A risk-score prediction of postoperative acute kidney injury following lung transplantation: a retrospective cohort study

Abstract

Background

Acute kidney injury (AKI) following lung transplantation (LTx) is correlated with high mortality rates. We aimed to establish a risk-score model for AKI prediction in LTx.

Methods

We retrospectively reviewed data from the Institutional Lung Transplant Database from 2016 to 2022. The primary endpoint was to establish a risk-score model to predict AKI. The secondary endpoint was the impact of AKI on postoperative rehabilitation and survival incidence at the 1-year follow-up.

Results

Of 415 patients, 27% (n = 112) developed AKI within 48 h after LTx. Multivariable analysis revealed that body mass index, diabetes, plasma infusion, surgical time, and postoperative extracorporeal membrane oxygenation (ECMO) assistance were risk factors for AKI. This risk score was created and calibrated based on these five factors, ranging from 0 to 16 points, with the associated prediction of postoperative AKI ranging from 3 to 99% (Hosmer–Lemeshow χ² = 7.502; P = 0.484). Good discrimination was verified by developing and validating the datasets [Area Under the Curve (AUC) = 0.813 and 0.782, respectively]. Based on score distribution, patients were classified into three risk levels: low risk (0–3), moderate risk (3–7), and high risk (7–16). AKI is associated with prolonged stay length of intensive care unit and postoperative hospital (P < 0.001 and P = 0.003), and has an impact in the 3-to-6-month survival (P = 0.008 and P = 0.006).

Conclusions

A risk-score model based on perioperative variables effectively predicted the risk of AKI within 48 h after LTx. This model may be useful in early decision-making regarding AKI treatment.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13019-025-03782-3.

Background

Lung transplantation (LTx) is the only available therapy for treating end-stage lung disease to extend patient lifespans and improve their quality of life; however, post-transplant complications continue to lead to significant mortality and morbidity [1].

Acute kidney injury (AKI) is one of the common complications of LTx. Etiology and comorbidities are used to determine AKI stage and types, and mortality rates vary [2, 3]. The occurrence of early postoperative AKI increases mortality by six times [4, 5]. Identifying perioperative predictive factors for AKI is crucial for preoperative risk stratification [6] and developing a risk-factor predictive model for early intervention in AKI.

Several studies have described AKI after LTx and summarized its risk factors; however, few have focused on constructing predictive models. AKI after LTx is associated with different transplant procedures, perioperative medical conditions, and post-transplant clinical settings. Independent predictors of AKI have been identified [5, 7–9], but they are based only on preoperative characteristics or are studied in a relatively small population [6, 10]. Grimm ever established the prediction model according to the preoperative parameters [6]. But the intraoperative changes, including the amount of bleeding, the surgical duration, and the intraoperative extracorporeal membrane oxygenation (ECMO) assistance, even other potential influential factors, have a more direct impact on postoperative AKI. It is important to identify intraoperative factors that lead to postoperative AKI. Additionally, the treatment methods and drug regimens in lung transplantation have changed over the past 10 years. Therefore, it is necessary to establish a current risk-score prediction based on perioperative conditions.

We carried on a retrospective investigation of LTx in a single center over 6 years. The primary endpoint was to establish a complete and comprehensive risk-score prediction for the detection of AKI within 48 h after LTx. The secondary endpoint was the impact of AKI on postoperative rehabilitation, and survival incidence at the 1-year follow-up.

Methods

Study design

Patients were included in this retrospective cohort study, who underwent LTx from January 1, 2016 to July 31, 2022 at the First Affiliated Hospital of Guangzhou Medical University. The study was authorized by our ethics committee (approval number: ES-2023-K004-01). Due to the retrospective characteristics, the necessity of informed consent was waived.

