Objectives: The purpose of this study was to assess and predict risk factors for death within 30 days after orthotopic liver transplant and to develop a nomogram to predict mortality after liver transplant.
Materials and Methods: We retrospectively studied 185 patients who underwent orthotopic liver transplant at Sichuan Provincial People’s Hospital from January 1, 2010, to December 31, 2018. Multivariable logistic regression analyses were used to identify independent risk factors. A nomogram model was developed to predict mortality after liver transplant. The perfor-mance of the prediction model was assessed and validated by receiver operating characteristic curve and bootstrap methods (1000 replications).
Results: Multivariable logistic regression analyses revealed that tracheal extubation time, postoperative infection, and intraperitoneal hemorrhage post-transplant were independent risk factors for mortality after liver transplant. The receiver operating characteristic curve of the nomogram prediction model was 0.896 (96% CI, 0.803-0.989), and the mean absolute error of internal validation by bootstrap (1000 replications) was 0.019 (n = 184). These results showed that the nomogram model had an excellent prediction accuracy.
Conclusions: A nomogram model can provide clinicians with an individualized risk assessment of perio-perative mortality in liver transplant recipients.
Key words : Hepatic, Prediction model, Risk factors, Transplant
Liver transplantation (LT) is currently the most effective treatment method for end-stage liver disease, including end-stage acute liver failure, cirrhosis in adults, hepatic malignancies, and congenital biliary atresia in children.1-4 With the continuous development of surgical techniques, perioperative management, and immunosuppressive therapy,5 survival can now reach 80% to 90% after 1 year and 70% to 80% at 5 years.6 However, as a result of shortages of donor organs, many patients die while on wait lists before undergoing LT.7 Therefore, a major challenge to physicians is to improve the success rate of LT and to utilize the living donors more effectively.8-10
Mortality after LT is highly prevalent within 30 days after transplant.11 Therefore, we retrospectively studied the risk factors associated with mortality within 30 days after LT. Although a number of studies have predicted the risk factors after LT,12-14 because of numerous factors affecting LT outcomes, precisely predicting postoperative mortality is challenging.
The nomogram uses various clinicopathological parameters to generate specific information that can affect clinical management or inform the clinician of the patient’s progress.15 A nomogram can evaluate outcomes of high-risk patients through a total score calculation, is intuitive, and is easy to understand. Nomograms are widely used in all aspects of clinical practice to predict the disease process.15,16 Although various nomograms have been established to estimate the incidence or risk of different diseases, as far as we know, nomograms are rarely used to predict mortality after orthotopic LT.
Materials and Methods
The study cohort included patients who received orthotopic LT at Sichuan Provincial People’s Hospital (Chengdu, China). All patient data were obtained from the hospital’s electronic medical record system. The study was approved by the Ethics Committee of Sichuan Provincial People’s Hospital (no. 2018-267) and has been registered with the Chinese Clinical Trial Registry (no. ChiCTR2000033742). Patient consent was not required for this study because it was a retrospective study without interventions that would affect patients’ rights and health. Patient data were anonymous at the time of collection.
This study was a retrospective analysis of 185 patients who underwent orthotopic LT at Sichuan Provincial People’s Hospital from January 1, 2010, to December 31, 2018. Exclusion criteria included blood group incompatibility between donor and recipient, patients who were pregnant or breastfeeding, those aged <18 years, patients with a previous history of liver surgery, multiple organ combination transplant recipients, patients with malignancy other than liver cancer, and patients with death in the operating room.
The preoperative variables selected for the study included recipient age, sex, body mass index (BMI), primary cause of the liver disease (liver cancer, severe hepatitis, chronic end-stage liver disease), nonhepatic phase duration, operative time, the volume of bicarbonate injection, the volume of saline injection, the volume of blood transfusions, total fluid volume, postoperative infection, postoperative intraperitoneal hemorrhage, Model for End-Stage Liver Disease (MELD) score, length of intensive care unit (ICU) stay, tracheal extubation (TE) time, and length of hospital stay.
