SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES

Adult-to-adult living donor liver transplantation (ALDLT) has been accepted as an important option for end-stage liver disease, but information regarding the risk factors remains fragmentary. We aimed to establish a predictive model for 90-day survival. In the first step, a total of 286 cases who had received primary ALDLT using a right lobe graft between 1998 and 2004 were randomly divided into 2 cohorts at a ratio of 2:1 (191 vs. 95 recipients). The larger cohort of patients was used to develop a model. The outcome was defined as 90-day survival, and a total of 39 preoperative and operative variables, including the period of surgery (1998-2001 vs. 2002-2004), were included using Cox's proportional hazard regression model. Two mismatches of human leukocyte antigen (HLA) type DR (hazard ratio [HR] = 4.45; confidence interval [CI] = 1.96-10.1), loge[blood loss volume] (HR = 2.43; CI = 1.64-3.60), period of surgery (1998-2001 vs. 2002-2004) (HR = 2.41; CI = 1.04-5.57), and loge[serum C-reactive protein or CRP] (HR = 1.64; CI = 1.13-2.38) were found to be independent risk factors. In the second step, we tried to establish a realistic survival model. In this step, we created 2 models, 1 that used all 4 variables (model 1) and 1 (model 2) in which blood loss volume was replaced with the past history of upper abdominal surgery and Model for End-Stage Liver Disease (MELD) score (≥25), both of which showed associations with blood loss volume. These models were applied to the smaller cohort of 95 patients. Receiver operating characteristic analyses demonstrated that both models showed similar significant c-statistics (0.63 and 0.62, respectively). In conclusion, model 2 can provide a rough estimation of the 90-day survival after ALDLT. Liver Transpl 12:904–911, 2006. © 2006 AASLD.

Adult-to-adult living donor liver transplantation (ALDLT) has been established as a safe procedure for both donors and recipients,1–3 and is now accepted as the first option for end-stage liver disease in Japan. In Western countries, a recent chronic shortage of deceased donor livers has caused a bottleneck in liver transplants, which in turn has fueled the spread of ALDLT.4–6

For effective use of donated liver grafts, it is important to detect the preoperative risks associated with survival after liver transplantation. A number of prior studies have attempted to determine risk factors that can predict patient survival after deceased donor liver transplantation. We have experience with 286 cases of ALDLT using a right lobe graft in our institution since February 1998. At the present time, 78 cases have died, resulting in a mortality rate of 27.3%. The observation that the majority of mortality or graft failure (53.8%) occurs within the initial 90-day period strongly suggests that there are unexplored risk factors for short-term survival after ALDLT.

The major aim of the present study was to build a quantitative predictive model for ALDLT focusing on the 90-day survival. This goal was achieved via 2 steps. After dividing the total population into 2 cohorts at a ratio of 2:1, the larger cohort was used to develop the quantitative survival model and then the smaller cohort was used to test the model. In the second step, we created 2 models, 1 using all the variables and the other using only the preoperative variables. The results indicated that the 2 models were both able to predict the 90-day survival, indicating that the model using only the preoperative variables is sufficiently accurate for risk communication to patients.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES

Patient Population

Between November 1994 and October 2004, 410 adult recipients underwent ALDLT for end-stage liver disease at our institution. We limited the study sample to 337 patients (82.2%) who received a right lobe graft.

Indications for ALDLT included malignant neoplasm, viral hepatitis, primary biliary cirrhosis, and other etiologies. The follow-up periods ranged from 4 days to 83 months, with a median follow-up of 22 months.

Subsequently, 51 recipients (36 fulminant hepatic failures and 15 retransplantations) were excluded from the 337 right lobe recipients, such that a final total of 286 recipients were studied.

We randomly allocated the 286 cases into 2 cohorts at a ratio of 2:1: cohort A (n = 191) and cohort B (n = 95). Cohort A was examined to establish a survival model after ALDLT and cohort B was used to validate the model.

