Abstract: TH-PO1018
Analysis of Japanese Histological IgA Classification Using Probabilistic Analysis Associated with the Bayesian Theorem
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
- Nakata, Takeshi, Oita University Faculty of Medicine, Yufu-shi, Japan
- Aoki, Kohei, Oita University Faculty of Medicine, Yufu-shi, Japan
- Fukuda, Akihiro, Oita University Faculty of Medicine, Yufu-shi, Japan
- Fukunaga, Naoya, Oita University Faculty of Medicine, Yufu-shi, Japan
- Tanigawa, Masato, Oita University Faculty of Medicine, Yufu-shi, Japan
- Shibata, Hirotaka, Oita University Faculty of Medicine, Yufu-shi, Japan
Background
Regarding the pathological diagnosis of IgA glomerulonephritis(IgAGN), the Oxford classification is widely used globally, although in Japan, the Japanese Society of Nephrology(JSN) published its original pathological classification. For appropriate classification, the Oxford classification requires more than 8 glomeruli, whereas the Japanese classification requires more than 10 glomeruli. However, no study has yet investigated the relationship between the total number of glomeruli and pathological classification by probabilistic analysis. The present study aimed to report how the total number of glomeruli affects the “lumped system” of the Japanese histological IgA classification using probabilistic analysis associated with the Bayesian theorem.
Methods
Ninety-nine patients from 2000 to 2009 diagnosed IgAGN by renal biopsy at Oita University Hospital were included. Certified pathologist diagnosed IgAGN using light microscopy and fluorescence microscopy. We used the third edition of IgAGN classification of JSN. We used Bayesian theorem for Probabilistic analysis. We used three models of the prior distribution. First is actual distribution; Second is a similarity of actual distribution by using the beta function, third is no information for the prior distribution by using the beta function. The ethical committee of Oita University approved this survey.
Results
The median total number of collected glomeruli was 12 [Quartile 7,19]. When the cut-off level was set to less than 60% of the posterior probability, 21 cases (33%) were excluded (7 cases had more than 10 glomeruli, 14 cases had less than 9 glomeruli). When cases with less than 9 glomeruli were excluded before the Bayesian probability test, only 8 cases (12%) showed less than 60% of the posterior probability. However, 19 cases with less than 9 glomeruli showed more than 60% of the posterior probability. Thus, these 19 cases were considered as exclusion cases for classification. The results were the same using the three models of the prior distribution.
Conclusion
It may be better when using the IgA pathological classification of JSN to adopt the probabilistic analysis associated with the Bayesian theorem instead of considering only the total number of obtained glomeruli.