The inclusion criteria were: adults aged 18 years or older who underwent LTx (single or double LTx). The exclusion criteria were: (i) re-transplantation, (ii) multi-organ transplantation, (iii) donation due to cardiac death, (iv) donor of ex vivo lung perfusion (EVLP), (v) preoperative renal replacement therapy (RRT) due to chronic kidney disease, and (vi) incomplete information.

Data collection

All LTx procedures were performed according to previously reported standard techniques [11, 12]. Patient variables were acquired from electronic medical records. The preoperative and intraoperative variables were included to establish the risk score, such as age, sex, body mass index (BMI), comorbidity, primary disease, Acute Physiologic Assessment and Chronic Health Evaluation (APACHE), left ventricular ejection fraction (LVEF), preoperative mechanical ventilation and extracorporeal membrane oxygenation (ECMO) support, types of transplantation, values of mean arterial pressure (MAP) and pulmonary artery systolic pressure (PASP) after anesthesia, intraoperative ECMO support, surgical time, transfusion volume, blood lose, preoperative values of white blood cell (WBC) and hemoglobin (HB), b-type natriuretic peptide precursor (pro-BNP), albumin, blood gas analysis, blood urea nitrogen (BUN), and serum creatinine (Scr). Postoperative ECMO support significantly affects the development of AKI. LTx patients will be decided whether to continue ECMO assistance at the end of surgery immediately. Therefore, postoperative ECMO support was included as an analysis factors in the AKI model.

Postoperative rehabilitation, complications within postoperative seven days, and 1-year survival were assessed. The laboratory findings included preoperative data from the day closest to the transplantation day as well as data from 48 h after surgery. The standards for perioperative ECMO support and ECMO weaning complied with the ELSO guidelines [11, 13].

Postoperative complications were mainly diagnosed by laboratory or imaging examinations. Cardiac-dysfunction meant LVEF < 50%, acute rejection was diagnosed by dyspnea and sputum, wtih significant decreasing PaO₂, and pleural effusion or infiltrative shadows in chest radiograph.

AKI definition and outcomes of study

The determination of AKI was on the basis of the Kidney Disease: Improving Global Outcomes (KDIGO) criteria [14], which specifies AKI as SCr increasing 26.5 µmol/L within 48 h. AKI is divided into three stages when the values compared to baseline: Stage 1, SCr increasing by 1.5–1.9 times; Stage 2, SCr increasing by 2.0–2.9 times; Stage 3, SCr increasing by 3.0 times or to 353.6 µmol/L or requiring RRT.

The primary endpoint was the establishment of a risk-score prediction for postoperative AKI after LTx. The secondary endpoint aimed to investigate the impact of AKI on postoperative rehabilitation, and the survival incidence within 1-year follow-up.

Statistical analysis

Normal distribution data were represented as mean and standard deviation, and compared by Student’s t-test. Continuous data with non-parametric dispersion were showed with median (interquartile ranges) and evaluated by Mann–Whitney U-test. Dichotomous variables were expressed by percentages of the total number (%) and analyzed with Fisher’s exact test. A P-value < 0.05 was regarded as statistically significant. The survival incidence was evaluated by Kaplan–Meier statistics and the statistical significance was assessed by log-rank analysis. If the missing data was less than 10%, it will be replaced by the mean value, otherwise, the variable would not be included in the modeling.

The preoperative demographic characteristics and intraoperative variable quantities, with univariate significance level of P < 0.05 were first identified using univariable binary logistic regression. These statistically significant variables were then entered into a multivariate logistic regression model to identify independent risk factors to create the risk score model. The training and validation datasets were created by randomizing all patients in a 7:3 ratio. According to the training dataset, a risk model and score were established. A backward stepwise logistic regression model was selected to establish a risk score. All listed candidate predictive factors with significant (P < 0.05) in the univariable analysis and had clinical or biological significance were contained in the establishment of the risk model. Based on this model, a risk-score table was developed using a scoring method analogous to that described by Sullivan et al. [15].

