A Prediction Model for Acute Kidney Injury in Adult Patients With Minimal Change Disease

Abstract

Background: Early prediction of acute kidney injury (AKI) can allow for timely interventions, but there are still few methods that are easy and convenient to apply in predicting AKI, specially targeted at patients with minimal change disease (MCD). Motivated by this, we aimed to develop a predicting model for AKI in patients with MCD within the KDIGO criteria.

Methods: Data on 401 hospitalized adult patients, whose biopsy was diagnosed as MCD from 12/31/2010 to 15/7/2021, were retrospectively collected. Among these data, patients underwent biopsy earlier formed the training set (n = 283), while the remaining patients formed the validation set (n = 118). Independent risk factors associated with AKI were analyzed. From this, the prediction model was developed and nomogram was plotted.

Results: AKI was found in 55 of 283 patients (19%) and 15 of 118 patients (13%) in the training and validation cohorts, respectively. According to the results from lasso regression and logistic regression, it was found that four factors, including mean arterial pressure, serum albumin, uric acid, and lymphocyte counts, were independent of the onset of AKI. Incorporating these factors, the nomogram achieved a reasonably good concordance index of 0.84 (95%CI 0.77-0.90) and 0.75 (95%CI 0.62-0.87) in predicting AKI in the training and validation cohorts, respectively. Decision curve analysis suggested clinical benefit of the prediction models.

Conclusions: Our predictive nomogram provides a feasible approach to identify high risk MCD patients who might develop AKI, which might facilitate the timely treatment.

Keywords: acute kidney injury (AKI); minimal change disease; nephrotic syndrome; nomogram; prediction model.

Figure 1

Flow diagram of the patient cohort. To build up the prediction model, the study population was divided into training and validation cohorts.

Figure 2

Texture feature selection using LASSO logistic regression and the predictive accuracy of the radiomics signature. (A) Selection of the tuning parameter (λ) in the LASSO model via 10-fold cross validation based on minimum criteria. The optimal λ value of 0.0577 with log(λ) = −2.85 was selected. (B) LASSO coefficient profiles of the 25 texture features. The dotted vertical line was plotted at the value selected using 10-fold cross-validation in (A).

Figure 3

The nomogram to estimate the risk of AKI in MCD patients. To use the nomogram, search the position of each variable on the corresponding axis, draw a line to the points axis for the number of points, add the total points, and draw a line from the total points axis to determine the AKI probabilities at the lower line of the nomogram. For example, doctor checked an MCD patient blood pressure is 120/90 mmHg, and her laboratory test revealed serum albumin 20 g/L, serum urci acid 400 μmol/L, lymphocyte counts 2.5*10⁹/L. Based on the nomogram, her points were 29, 62, 36 and 55, respectively. The total points was 182 and the probability of AKI was 0.2.

Figure 4

ROC curves of the AKI model. (A) Prediction of AKI (AUC = 0.84 95%CI 0.77–0.90). (B) Prediction of AKI by the bootstrap (500 resample) method. Blue shading shows the bootstrap estimated 95% CI with the AUC (AUC = 0.83 95%CI 0.75–0.89).

Figure 5

Decision curve analysis depicting the clinical efficiency in our cohort.