Objectives: We hypothesized that the combination of APACHE II and Model for End-Stage Liver Disease systems would work satisfactorily in patients admitted to intensive care unit after living-donor liver transplant.
Materials and Methods: Data were retrospectively collected from the database of our surgical team. The study included 38 patients (hepatitis B virus cirrhosis, 47.4%; hepatocellular carcinoma, 28.9%; other diseases, 23.7%). Laboratory values were obtained. Vital signs, Glasgow Coma scale scores, and urine output were abstracted. Variables included age, sex, acute physiology score, APACHE II score, APACHE II-predicted intensive care unit and hospital mortality, predicted length of intensive care unit, and hospital stay. Patients’ actual length of intensive care unit and hospital stays, intensive care unit and hospital discharge status, and discharge location were recorded. Standardized mortality ratios were calculated. Discrimination and calibration of APACHE II were assessed. All patients were divided into 3 groups: Model for End-Stage Liver Disease score: >25, 18 to 25, and <18. Predicted hospital mortality was calculated and compared.
Results: Mean APACHE II scores of survivors and nonsurvivors were 13.03 and 23.67. Mean risk of death was 7.05% and 25.07%. APACHE II scores and risk of death between survivors and nonsurvivors was significantly different (P < .001). The cutoff value of APACHE II score and Model for End-Stage Liver Disease score in the receiving operating characteristic curve was 20 and 25. Patients with APACHE II scores greater than 20 or Model for End-Stage Liver Disease scores greater than 25 had higher predicted hospital mortality after living-donor liver transplant.
Conclusions: The modified APACHE II model provides an accurate prognosis of patients receiving a living-donor liver transplant. The combined application of Model for End-Stage Liver Disease score and APACHE II score can improve the predictive accuracy.
Key words : Retrospective study, Intensive care unit, Posttransplant outcome
The intensive care unit (ICU) plays a vital role in liver transplant. Several scoring systems have been used to quantify the severity of illness of ICU patients and to predict their survival. Prognostic systems have been used to justify developing progressive care units by identifying a group of ICU patients at low risk for mortality.1-3 Such systems also may provide objective assessment for developing ICU discharge criteria and may identify those patients likely to require ICU readmission.4 Both of these issues were of interest to our surgical team.
This study sought to validate the Acute Physiology and Chronic Health Evaluation (APACHE) II and the Model for End-Stage Liver Disease (MELD) scoring systems in a cohort of patients admitted to the ICU after living-donor liver transplant (LDLT), especially to assess the effect of the addition of the MELD score on the predictive ability of APACHE II score.
Materials and Methods
Thirty-eight consecutive patients (7 females, 31 males; mean age, 39 ± 16 y; age range, 4-66 y;) who underwent LDLT between June 2006 and May 2009 were identified in the database of our surgical team. All grafts were ABO compatible and graft types were composed of 32 right lobes, 5 left lobes, and 1 dual graft of the right lobe and the left lobe. The indications for LDLT included hepatitis B virus cirrhosis in 18 recipients, hepatocellular carcinoma in 11 recipients, Wilson disease in 6 recipients, biliary cirrhosis in 1 recipient, sclerosing cholangitis in 1 recipient, and multiple intrahepatic stones with liver fibrosis in 1 recipient. According to the Child-Pugh classification, 12 patients were classified as Child C, 19 as Child B, and 7 as Child A. This study was approved by the Ethics and Indications Committee of Shanghai Jiao Tong University and performed according to the principles outlined in the 1975 Declaration of Helsinki. Written informed consent was obtained from each patient.
Measurement of main outcomes and variables
For each patient, APACHE II score was calculated as follows: (1) Original APACHE II: Ln (risk of death [ROD]/1-ROD) = -3.517 + (APACHE II score × 0.416) -0.6135,6; and (2) the modified APACHE II: Ln (ROD)/1-ROD = -3.517 + (APACHE II score × 0.416) -1.076.7 Data were collected for the first ICU admission after LDLT. Data for ICU readmissions were not used. The laboratory values used for the APACHE II scoring were computerized. Vital signs, Glasgow Coma scale scores, and urine output were abstracted according to a formalized protocol. The abstracted data included age, sex, acute physiology score, APACHE II score, APACHE II-predicted ICU and hospital mortality, and predicted length of ICU and hospital stay. The acute physiology score for each patient was calculated as described previously.5 In addition, patient actual length of ICU and hospital stays, and ICU and hospital discharge status and discharge location were recorded. In addition, data regarding duration of anesthesia, surgery, and intraoperative blood loss were recorded. The patients were divided into 3 groups: MELD score: > 25, 18 to 25, and < 18. The sum of ROD value of each patient divided by the total number of patients equals to the predicted hospital mortality.5 Predicted hospital mortality in these 3 groups was calculated and then compared.