In the training and validation datasets, the prediction accuracy of the risk score was evaluated by both discrimination, measured using calibration, and the Area Under the Curve (AUC), assessed by the Hosmer–Lemeshow chi-squared test and calibration plot. Subgroup validation (non-AKI and AKI groups) was conducted to further evaluate the stability of the risk score.

SAS (SAS Institute, version 9.4; Cary, NC) was applied to conduct the analyses. Figures were plotted with R software (R Foundation for Statistical Computing, version 3.6.2; Vienna, Austria). The Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis statement [16] was strictly applied in our study.

Results

Population characteristics

Out of 452 LTx patients, 415 met the inclusion criteria, and 112 patients (27%) developed AKI within postoperative 48 h. (Fig. 1)

Fig. 1.

The flowchart of this study. LTx, lung transplantation; AKI, acute kidney injury

Preoperative characteristics between AKI and non-AKI groups

In the AKI group, the BMI, the incidence of diabetes, and preoperative assistance with mechanical ventilation, and ECMO were significantly higher when compared to the non-AKI group (20.9 vs 19.73 kg/m², P = 0.005; 17 vs 7.6%, P = 0.005; 31.2 vs 11.2%, P < 0.001; 17 vs 4.6%, P < 0.001, respectively). In terms of primary disease, non-specific interstitial pneumonia (NSIP) (44.6%) was most common in the AKI group, followed by idiopathic pulmonary fibrosis (IPF) (21.4%), while the chronic obstructive pulmonary disease (COPD) (34%) and NSIP (35%) were common in the non-AKI group (P < 0.01). (Table 1).

Table 1.

Preoperative characteristics of patients

Variable	Non-AKI group (n = 303)	AKI group (n = 112)	P value
Median Age (y)	57.64 ± 11.37	55.50 ± 11.94	0.102
BMI (kg/m2)	19.73 ± 3.61	20.90 ± 4.05	0.005
Sex (n, %)			0.931
Male  Female	250 (82.5%) 53 (17.5%)	92 (82.1%) 20 (17.9%)
Comorbidity (n, %)
Coronary heart disease	25 (8.3%)	9 (8%)	0.944
Diabetes mellitus  Hypertension	23 (7.6%) 32 (10.6%)	19 (17%) 14 (12.5%)	0.005 0.577
Primary disease (n, %)			< 0.001
COPD  NSIP  Silicosis  Bronchiectasia  IPF  Lymphatic vessel disease  Occlusive bronchitis	103 (34%) 106 (35%) 19 (6.3%) 18 (5.9%) 41 (13.5%) 12 (4.0%) 4 (1.3%)	16 (14.3%) 50 (44.7%) 3 (2.7%) 7 (6.2%) 24 (21.4%) 10 (8.9%) 2 (1.8%)
APACHE	15.37 ± 3.14	15.13 ± 3.43	0.509
LVEF value%	69.98 ± 6.29	69.96 ± 5.95	0.985
Pre-op mechanical ventilation (n, %)			< 0.001
None  Yes	269 (88.8%) 34 (11.2%)	77 (68.8%) 35 (31.2%)
Pre-op ECMO (n, %)			< 0.001
None  Yes	289 (95.4%) 14 (4.6%)	93 (83%) 19 (17%)

BMI, Body mass index; NSIP, Nonspecific interstitial pneumonia; COPD, Chronic obstructive pulmonary disease; IPF, idiopathic pulmonary fibrosis; APACHE, Acute physiologic assessment and chronic health evaluation; LVEF, Left ventricular ejection fraction; ECMO, Extracorporeal membrane oxygenation; Pre-op mechanical ventilation, Pre-operative mechanical ventilation

Intraoperative conditions between AKI and non-AKI groups

In the AKI group, more double-lung transplantation procedures were performed (68.8 vs 40.6%, P < 0.001), PASP was higher (46.68 vs 42.94 mmHg, P = 0.013), and more patients required intraoperative ECMO assistance (68.8 vs 32%, P < 0.001). The surgical time was longer (455 vs 319 min, P < 0.001). The blood loss (1000 vs 550 mL, P < 0.001), infusion volume of crystalloids (1625 vs 1250 mL, P = 0.005), red blood cell (1550 vs 800 ml, P < 0.001), plasma (1000 vs 600 mL, P < 0.001), and 20% albumin (300 vs 200 mL, P = 0.002) were also higher in the AKI group (Table 2).