Operative Techniques and Postoperative Therapies

Each patient received a right lobe (right hemiliver) graft from a living donor as previously reported.7 The following vascular anastomoses were performed: the donor's right and middle hepatic veins to the recipient's right and middle hepatic vein remnants with a caval extension, the donor's right portal vein to the recipient's portal vein, and the donor's right hepatic artery to the recipient's hepatic artery. The bile duct was reconstructed by an anastomosis of the donor's bile duct to the recipient's bile duct. The immunosuppression protocol basically consisted of tacrolimus and low-dose steroids, and patients who experienced an acute rejection were treated with a bolus of methylprednisolone for 3 days as previously reported.8

Clinical Data

A comprehensive clinical and laboratory database was constructed retrospectively from the clinical charts of the patients. Age, gender, height, body weight, causes of liver disease, laboratory data in the last determination before ALDLT, preoperative therapy, past histories, and lifestyles were obtained from these clinical charts (Table 1), as were donor and operative factors. The Model for End-Stage Liver Disease (MELD) score was calculated using the serum total bilirubin, international normalized ratio for prothrombin time test, and serum creatinine, and the history of dialysis within 1 week according to the current United Network for Organ Sharing guidelines for liver transplantation.9 Intermittent hemodialysis was required for 3 cases before the transplantation and their creatinine levels were capped at 4 mg/dL.

Table 1. Variables Analyzed
  1. NOTE: The laboratory data and physical data were obtained just before the operation. The variables included 27 recipient factors, 4 donor factors, 3 organ factors, 4 operative factors, and the period of surgery.

  2. Abbreviations: BMI, body mass index; MELD, Model for End-Stage Liver Disease; AST, aspartate transaminase; WBC, white blood cell; CRP, C-reactive protein; PT/INR, Prothrombin time/international normalized ratio; HLA, human leukocyte antigen; GRWR, graft/recipient weight ratio.

A. Preoperative Factors
 I. Recipient
  Age, gender, BMI
  MELD score
  Cause of liver disease
  Laboratory data
   Serum creatinine, bilirubin, albumin, AST, WBC, CRP, PT/INR
   Blood culture
   Preoperative cultural screening (sputum, nasal discharge, urine, and stool)
  Physical data
   Encephalopathy, ascites, body temperature
   Preoperative therapy
   Diuretics, steroids
   Preoperative hemodialysis (intermittent hemodialysis or continuous hemodiafiltration)
   Past history
   Hypertension, diabetes mellitus, variceal bleeding, spontaneous bacterial peritonitis
   Upper abdominal surgery
  Lifestyle
   Smoking, alcohol consumption
 II. Donor
  Age, gender, BMI, relationship
 III. Organ
  ABO blood group mismatch, HLA mismatching, GRWR (%)
B. Operative factors
 Cold ischemic time, warm ischemic time, duration of operation
 Blood loss volume during operation
C. Period of surgery

Qualitative data, such as encephalopathy and ascites, were trichotomized according to other reports into none, moderate, or severe/refractory.10 The maximum body temperature during the 24 hours prior to the transplantation was described as a continuous value.

The history of hypertension or diabetes mellitus was defined as positive when the patient had a history of treatment with antihypertensive drugs, sulfonylurea, or insulin. Spontaneous bacterial peritonitis was diagnosed when the ascitic fluid polymorphonuclear cell count was greater than 250 cells/mm3 in the absence of secondary peritonitis11 in our institution or a referral center. The history of upper abdominal surgery excluded laparoscopic surgery undergone before the transplantation.

We assumed that as the total number of such operations increased, implicit factors, such as experience of the surgical procedure and advancement of postoperative therapies, would have improved the technology of ALDLT. The timing of the ALDLT was therefore divided into 2 periods: 1998-2001 and 2002-2004. A total of 143 patients had undergone the surgery by the end of 2001.