Descriptive data were summarized as means ± standard deviation. A chi-square analysis was used to compare categorical variables, and the t test and rank sum tests were used to compare continuous variables. Statistical tests were 2-tailed and were considered statistically significant with P values < .05. Standardized mortality ratio was calculated by dividing the observed rate by the predicted rate. Discrimination of a prognostic model was the ability of the model to distinguish between survivors and nonsurvivors. Discrimination of the APACHE II-predicted mortality for prediction of in-hospital mortality was analyzed by calculating the area under the receiver operating characteristic (ROC) curve (AUC).8 Calibration of a model is the degree of agreement between predicted mortality and actual mortality. The Hosmer–Lemeshow test was used to determine calibration of the model. A model with good calibration should have a Hosmer–Lemeshow test score close to the degrees of freedom, which was equal to the number of categories minus 2, and a P value > .05. The Brier score was used as an overall assessment of the model’s performance, in which a lower score indicated a better performance. Statistical analyses were performed with SPSS software (SPSS: An IBM Company, version 13.0, IBM Corporation, Armonk, NY, USA).
Differences in APACHE II scores, including original and modified ROD value of APACHE II model between survivors and nonsurvivors was statistically significant (P < .001) (Table 1). Six patients died during hospitalization; 2 died of small-for-size syndrome. The observed hospital mortality was 15.79%. When we assessed the prognosis of patients with the APACHE II model, small-for-size syndrome became an important surgical factor. Therefore, we removed the 2 patients who died of small-for-size syndrome to predict the prognosis more accurately. After removing the 2 cases, there were 4 patients that died, and the observed hospital mortality was 11.11%. The predicted hospital mortality of the original APACHE II model and the modified APACHE II model was 14.37% and 9.90% (Figure 1).
The 36 patients were divided into 3 groups according to MELD scores (Table 2). There were 16 patients in group A (MELD < 18), of which 6 patients had liver cirrhosis, 8 had hepatocellular carcinoma, and 2 had hepatolenticular degeneration. The mean age was 44 ± 13 years old. Intraoperative blood loss was 4334 ± 3894 mL. Anhepatic phase was 75 ± 51 minutes. The mean MELD score was 14 ± 2. The APACHE II score was 13.13 ± 2.92. There were 9 patients in group B (MELD, 18-25), of which 3 patients had liver cirrhosis, 2 had hepatocellular carcinoma, 3 had hepatolenticular degeneration, and 1 had primary biliary cirrhosis. The average age was 31 ± 19 years. Intraoperative blood loss was 4961 ± 5302 mL. The anhepatic phase was 94 ± 46 minutes. The mean MELD score was 21 ± 2. The APACHE II score was 13.00 ± 4.54. There were 11 patients in group C (MELD > 25), of which 9 patients had liver cirrhosis, 1 had hepatolenticular degeneration, and 1 had intrahepatic bile duct stones with liver fibrosis. The average age was 38 ± 17 years. Intraoperative blood loss was 5783 ± 3855 mL. The anhepatic phase was 69 ± 29 minutes. The mean MELD score was 31 ± 4 minutes. The APACHE II score was 18.10 ± 5.88. The analysis showed that there was no significant difference between the APACHE II score and ROD value between groups A and B, but there was statistically significant difference between groups A and C, or B and C (P < .05) (Figure 2).
The ROC curve showed that area under the ROC curve of the APACHE II scoring system and the MELD scoring system was 0.996 and 0.863. Both had good diagnostic values. The cutoff value of the APACHE II score and the MELD score was 20 and 25 (Figure 3).
Liver transplant is the only effective treatment for end-stage liver diseases. Therefore, evaluating and predicting patients undergoing liver transplant is a critical issue. Several programs and applications of grading and scoring end-stage liver diseases have been proposed, but they are not satisfactory. The APACHE II system, an integrated method of evaluating the severity of illness, was proposed by Knaus.5 At present, this scoring system is applied to evaluate preoperatively and predict postoperatively the outcomes of patients undergoing an orthotopic liver transplant (OLT).