Table 2.

Comparison of intraoperative conditions

Variables	Non-AKI group (n = 303)	AKI group (n = 112)	P value
Types of transplantation (n, %)			< 0.001
Double lung transplantation  Right lung transplantation  Left lung transplantation	123 (40.6%) 84 (27.7%) 96 (31.7%)	77 (68.8%) 15 (13.4%) 20 (17.9%)
MAP (mmHg)	78.96 ± 12.13	80.75 ± 16.87	0.233
PASP (mmHg)	42.94 ± 12.56	46.68 ± 15.51	0.013
Intra-op ECMO (n, %)			< 0.001
None  Yes	206 (68.0%) 97 (32.0%)	35 (31.2%) 77 (68.8%)
Intraoperative vasoactive drugs			0.465
Norepinephrine ≤ 0.1ug/kg/min  Norepinephrine 0.1-0.3ug/kg/min  Dopamine ≤ 5ug/kg/min  Dopamine 6-10ug/kg/min	185 (61.0%) 22 (7.3%) 89 (29.4%) 7 (2.3%)	68 (60.7%) 22 (19.6%) 19 (17.0%) 3 (2.7%)
Surgical time (min)	319 (245, 445)	455 (370, 559)	< 0.001
Blood loss (ml)	550 (200, 1000)	1000 (500, 3000)	< 0.001
Crystalloids (ml)	1250 (500, 1750)	1625 (1000, 2438)	0.005
RBC infusion (ml)	800 (0, 1400)	1550 (725, 3100)	< 0.001
Plasma infusion (ml)	600 (0, 1000)	1000 (400, 1800)	< 0.001
20% albumin infusion (ml)	200 (0, 300)	300 (100, 500)	0.002
5% albumin infusion (ml)	0 (0, 400)	0 (0, 525)	0.765

MAP, Mean arterial pressure; PASP, Pulmonary artery systolic pressure; Intra-op ECMO, Intra-opearative extracorporeal membrane oxygenation; RBC, Red blood cell

Risk score validation and risk-score establishment

The percentage of missing data for each variable was about 1–5%. BMI, diabetes, surgical time, intraoperative plasma infusion, and postoperative ECMO assistance were selected as risk factors to establish a risk model (Table S1 and S2). In the development dataset, the risk score model was used for discrimination, with a high χ² statistic calibration of 7.502 (P = 0.484) and an AUC of 0.813 [95% confidence interval (CI): 0.760–0.865] (Fig. 2). When the risk score model was applied to the split sample from the validation datasets, the validation results were the same as those from the development datasets, accompanied by a high χ² statistic calibration of 7.435 (P = 0.491) and an AUC of 0.782 (95% CI: 0.696–0.868) (Fig. 2). The ROC curves for the development and validation datasets were similar (Fig. 2).

Fig. 2.

The area under the ROC curves and calibration plot with the Hosmer–Lemeshow test for goodness-of-fit result for the risk score. ROC, receiver operator characteristic

The receiver operating characteristic curves (Fig. 2) from the development and validation datasets coincided with AKI proportions (Fig. 3). The predicted AKI incidence corresponded well with the observed incidence (Fig. 3). Risk scores for the predictive factors were indicated in Table 3.

Fig. 3.

The risk score level and corresponding predicted risk

Table 3.

