Development and validation of a discrimination model between primary PLA2R-negative membranous nephropathy and minimal change disease confirmed by renal biopsy

Abstract

Membranous nephropathy (MN) and minimal change disease (MCD) are two common causes leading to nephrotic syndrome (NS). They have similar clinical features but different treatment strategies and prognoses. M-type phospholipase A2 receptor (PLA2R) is considered as a specific marker of membranous nephropathy. However, its sensitivity is only about 70%. Therefore, there is a lack of effective and noninvasive tools to distinguish PLA2R-negative MN and MCD patients without renal biopsy. A total 949 patients who were pathologically diagnosed as idiopathic MN or MCD were enrolled in this study, including 805 idiopathic MN and 144 MCD. Based on the basic information and laboratory examination of 200 PLA2R-negative MN and 144 MCD, we used a univariate and multivariate logistic regression to select the relevant variables and develop a discrimination model. A novel model including age, albumin, urea, high density lipoprotein, C3 levels and red blood cell count was established for PLA2R-negative MN and MCD. The discrimination model has great differential capability (with an AUC of 0.904 in training group and an AUC of 0.886 in test group) and calibration capability. When testing in all 949 patients, our model also showed good discrimination ability for all idiopathic MN and MCD.

Machine learning-based diagnostic prediction of minimal change disease: model development study

“…Another study on adult nephrotic syndrome patients showed high discrimination ability for MCD with an AUROC of 0.827 using two predictor variables 13 . In a study on a discrimination model between primary PLA2R-negative membranous nephropathy and MCD, logistic regression maintained an AUROC of 0.904 during development and 0.886 during validation 12 . While the previous studies have examined the effectiveness of predictive models in diagnosing MCD, all the models were classical statistical methods such as logistic regression.…”

“…The need for non-invasive diagnostic methods for MCD has been discussed, but studies using predictive models are limited 8, [11][12][13]32,33 . A study using logistic LASSO regression on 1009 adult nephrotic syndrome patients demonstrated the performance of AUROC 0.880, indicating the utility of predictive models for diagnosing MCD 11 .…”

“…However, these biomarkers cannot yet be measured in general medical facilities without advanced equipment, so their clinical utility remains uncertain. Developing predictive models using clinically measurable parameters could be bene cial, but research on such models for diagnosing MCD is limited [11][12][13] . Previous studies on diagnostic predictive models have employed logistic regression.…”

Serological Biomarker-Based Machine Learning Models for Predicting the Relapse of Ulcerative Colitis

“…The nomogram transforms the complex regression equation into a visual graph, making the results of the prediction model more readable. 33 The six variables in the nomogram are routine clinical variables that are readily available to clinicians. The decision curve showed the clinical utility of our model, suggesting that it may be beneficial for clinicians to use the nomogram to predict UC relapse.…”

Construction of a Nomogram Discriminating Malignancy-Associated Membranous Nephropathy From Idiopathic Membranous Nephropathy: A Retrospective Study

“…Clinical prediction models have rapidly developed in renal disease, due to their scientific nature, accuracy, and simplicity. Examples of such models include those used for the distinction between MN and minimal change disease and between diabetic kidney disease and diabetes mellitus (20)(21)(22). However, MMN-related clinical models are currently lacking.…”