Definitions in the model
This study examined mortality related to LT within 30 days after the operation, and the main prognostic outcome was survival or no survival. The diagnosis of postoperative infection was based on clinical manifestations and organism isolation.17 Intraperi-toneal hemorrhage after LT was diagnosed when there was a large amount of bloody drainage from the abdominal drainage tube, the hemoglobin level dropped rapidly, or there was instability in blood pressure.18,19
We performed all statistical analyses with IBM SPSS statistical software (version 25.0) and R software (version 3.4.6). Clinical features were summarized as categorical variables and presented as counts and percentages; continuous variables are presented as means and standard deviation or medians and interquartile range. Continuous variables were assessed by t tests, and categorical variables were assessed by chi-square tests. Statistical significance was defined as P < .05. Clinically relevant factors associated with nonsurvival (P < .05) in univariate analysis were included in subsequent multivariable logistic regression analyses to identify independent risk factors. The performance of the nomogram model was assessed by receiver operating characteristic curve (ROC)20 and bootstrap methods (1000 replications).
During the study period, 185 patients underwent LT. One patient died during LT. No primary nonfunction of the liver and no pulmonary embolisms
occurred in the patients. In total, 184 patients were included, with 143 men (77.7%) and 41 women (22.3%) analyzed in the study who had median age of 47 years. Indications for LT included end-stage liver disease (n = 90, 48.9%), liver cancer (n = 88, 47.8%), and severe hepatitis (n = 6, 3.3%). The main complications after LT were postoperative infection (n = 25, 13.6%) and postoperative intraperitoneal hemorrhage (n = 7, 3.8%). Compared with the survival group, the study indicators were significantly more prevalent in the nonsurvival group. The significantly prevalent indicators in the nonsurvival group included TE time (P < .001), postoperative infection (P < .001), postoperative intraperitoneal hemorrhage (P = .001), ICU length of stay (P = .003), hospital length of stay (P = .007), 0.9% saline volume (P < .001), blood trans-fusion volume (P = .029), total fluid volume (P = .016), BMI (P = .017), and MELD score (P = .012) (Table 1).
Mortality risk factors
In the multivariable logistic regression analyses, risk factors for LT mortality included TE time, ICU length of stay, hospital length of stay, postoperative infection, postoperative intraperitoneal hemorrhage, saline injection volume, blood transfusion volume, and total fluid volume. Multivariable logistic regression analyses showed that TE time, postoperative infection, and postoperative intraperitoneal hemorrhage were independent risk factors for nonsurvival (Table 2).
Nomogram model development and validation
A nomogram to predict mortality after LT was developed based on 3 independent risk factors identified by multivariate logistic regression analysis (Figure 1). For example, for a patient with TE at 5 days after LT with postoperative infection and without postoperative intraperitoneal hemorrhage, the total score was 102 (56 + 46 + 0) and the corresponding risk of mortality post-LT was 70% (Figure 2).
A ROC curve was used to assess the predictive ability of the model for death in LT patients, and the result showed an area under curve of 0.896 (96% CI, 0.803-0.989) (Figure 3). The internal calibration was constructed using bootstrap resampling 1000 times, and the results showed high consistency in the nomogram (Figure 4).
Mortality after LT is the consequence of multiple interacting factors and cannot be explained by a single etiological factor.21 According to our results, we found hospital length of stay, ICU length of stay, BMI, MELD score, TE time, postoperative intraperitoneal hemorrhage, and postoperative infection to be risk factors for mortality after LT. Among them, postoperative infection, postoperative intraperitoneal hemorrhage, TE time, and BMI were the most independent prognostic factors for mortality after LT. Although our study showed that BMI was an independent factor for mortality after LT, whether BMI is a risk factor for LT remains controversial.22,23 Several studies have shown that post-LT outcomes of morbidly obese (BMI ?40) patients are similar in terms of patient and graft survival outcomes compared with outcomes of nonobese patients.22,24-26 Therefore, BMI was excluded as a risk factor for death after LT in our study. Thus, we developed a nomogram model based on 3 factors to accurately predict mortality risk after LT.