Risk scores for all predicting variables

Risk factors	Score
BMI
≤ 18.5  18.6–24  25–28	0 2 3
Diabetes
Yes  No	3 0
Surgical time, min
≤ 300  300–500  > 500	0 2 3
Intraoperative plasma infusion, ml
≤ 1000  1000–3000  > 3000	0 0 5
Post-op ECMO assistance
Yes  No	2 0

BMI, Body mass index; Post-op ECMO, Post-operative Extracorporeal membrane oxygenation

Predictive grading of the risk score model

According to the score distribution, the risk scores were graded into three levels of prediction probability for AKI after LTx: low-risk, 3–12% (0–3 points]; moderate-risk, 13–55% (3–7 points]; and high-risk, 56–99% (7–16 points] (Fig. 3).

Impact of AKI on post-operative outcomes

Patients in the AKI group demanded more postoperative ECMO assistance (53.6 vs 21.1%, P < 0.001). The stay lengths of ICU, post-hospital, and total hospital were longer (17.5 vs 9 days, P < 0.001; 35.5 vs 27 days, P = 0.003; and 45.5 vs 35 days, P = 0.002, respectively). The amount of patients without postoperative complications was lower in the AKI group (33.9 vs 58.1%, P < 0001), with more hemorrhage (10.7 vs 3.3%), higher pneumonia (15.2 vs 10.9%), more cardiac-dysfunction (8.9 vs 3.6%), and more pneumothorax (6.2 vs 2.0%) in the AKI group (Table 4).

Table 4.

Impact of acute kidney injury on Post-LTx outcomes

Variable	Non-AKI group (n = 303)	AKI group (n = 112)	P value
Post-op ECMO (n, %)			< 0.001
None  Yes	239 (78.9%) 64 (21.1%)	52 (46.4%) 60 (53.6%)
ICU stay (day)	9 (6, 16)	17.5 (10, 34.75)	< 0.001
Post-op hospital stay time (min)	27 (20, 41)	35.5 (23, 55)	0.003
Total hospital stay time (min)	35 (24, 56)	45.5 (31, 75.75)	0.002
3 months of survival (n, %)			0.008
Death  Living	49 (16.2%) 254 (83.8%)	31 (27.7%) 81 (72.3%)
6 months of survival (n, %)			0.006
Death  Living	63 (20.8%) 240 (79.2%)	38 (33.9%) 74 (66.1%)
12 months of survival (n, %)			0.233
Death  Living	100 (33%) 203 (67%)	44 (39.3%) 68 (60.7%)
Complications within post-op 7 days (n, %)			< 0.001
None  Hemorrhage  Acute rejection  Pneumonia  Cardiac-dysfunction  Arrhythmias  Pneumothorax  Pleural effusion  Delirium  Death  Reoperation  Thrombus	176 (58.1%) 10 (3.3%) 22 (7.3) 33 (10.9%) 11 (3.6%) 21 (6.9%) 6 (2.0%) 11 (3.6%) 6 (2.0%) 4 (1.3%) 3 (1.0%) 0	38 (33.9%) 12 (10.7%) 7 (6.3%) 17 (15.2) 10 (8.9%) 5 (4.5%) 7 (6.2%) 3 (2.7%) 2 (1.8%) 7 (6.2%) 3 (2.7%) 1 (0.9%)

Post-op ECMO, Post-opearative Extracorporeal membrane oxygenation; ICU, Intensive care unit

Impact of AKI on survival within 1 year

AKI at postoperative 48 h was related to significantly decreased survival at postoperative 3 months (72.3 vs 83.8%, P = 0.008), and 6 months (66.1 vs 79.2%, P = 0.006). The survival rate showed no difference at postoperative 12 months between groups (60.7 vs 67%, P = 0.233). In total, 335 patients survived and 80 died by postoperative 3 months, 314 patients survived and 101 died by postoperative 6 months, and 271 patients survived and 144 died by postoperative 12 months (Table 4; Fig. 4).

Fig. 4.

