, Juntendo University, Tokyo, Japan
  • Liu, Zhi-Hong, Nanjing University, Nanjing, China
  • Espino-Hernandez, Gabriela, BC Renal Agency, Vancouver, British Columbia, Canada
  • Reich, Heather N., University of Toronto, Toronto, Ontario, Canada
  • Cattran, Daniel C., University of Toronto, Toronto, Ontario, Canada

Group or Team Name

  • International IgAN Collaboration
Background

Predicting renal outcome in IgA nephropathy (IgAN) is challenging. A prediction model is needed to improve risk stratification that is properly validated in multiple ethnic groups worldwide and can be used in clinical practice. To overcome these obstacles, we used large datasets from international collaborators to generate an accurate prediction model in IgAN.

Methods

The derivation dataset was from European, Japanese and Chinese adult cohorts; the validation dataset was from separate North/South American, European, Chinese and Japanese cohorts. Time from biopsy to the composite outcome (50% decline in eGFR or ESRD) was analyzed using Cox survival models.

Results

The validation dataset (N=2784) is 42% Caucasian, 21% Japanese and 37% Chinese; 18% (N=495) experienced the composite outcome over a median 4.8 years of follow-up. Two models were considered: a reduced model containing eGFR, blood pressure, proteinuria at biopsy, and MEST score; and a full model that also contained age, sex, race, use of RAS blockade/immunosuppression, and crescents. Compared to the reduced model, the full model had improved prediction with better AIC (5679 vs 5648), R2 (19 vs 21%), ΔC-statistic (0.01, 95%CI 0.008-0.03), continuous NRI (0.17, 95%CI 0.11-0.27) and IDI (0.03, 95%CI 0.01-0.05), with similar calibration curves. We will next externally validate the full model in the validation dataset (N=1401), and convert the model into web and mobile app calculators for implementation in clinical practice.

Conclusion

Using the largest and most diverse datasets to date in IgAN, we will generate the first risk prediction tool that is externally validated in multiple ethnic groups worldwide and can be easily implemented in clinical practice using web/app-based calculators. We expect the prediction model will become the international standard for risk stratification in IgAN, and will facilitate both clinical trial recruitment of high-risk patients and testing the added prediction benefit of novel biomarkers.