Basile-Filho9 have indicated that the APACHE II scoring system is most effective in predicting early mortality in a retrospective study that included 63 patients who underwent an OLT. However, this system is rarely used to evaluate patients undergoing an LDLT. Only Chung’s10 retrospective study illustrates that the APACHE II score is correlated with mortality after an LDLT. Although the APACHE II scores between survivors and nonsurvivors were significantly different (P < .001), there was no classification diagnosis for an LDLT in the classification diagnosis list. Risk of death could be calculated only by using the classification diagnosis weight of gastrointestinal surgery (-0.613). However, it was discovered that predicting patient mortality using the APACHE II was too high for clinical practice. Therefore, Angus has proposed a classification diagnosis weight of -1.076 for OLT.6,7 Our study reveals that in predicting an LDLT, consistency between the modified predicted hospital mortality (9.9%) and the observed hospital mortality is superior to that between the originally predicted hospital mortality (14.37%) and the observed hospital mortality. Thus, the modified predicted hospital mortality has a better predictive accuracy.
The MELD score has previously been used to assess and predict survival of patients with end-stage liver disease who are waiting for liver transplant; it also has been used as selective criteria for an OLT. In a retrospective study of 44 patients admitted for an OLT, Basile-Filho indicate that the MELD scoring system is applicable in predicting early mortality after OLT.11 The dynamic changes in the MELD score have predictive effect on survival after an OLT and serve as an easily applicable tool for temporarily withholding deteriorating patients and allocating available organs to recipients with a higher chance for posttransplant survival.12 In conflict, the MELD score has been reported to have a low overall accuracy for predicting posttransplant survival of patients undergoing an OLT.13 Therefore, using the MELD scoring to predict postoperative prognosis is debatable.
In this study, we showed that in the group with a MELD score less than 18 and the group with a MELD score of 18 to 25, the APACHE II scores and the ROD values were not significantly different, and the predicted hospital mortality was lower than 8%. However, in patients with a MELD score higher than 25, the APACHE II score and the ROD significantly increased, and the predicted hospital mortality was higher than 15%. In the ROC curve, the APACHE II and MELD scores had high diagnostic values in predicting patient outcomes. Patients with APACHE II scores higher than 20 or MELD scores higher than 25 had a significantly increased mortality after an LDLT. The combined application of the MELD and APACHE II scores would improve the predictive accuracy. The higher the MELD score is, the higher the ROD would be.
In conclusion, the APACHE II scoring system is a good predictive system for an LDLT. Patients with APACHE II scores higher than 20 or MELD scores higher than 25 have greater predicted hospital mortality after an LDLT, and the combined application of the 2 scoring systems can improve the predictive accuracy. However, whether the combined application of these 2 scoring systems can predict patient outcomes in OLT is unknown. Further study is warranted to clarify this point.
Volume : 13
Issue : 1
Pages : 41 - 45
DOI : 10.6002/ect.2013.0289
From the 1Department of Surgery, Shanghai Jiao Tong University Affiliated Sixth
People’s Hospital; the 2Department of Surgery, Changzheng Hospital, Second
Military Medical University; and the 3Department of Surgery, Ruijin Hospital,
Shanghai Jiao Tong University School of Medicine, Shanghai, China
Acknowledgements: Zheng-Yun Zhang and Rui Chen contributed equally to this work. The authors have no conflicts of interest, and there was no funding for this study.
Corresponding author: Guang-Wen Zhou, Department of Surgery, Shanghai Jiao Tong University Affiliated Sixth People’s Hospital. No, 600, Yishan Road, 200233, Shanghai, China
Phone: +86 21 6436 9181
Fax: +86 21 6470 1361
Table 1. Comparison of APACHE II Score and ROD Between Survivors and Nonsurvivors
Table 2. Relevant Information of Patients in Different MELD Groups
Figure 1. Comparison of Observed Hospital Mortality and Predicted Hospital Mortality
Figure 2. Comparison of APACHE II and ROD Among MELD Groups
Figure 3. Receiver Operator Characteristic Plot of Hospital Mortality Predicted Using APACHE II and MELD Scoring Systems