Kaplan–Meier survival curves after LTx according to the incidence of postoperative AKI in follow-up after LTx. LTx, lung transplantation; AKI, acute kidney injury

Distribution of AKI stage postoperatively at 48 h

At 48 h postoperatively, 112 (27%) patients developed AKI. AKI stage 1 was observed in 65 (15%) patients, AKI stage 2 in 28 (7%), and AKI stage 3 in 19 (5%) (Fig. S1).

Correlation of ECMO and AKI in postoperative 48 h

At 48 h postoperatively, 19 (57.6%) of 33 patients who required preoperative ECMO developed AKI; 77 (44.3%) of 174 requiring intraoperative ECMO assistance developed AKI; and 60 (48.4%) of 124 needing postoperative ECMO developed AKI. The correlation between perioperative ECMO and AKI at postoperative 48 h was significant (P < 0.001) (Table S3).

Laboratory findings

In the AKI group, the preoperative values of BUN and eGFR were higher than those in the non-AKI group (5.75 vs 5.10 mmol/L, P = 0.033; and 89.34 vs 79.10 µmol/L, P = 0.001), while those of Hb, PaO₂, and P/F were lower (112.98 vs 118.50 g/L, P = 0.039; 104.9 vs 114 mmHg, P = 0.046; and 171.97 vs 186.9, P = 0.046, respectively) (Table S4).

In the AKI group, the postoperative 48-h values of BUN and Scr were higher (18.6 vs 12.6 mmol/L, P < 0.001; and 122.85 vs 71.6 umol/L, P < 0.001), while those of eGFR and Hb were lower (45.84 vs 73.16 umol/L, P < 0.001; and 88.78 vs 95.47 g/L, P < 0.001). When compared with the non-AKI group, the PaO₂ of the AKI group was higher (128 vs 124 mmHg, P = 0.048), while the pH and PaCO₂ were lower (7.42 vs 7.44, P = 0.001; and 40.39 vs 43.0 mmHg, P < 0.001) (Table S5).

Discussion

This retrospective study established a risk score for AKI within 48 h after LTx. BMI, diabetes, surgical time, plasma infusion, and postoperative assistance with ECMO were risk factors for postoperative AKI. A score of 7–16 points in the risk model predicted a 56–99% risk of AKI. The patients in AKI group was associated with prolonged stay of intensive care unit and postoperative hospital, and has lower three and six months survivals.

The risk score model developed in this study consisted of five factors, showing high levels of variability and good calibration in predicting postoperative AKI. It can offer more scientific guidance for clinical decision-making in the early treatment of postoperative AKI. The risk score was graded into three levels mainly based on clinical relevance considerations. At low risk (0–3 points), the predicted risk of AKI at postoperative 48 h was 3–12%, which is relatively low. Attention must be paid to postoperative dynamic changes of renal function and urine volume to maintain hemodynamic stability and maintain renal perfusion. Moderate risk (3–7 points) indicated the predicted risk of AKI of 13–55%. In addition to closely monitoring the real-time clinical changes in renal function and urine volume, this requires the use of more renal-protective drugs, minimizing the use of nephrotoxic drugs and blood transfusion, and paying more attention to drug blood concentration to reduce renal injury. The predicted high risk of AKI at postoperative 48 h was 56–99% at 7–16 points. High-risk patients with AKI should consult nephrologists as early as possible because prompt participation and coordination of kidney disease treatment could facilitate the timely diagnosis and mitigation of postoperative AKI. In addition to the above precautions, the use of larger doses of diuretics after surgery and ensuring effective renal perfusion are necessary. Otherwise, the postoperative renal replacement therapy should be considered as early as possible. This risk score has been attempted to be implemented in our clinical practice. For obese patients with diabetes, as well as those requiring ECMO assistance after surgery, they were stratificated according to the above risk-score level and treated based on the above decision-making process. However, whether it can achieve the ultimate beneficial effect still requires prospective observation to verify.

