Patients and ethical approval

This retrospective study was approved by the Ethics Committee of the First Affiliated Hospital of Xi’an Jiaotong University (Shaanxi, China), No. XJTU1AF2015LSL − 058. All participants signed the informed consent form. The study protocol complied with the Declaration of Helsinki and Istanbul principles. Kidneys for transplantation were obtained from the Coordination Group of Shaanxi Red Cross Organization and harvested by the Organ Procurement Organization (OPO). No organs were harvested from executed prisoners. The immediate relatives of the donors voluntarily offered organ donation. Organ allocation was performed based on the China Organ Transplant Response System, and the process was kept as double-blind between donors and recipients.

Immunosuppressive regimen

On the day of transplantation surgery, all recipients were intravenously administered induction therapy by using rabbit anti-thymocyte globulin (r-ATG, 50 mg or 75 mg), ATG-Fresenius (ATG-F, 200–300 mg), or basiliximab (40 mg). The dose of r-ATG or ATG-F was tapered until discontinuation on postoperative day 5, and basiliximab 40 mg was provided again on postoperative day 4. Since the first day after transplantation, each recipient received the triple immunosuppressive regimen consisting of a mycophenolic acid drug (MPA: enteric-coated mycophenolate sodium or mycophenolate mofetil), a calcineurin inhibitor (CNI: tacrolimus or cyclosporine A), and prednisone.

Study design

The inclusion criteria for donors were as follows: (i) had clear identity and met the medical and ethical conditions for organ transplantation; (ii) had no history of kidney disease, drug abuse, and active infection diseases such as HIV and HBV; (iii) had no history of diabetes mellitus with severe complications; and (iv) had no history of malignant tumor.

Recipients were excluded if: (1) recipients who developed graft failure within 48 h of the transplant operation; (2) had a positive cross match or positive panel-reactive antibody (over 30%); (3) had an active infection, hepatitis, or abnormal hepatic function; or (4) had leukopenia (leukocytes < 3000/mm3), thrombocytopenia (platelets < 100,000/mm3), or severe anemia (hemoglobin < 60 g/L); (4) recipients who received re-transplantation or dual kidneys; (5) children’s kidney (6) recipients who received combined liver transplant; (7) Recipients with body mass index (BMI) < 28 kg/m2.

The donor scoring system included the donor’s age, primary disease, sCr levels prior to organ recovery, history of hypertension, CPR incidence and hypotension duration. The value of donor clinical scores in predicting graft performance was previously developed and validated from a thousand-patient cohort at our center [12]. Donors above 16 years of age with confirmed identity; with no history of kidney diseases, diabetes, drug abuse, and uncontrollable psychotic symptoms; who were not actively infected with hepatitis B and C viruses, human immunodeficiency virus, bacteria, and fungi; and in whom the isolated renal had a warm ischemia time (WIT) < 30 min and a cold ischemia time (CIT) < 12 h were included in the study. At least one kidney from each donor was used for single renal transplantation. Executed prisoners were excluded from the study.

DGF was considered as the primary outcome of this study, and it was defined as the need for dialysis [13] in the first week after kidney transplant surgery. Two independent datasets were used in this study, including the training cohort and the validation cohort. The training cohort was used to construct the predictive model and included 492 kidney transplant recipients between May 2018 and December 2019. The independent validation cohort was used to test the predictive model and included 105 recipients who underwent transplantation surgery between January 2020 and April 2020.

Variables and samples

Three transplant surgeons independently assessed the clinical characteristics of the included donors and recipients. The donor clinical characteristics included age, gender, body mass index (BMI), ABO blood type, cause of death (cardiac death), hypertension history (presence or absence), hypotension procedure (systolic pressure < 100 mmHg, presence or absence), cardiopulmonary resuscitation (CPR) procedure (presence or absence), terminal renal function (including serum creatinine (SCr) and blood urea nitrogen (BUN) levels), and urine volume. The recipient clinical characteristics included age, gender, BMI, primary disease, ABO blood type, dialysis method (hemodialysis or peritoneal dialysis) and duration, human leukocyte antigen (HLA) mismatch, pre-transplant panel reactive antibody (PRA) level (positive or negative), and type of induction therapy (r-ATG, ATG-F, or basiliximab).