Postoperative infection remains the most common complication and mortality factor after LT.22,23 We also found that postoperative infection was a major cause of mortality after LT. We observed that 55% of patients in the nonsurvivor group had postoperative infections compared with 8.5% of patients in the survivor group (P < .001). The cause of post-LT infections may be related to the patient’s usual immunocompromised status secondary to immunosuppressive therapy, poor liver function, and clinical conditions. Patients whose liver function does not recover after LT are more likely to be infected.27,28
Postoperative intraperitoneal hemorrhage is another potentially fatal complication after LT that usually requires emergency surgical treatment.19,29 Several studies have demonstrated low short-term survival in patients with postoperative intraperitoneal hemorrhage.19,29,30 We observed that postoperative intraperitoneal hemorrhage was an independent risk factor for mortality after LT. The following factors may increase the incidence of postoperative intrape-ritoneal hemorrhage. During LT, the graft may develop an ischemia-reperfusion injury, which may lead to graft loss or primary failure.31 In addition, the recipient may develop a hypocoagulable state after LT. Early thrombocytopenia after LT is common due to platelet activation and depletion after graft reperfusion.32 Moreover, the coagulation changes in LT patients are extremely complex, and patients may develop coagulopathies postoperatively.33
After LT surgery, patients are returned to the ICU with endotracheal intubation. Longer intubation times are associated with higher complications, including ventilator-associated pneumonia and increased mortality.34 Glanemann and colleagues35 concluded that LT patients did not benefit from mechanical ventilation. In the setting of mechanical ventilation, increased pulmonary vascular resistance due to lung inflation may increase right ventricular afterload, and the inferior vena cava and hepatic vein reflux caused by related tricuspid regurgitation may produce venous congestion.36 Conversely, decreased intrapleural pressure in spontaneously breathing patients can improve venous return, including hepatic blood flow. Therefore, spontaneous ventilation can improve hepatic venous return and liver graft circulation, thereby enhancing liver graft recovery.37 This is consistent with our study that TE time was a risk factor in patients after LT surgery.
Although Child-Pugh classification is commonly used for LT, there are some problems. First, the extent of ascites and encephalopathy are both subjective assessments by physical examination alone. Second, both ascites and encephalopathy may be affected by treatments such as diuretics, albumin infusions, and lactulose, and it is unclear whether ascites and encephalopathy are scored best or worst or independent of specific treatments.38,39 Several studies have demonstrated that Child-Pugh score was not a predictor of mortality in patients after LT.40 To reduce subjective assessments, we did not use Child-Pugh in this study but instead used MELD score.
The study has several limitations. First, the cohort of patients was from a single center with potential selection bias, including ethnicity, graft size, living donor age, surgical techniques, and surgical indications. Second, more than 30% of data for cold ischemia time (CIT) were missing for this study, so we excluded this variable. However, all LT procedures were performed according to national guidelines for donation after cardiac death in China,41 for which CIT is less than 12 hours. Also, it is still controversial whether CIT of within 12 hours has a significant effect on LT survival.42-44 Moreover, the preservation method plays a key role in the consequences of CIT43; older donor age and CIT have a cumulative effect on LT survival.45,46 Therefore, the absence of CIT did not affect the results of our study. Finally, the relatively small sample size may also limit the generality of our conclusions. Therefore, the nomogram is likely to be influenced by single-center perioperative care and clinical management, and multicenter validation is required in future studies.
Our results suggest that reducing TE time, pos-toperative infection, and intraperitoneal hemorrhage during the perioperative period of LT can signifi-cantly improve patient outcomes. In addition, the nomogram model that we developed to predict mortality risk in LT shows good distinguishing power in predicting the prognosis after LT. Although it cannot replace the clinician’s judgment of the prognosis, it can provide a convenient tool for individualized assessment of mortality in patients after LT.
Volume : 20
Issue : 12
Pages : 1099 - 1104
DOI : 10.6002/ect.2021.0431
From the 1Information Center, West China Hospital, Sichuan University, Chengdu, Sichuan Province, China; the 2Department of Medical Informatics, West China Medical School, Chengdu, Sichuan Province, China; the 3Department of Anesthesiology, Sichuan Provincial People’s Hospital, University of Electronic Science and Technology of China, Chengdu, China; and the 4Chinese Academy of Sciences Sichuan Translational Medicine Research Hospital, Chengdu, China
Acknowledgements: The authors have not received any funding or grants in support of the presented research or for the preparation of this work and have no declarations of potential conflicts of interest. *Jialin Liu and Jiacen Li contributed equally to this work.
Corresponding author: Xi Yang, Department of Anesthesiology, Sichuan Provincial People’s Hospital, Chengdu, Sichuan Province, China
Table 1. Clinicopathologic Characteristics of Liver Transplant Recipients
Figure 1. Nomogram Prediction Model for Mortality After Liver Transplant
Table 2. Univariate and Multivariable Analyses of Mortality Risk Factors After Liver Transplant
Figure 2. Example of Nomogram Prediction
Figure 3. Receiver Operating Characteristic Curve for the Nomogram
Figure 4. Calibration Curve of the Nomogram