Development and validation of a novel nomogram model for predicting delayed graft function in deceased donor kidney transplantation based on pre-transplant biopsies

Abstract

Background: Delayed graft function (DGF) is an important complication after kidney transplantation surgery. The present study aimed to develop and validate a nomogram for preoperative prediction of DGF on the basis of clinical and histological risk factors.

Methods: The prediction model was constructed in a development cohort comprising 492 kidney transplant recipients from May 2018 to December 2019. Data regarding donor and recipient characteristics, pre-transplantation biopsy results, and machine perfusion parameters were collected, and univariate analysis was performed. The least absolute shrinkage and selection operator regression model was used for variable selection. The prediction model was developed by multivariate logistic regression analysis and presented as a nomogram. An external validation cohort comprising 105 transplantation cases from January 2020 to April 2020 was included in the analysis.

Results: 266 donors were included in the development cohort, 458 kidneys (93.1%) were preserved by hypothermic machine perfusion (HMP), 96 (19.51%) of 492 recipients developed DGF. Twenty-eight variables measured before transplantation surgery were included in the LASSO regression model. The nomogram consisted of 12 variables from donor characteristics, pre-transplantation biopsy results and machine perfusion parameters. Internal and external validation showed good discrimination and calibration of the nomogram, with Area Under Curve (AUC) 0.83 (95%CI, 0.78-0.88) and 0.87 (95%CI, 0.80-0.94). Decision curve analysis demonstrated that the nomogram was clinically useful.

Conclusion: A DGF predicting nomogram was developed that incorporated donor characteristics, pre-transplantation biopsy results, and machine perfusion parameters. This nomogram can be conveniently used for preoperative individualized prediction of DGF in kidney transplant recipients.

Keywords: Delayed graft function; Kidney transplantation; LASSO regression; Nomogram; Pre-transplant biopsy.

Fig. 1

Variable selection using the least absolute shrinkage and selection operator (LASSO) binary logistic regression model. (A) Tuning parameter (λ) selection in the LASSO model used 10-fold cross-validation with the minimum criteria. The binomial deviance was plotted versus log(λ). Dotted vertical lines were drawn at the optimal values by using the minimum criteria and the 1-standard error of the minimum criteria (the 1-SE criteria). A λ value of 0.012, with log (λ), -4.423 was chosen (1-SE criteria) according to 10-fold cross-validation. (B) LASSO coefficient profiles of the 28 variables. A coefficient profile plot was produced against the log(λ) sequence. A vertical line was drawn at the value selected using 10-fold cross-validation, where optimal λ resulted in 12 nonzero coefficients (except for the intercept)

Fig. 2

The developed preoperative DGF prediction nomogram. The nomogram was constructed using the data from the development cohort. Ten variables with nonzero coefficients selected by LASSO regression were presented. Terminal BUN level, initial perfusion resistance, WIT, and Banff score were entered into the logistic model as continuous variables, while the other variables were entered as dichotomous variables

Fig. 3

Predictive performance of the preoperative DGF prediction nomogram. (A) and (B) show the ROC curves of the nomogram in the development and validation cohorts, respectively. (C) and (D) present the calibration curves of the nomogram in the development and validation cohort, respectively. The calibration curve shows the calibration of the nomogram in terms of the agreement between the predicted risk of DGF and the observed risk of DGF. The 45° dotted line represents a perfect prediction, and the solid line represents the predictive performance of the nomogram. The solid line shows a closer fit to the dotted line, which indicates better predictive accuracy of the nomogram

Fig. 4

Decision curve analysis for the DGF prediction nomogram. The y-axis measures the net benefit. The red line represents the DGF prediction nomogram. The gray line represents the assumption that all recipients have DGF. The black line represents the assumption that no recipients have DGF. The net benefit was calculated by subtracting the proportion of all patients who are false positive from the proportion of those who are true positive, weighting by the relative harm of forgoing treatment when compared with negative consequences of an unnecessary treatment. The decision curve showed that if the threshold probability of a patient or doctor is > 5% and < 70%, the use of the nomogram in the current study to predict DGF adds more benefit than that achieved with the treat-all-patients scheme or the treat-none scheme