Grimm established a predictive model just based on preoperative parameters [6]. However, perioperative changes, including the amount of bleeding, surgical duration, and postoperative ECMO assistance had a more direct impact on the risk of postoperative AKI. In addition, Grimm’s research comes from a large national database and is easily influenced by data input and variability in definitions of postoperative outcomes by different institutions [6]. Moreover, Grimm only analyzed the variables collected by the United Network for Organ Sharing; other potentially influential factors were not explored. Our study set the diagnosis of AKI according to the KDIGO definition and comprehensively considered the important preoperative and intraoperative factors to establish a prediction model that can accurately and effectively predict early AKI at 48 h postoperatively and is more consistent with clinical practice. The model’s validity is further reinforced by the inclusion of surgery-related parameters, which are crucial for accurate risk prediction. The endpoint time-window of AKI within postoperative 48 h was more likely to reflect the risk relationship between pre - operative and intra - operative related factors and the occurrence of AKI, thereby avoiding the influence of immunosuppressants, anti-rejection drugs, and antibiotics, as well as the impact of secondary infections.

BMI and diabetes mellitus were risk factors for AKI within 48 h postoperatively, conforming to previous reports [5, 17, 18]. These mechanisms may be associated with intraperitoneal hypertension, proinflammatory factors, and fluid infusion [19, 20]. Patients with high BMI secrete a large amount of pro-inflammatory cytokines in their adipose tissue, which exacerbate renal tubular injury and induce AKI under surgical stress [21, 22]. Diabetes leads to glomerular microvascular sclerosis, and the ability of renal blood flow autoregulation decreases. In addition, under hyperglycemia, the generation of reactive oxygen species increases, inducing apoptosis of renal tubular epithelial cells and mitochondrial dysfunction, exacerbating AKI [23]. Several studies have shown an association between cystic fibrosis and AKI [24–26], but there are conflicting reports on whether COPD or idiopathic pulmonary fibrosis (IPF) are more likely to lead to postoperative AKI [5, 24–27]. Our results showed that the incidence of postoperative AKI was lower in patients with COPD than in those with IPF and NSIP.

ECMO continuously exposes blood to non-endothelial and non-physiological interfaces, thereby promoting the activation of inflammatory cytokines and leading to renal injury [28, 29]. ECMO support lasting > 2 days [30, 31] or postoperative mechanical ventilation > 3 days [30, 32] are independent risk factors for AKI in patients after LTx. In this study, over 44.3% of patients with perioperative ECMO assistance developed AKI within 48 h postoperatively. Therefore, ECMO should be carefully selected in patients who have undergone LTx, especially those who need postoperative ECMO support. Postoperative ECMO support was a risk factor for AKI, which referred to the situation where ECMO support was determined based on the patient’s condition at the end of the surgery, and was continued from intraoperative ECMO to postoperative ECMO support.

Intraoperative fluid volume is also closely related to postoperative AKI. However, 26.7% of transplantation centers in European and non-European countries do not routinely use albumin [33]. The results showed that 300 mL of a high albumin concentration (20%) increased the incidence of postoperative AKI, whereas a low albumin concentration (5%) appeared to be unrelated to AKI, which coincided with the guide in non-European centers that 5% albumin is also the preferred colloid solution [33]. Our results showed a high intraoperative transfusion volume of RBCs and plasma, which may be due to the high incidence of intraoperative ECMO, which was as high as 41.9%. The results also revealed that the average blood transfusion in AKI cases was 1550 ml, confirming that blood transfusions of > 1000 ml of packed RBCs are associated with a higher incidence of AKI [34]. Therefore, lower albumin concentrations and more conservative blood transfusions may be recommended.

