Abstract: TH-PO1023
Risk Prediction Model Including Therapeutic Options in IgA Nephropathy
Session Information
- Glomerular Diseases: Minimal Change Disease, FSGS, IgAN
November 07, 2019 | Location: Exhibit Hall, Walter E. Washington Convention Center
Abstract Time: 10:00 AM - 12:00 PM
Category: Glomerular Diseases
- 1203 Glomerular Diseases: Clinical, Outcomes, and Trials
Authors
- Matsuzaki, Keiichi, Kyoto University Health Service, Kyoto, Japan
- Suzuki, Hitoshi, Juntendo University Faculty of Medicine, Tokyo, Japan
- Imai, Takumi, Graduate School of Medicine, Kyoto University, Kyoto, Japan
- Aida, Rei, Graduate School of Medicine, Kyoto University, Kyoto, Japan
- Barbour, Sean, University of British Columbia, Vancouver, British Columbia, Canada
- Cattran, Daniel C., Toronto General Hospital, Toronto, Ontario, Canada
- Katafuchi, Ritsuko, National Hospital Organization Fukuokahigashi Medical Center, Koga, Fukuoka, Japan
- Suzuki, Yusuke, Juntendo University Faculty of Medicine, Tokyo, Japan
Background
Recently derived prediction model for kidney outcome of IgA nephropathy (IgAN) based on clinical and pathological predictor variables from multi-ethic cohorts (Barbour SJ, JAMA Intern Med. 2019) is a potentially valuable tool in terms of its application to daily clinical practice as well as for clinical trial design. However, since timing of intervention including immunosuppressive therapy is highly variable and is dependent on patient characteristics, pathology, physician preference, country bias etc. The prediction equation currently does not include therapeutic factors after renal biopsy. Therefore, we derived and validated model used at 6 months after renal biopsy with therapeutic factors in Japanese dataset from original multi-ethic cohorts.
Methods
The outcome was defined as the composite outcome of 50% decline in eGFR or ESKD. We derived the renal prediction model including treatment within 6 months after the biopsy by Cox regression modeling with variable selection using the derivation dataset. The performances of the Cox models and the prognostic scores were evaluated based on Akaike information criteria (AIC) and the discrimination ability measured by Chambless and Diao's C-statistics. The derived score was externally validated using the validation dataset by C-statistics, calibration slope.
Results
Japanese patients from the multi-ethnic cohorts were divided to 564 derivation and 582 validation datasets. Over a median 3.31 and 4.63 years of follow-up, 10.1% and 17.2% reached the outcome, respectively. Among the variables at biopsy and treatments, eGFR, blood pressure, proteinuria at biopsy, T and C score, immunosuppression (IS) use, tonsillectomy and interactions between IS and C score, eGFR and tonsillectomy were selected for the prediction model. The derived prediction model achieved better performance compared to the model including only the variables at biopsy (AIC: 599.1 vs 562.6, C: 0.83 95%CI 0.75 – 0.89, ΔC: 0.05, 95%CI 0.00-0.12). The performance of the score remained fine in the validation dataset (C: 0.92, 95% CI 0.86-0.96, calibration slope: 0.97, 95%CI 0.82-1.11).
Conclusion
We derived and validated the risk prediction model at 6 months after renal biopsy including therapeutic options. This model may be useful guide for appropriate treatment in Japan.
Funding
- Government Support - Non-U.S.