Abstract
BACKGROUND: The two most relevant pathologies of long-term peritoneal dialysis (PD) are simple sclerosis and encapsulating peritoneal sclerosis (EPS). The histological differentiation of those two entities is difficult. The Aim of the study was to establish a method to standardize and facilitate the differentiation between simple sclerosis and EPS METHODS: We investigated 58 peritoneal biopsies - 31 EPS patients and 27 PD patients. Two blinded investigators analyzed 20 histological characteristics in EPS and PD patients. RESULTS: THE FOLLOWING FINDINGS WERE SIGNIFICANTLY MORE COMMON IN EPS THAN IN PATIENTS ON PD WITHOUT EPS: fibroblast like cells (FLC) (p<0.0001), mesothelial denudation (p<0.0001), decreased cellularity (p = 0.008), fibrin deposits (p<0.03), Fe deposits (p = 0.05), podoplanin vascular (p<0.0001), podoplanin avascular (p<0.0001). Using all predictor variables we trained the classification method Random Forest to categorize future cases. Podoplanin vascular and avascular were taken together (p<0.0001), FLC (p<0.0001), mesothelial denudation (p = 0.0005), calcification (p = 0.0026), acellular areas (p = 0.0094), and fibrin deposits (p = 0.0336) showed up as significantly important predictor variables. Estimated misclassification error rate when classifying new cases turned out to be 14%. CONCLUSION: The introduced statistical method allows discriminating between simple sclerosis and EPS. The misclassification error will likely improve with every new case added to the database.
Braun, N