Surgical time was also a risk factor for AKI. More patients with AKI underwent bilateral lung transplantation (BLT) in this study. Jacques and Ishikawa showed that BLT was more likely to be associated with AKI [35, 36]. Of 200 BLT patients, 68.8% developed AKI within 48 h postoperatively, which reflected previous findings [5, 35, 37]. BLT requires a longer surgical time and more blood or colloid transfusion, resulting in fluid retention [38], a higher incidence of intraoperative hemodynamic fluctuation [5], and the need for extracorporeal circulation assistance, and may increase the physiologic stress and derangement associated with AKI. In addition, surgery may cause AKI by releasing inflammatory mediators that cause renal epithelial apoptosis and decrease renal perfusion [30]. These findings were also similar to those shown by Haase [39].

AKI was significantly associated with lower 3- and 6-month survival rates, which was cohered with previous reports on the relationship between AKI and short-term mortality after LTx [1, 35]; however, mortality within 1 year was not affected in this study, which required a longer follow-up to confirm.

The probability of AKI after LTx ranges from 9.4% to 85%, including the gamut of AKI severity from small SCr increases to acute dialysis [5, 34–37, 40–42]. This wide range may be owing to the adoption of different AKI definitions, different times of AKI determination [17, 43], and different transplant centers. The incidence of AKI within 48 h postoperatively was 27% in our study, lower than previously reported [34, 35, 40], which may be related to the early 48 h postoperative definition and rich experience of LTx in our center.

Study limitations

The study had several limitations worth considering. First, it was a retrospective study in a single-center, with the possibility of bias and confusion. However, a single academic medical center is conducive to the uniformity of the treatment process, measurement standards, and is more conducive to the summary of research objectives. Second, some factors, such as preoperative hydration status, and vascular complications, were not considered. However, APACHE, which is the most widely used and authoritative critical illness evaluation system, was used to evaluate patients before LTx. Third, it just excluded the patients with preoperative renal replacement therapy, the preoperative mild renal insufficiency was not excluded form the study. Since regardless of preoperative kidney function, as long as the changes of serum creatinine increasing within postoperative 48 h and meeting the above diagnostic criteria, AKI can be diagnosed.

Conclusions

BMI, diabetes, surgical time, plasma infusion, and postoperative assistance with ECMO were risk factors for postoperative AKI. This study established a risk score model according to preoperative and intraoperative variables to predict the risk of AKI at 48 h after LTx, which may be more consistent with clinical practice.

Abbreviations

AKI

Acute kidney injury

BMI

Body mass index

BLT

Bilateral lung transplantation

BUN

Blood urea nitrogen

COPD

Chronic obstructive pulmonary disease

ECC

Extracorporeal circulation

ECMO

Extracorporeal membrane oxygenation

eGFR

Estimated glomerular filtration rate

EVLP

Ex vivo lung perfusion

IPF

Idiopathic pulmonary fibrosis

ICU

Intensive care unit

KDIGO

Kidney Disease: Improving Global Outcomes

LTx

Lung transplantation

NSIP

Nonspecific interstitial pneumonia

PASP

Pulmonary artery systolic pressure

PGD

Primary graft dysfunction

P/F

PaO₂/FiO₂

RRT

Renal replacement therapy

RBC

Red blood cells

SCr

Serum creatinine

Author contributions

(I) Conception and design: Lan Lan, Furong Lin; (II) Administrative support: Xin Xu; (III) Provision of study materials or patients: Yingfen Li, Bu Long; (IV) Collection and assembly of data: Yahong Liu, Chao Yang, Yingfen Li; (V) Data analysis and interpretation: Yaoliang Zhang; (VI) Manuscript writing: Furong Lin; (VII) Final approval of manuscript: All authors.

Funding

This study was supported by Guangzhou Municipal Science and Technology Bureau, The Projcet of Basic and Applied Basic Research Jointly Funded by Municipality and University (Hospital) (Fund No. 202201020584).

Data availability

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

The study was authorized by our ethics committee (approval number: ES-2023-K004-01). Due to the retrospective characteristics, the necessity of informed consent was waived.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.