After organ procurement, the kidneys were preserved by static cold storage (SCS) or hypothermic machine perfusion (HMP). HMP was performed using the LifePort Kidney Transporter machine (Organ Recovery Systems, Chicago, IL, USA). The initial pump pressure was set as 30–40 mmHg. The machine recorded the following five parameters: pressure, temperature, resistance, flow rate, and duration. The following characteristics of organ transport were recorded by OPO and transplant surgeons: transport method (SCS or HMP), machine perfusion parameters (initial and terminal pressure, flow rate, and resistance), cold ischemic time (CIT), and warm ischemic time (WIT).

Pre-implantation biopsies were performed by the transplant surgeon by using a 16G Bard needle. One sample was obtained from each kidney; fixed in formaldehyde; embedded in paraffin; sectioned; and stained with hematoxylin and eosin, periodic acid-Schiff, Masson’s trichrome, and silver methenamine. Each section was evaluated by two pathologists independently. Light microscopy was performed, and Banff 2022 classification [14] was used to evaluate chronic histopathological changes in the kidney. Acute changes, including acute tubular injury (ATI) and arteriolar smooth muscle vacuolar degeneration, were also noted. Each chronic or acute lesion was recorded on the scale of 0–3 points according to the degree of severity. Remuzzi score [15] and Banff score were calculated according to semi-quantitative chronic histopathological changes. All biopsies were performed pre-implantation and after conducting HMP/SCS.

Development of a nomogram model

Continuous variables are reported as mean ± SD (standard deviation), and categorical variables are reported as primary frequencies (percentages). HMP parameters, histological lesions, and donor and recipient clinical characteristics were analyzed using the Mann-Whitney U test or the chi-square test to determine significant differences between the DGF and non-DGF groups. Significant variables were included in the multivariate model. However, because of the small number of events relative to the number of factors and to obtain an optimal model with as few factors as possible, we used L1-penalized LASSO regression for multivariate analysis [16]; this was the first-step variable selection process. This logistic regression model penalizes the absolute size of the regression coefficients according to the lambda value. With larger penalties, the estimates of weaker factors shrink toward zero; consequently, only the strongest predictors remain in the model, and weaker predictors are excluded. To avoid overfitting models to idiosyncratic relationships in the training cohorts, the variable selection process used 10-fold cross-validation to select the optimal level of tuning or penalization, as measured by the Bayesian information criterion. A nomogram model was developed using variables with nonzero coefficients through multivariate logistic regression.

Validation and performance of the nomogram

Calibration curves were plotted to calibrate the prediction nomogram, accompanied with the Hosmer-Lemeshow (H-L) test. A nonsignificant test result (P > 0.05) implies that the calibration of the model is inaccurate. To assess the discrimination of the prediction nomogram, a receiver operating characteristic curve (ROC) was plotted, and area under the curve (AUC) values were determined. Calibration curves and ROC of the validation cohort were obtained to validate this nomogram.

Decision curve analysis (DCA) was performed to determine the clinical usefulness of the nomogram by quantifying the net benefits at different threshold probabilities in the validation cohort [17].

Pre-transplant biopsies

Pre-implantation biopsies were performed by the transplant surgeon using a 16-g Bard needle. Two biopsies were performed for each donor kidney. One piece of tissue was embedded for immunofluorescence staining, including IgA, IgM, IgG, C3, C1q, and fibrin-related antigens. Another biopsy tissue was fixed with formaldehyde, embedded in paraffin, sectioned and stained with hematoxylin and eosin, periodic acid Schiff, Masson trichrome, and hexamine silver. The donor kidney biopsy tissue contained at least 25 glomeruli, and Remuzzi score was immediately performed according to the rapid biopsy results [15]. Remuzzi’s method was used to evaluate the chronic histopathological changes of the donor kidney, and the ATI of the donor kidney was evaluated. According to Remuzzi scoring criteria, the degree of glomerular sclerosis, renal tubular atrophy, interstitial fibrosis and arterial lumen stenosis of the donors were evaluated by pathologists with a score of 0–3. All biopsies were performed before transplantation, but histopathological diagnoses were determined after transplantation to avoid potential selection bias based on histopathological findings.

Statistical analysis

Statistical analysis was conducted using R version 4.0.0 (www.Rproject.org) and GraphPad Prism v9.0. LASSO logistic regression was performed using the “glmnet” package. The packages “rms,” “pROC,” and “DecisionCurve” were used to plot the nomogram, ROC and AUC, and DCA, respectively. The H-L test was performed using the “generalhoslem” package. The reported significance levels were two-sided and set at 0.05.