Risk Factors

The outcome was defined as graft failure or patient mortality in the 90-day period after ALDLT. Cases in which the patient died or suffered graft failure were defined as censored. A total of 39 potential prognostic variables (Table 1) were examined to identify risk factors using Cox's proportional hazard model.12 Qualitative variables, such as the presence or absence of a sign or symptom, were dichotomous. To lessen the influence of extreme values, we transformed some continuous variables, such as blood loss volume, cold ischemic time, C-reactive protein (CRP), etc., to their natural logarithms.

To identify potential risk factors, we used Cox's proportional hazard regression model throughout the study. First, we screened all the variables by univariate analysis and examined whether or not each individual factor was a risk factor at P < 0.05. Subsequently, we incorporated all significant factors examined interaction among the variables by multivariate analysis and finally isolated independent significant factors. The model estimation and assessment were performed using a computer and the SAS software (version 8.0; SAS Institute, Cary, NC) procedure PHREG.13

Quantitative Survival Modeling

Cox's proportional hazard regression model12 was applied to predict the 90-day survival. In the model, each individual patient is given a risk score as follows: R = (X, β) = X1β1 + X2β2 +… + Xkβk. Let S(t, X) give the probability of patients with a risk factor, R, i.e., (X1, X2, … Xk). Suppose that we know the survival function, S0(t), for individuals with the risk score R = 0, which we express as R0. From the proportional hazards assumption, we can obtain the following very simple formula for S(t, X): S(t, X) = {S0(t)}EXP (R−R0). Standard techniques presented by Kalbfleisch and Prentice14 allow us to estimate S0(t) from the data using the SAS procedure PHREG, and the regression coefficients (β1, β2, …. βk) can be estimated by applying the maximum likelihood estimation method to Cox's partial likelihood. Each coefficient βj has the simple interpretation that every unit increase in the covariate Xj elevates the risk of dying or graft failure by the multiplicative factor exp(βj).

Predictive Survival Models

We used the large cohort of 191 patients to build a predictive model. The R value was computed for each patient. We created 2 models, 1 using all the significant variables (model 1) and the other using only the preoperative variables (model 2). In model 2, a significant operative variable, blood loss volume, was replaced by preoperative variables that showed associations with it. The obtained R values were ranked and divided into 3 groups such that there were approximately equal numbers of deaths in each group. A mean survival rate of the patients within a risk group was used as the representative survival rate of this mortality score group. The survival rate for each patient was computed from the model. The actual survival curves observed were drawn by the Kaplan-Meier method.15 The predicted and observed survival curves were then compared graphically. We also plotted a receiver operating characteristic curve to validate the model.16 In the 2 models, the predicted variable was R and the outcome was defined as the 90-day mortality. The c-statistic reported corresponds to the area under the curve with a value of 0.5 corresponding to no apparent accuracy and a value of 1.0 corresponding to perfect accuracy.

RESULTS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES

The demographic features of cohorts A and B are shown in Table 2. There were no significant differences between the 2 cohorts, except for the recipient ages.

Table 2. Demographics of the Cohorts for Model Development (Cohort A) and Validation (Cohort B)
 Cohort ACohort BP*
  • *

    Continuous data were analyzed by Student's t-test.

Total number19195 
Number of male recipients106460.26
Number of male donors107540.90
Recipient age (mean ± SD)47.4 ± 12.044.1 ± 13.20.041
Donor age (mean ± SD)42.2 ± 11.541.1 ± 12.70.48
MELD score (mean ± SD)19.6 ± 7.0318.9 ± 7.850.46
Malignancy7742 
Viral hepatitis4415 
Primary biliary cirrhosis297 
Biliary atresia911 
Primary sclerosing cholangitis98 
Metabolic liver disease62 
Autoimmune hepatitis33 
Others147 
Chi-squared testχ2 = 11.7P = 0.11 

Univariate Analyses

A univariate proportional hazard regression model was used to examine potential risk factors for association with the 90-day mortality. Univariate analysis revealed 8 potential risk factors at a statistical level of P < 0.05 (Table 3). One operative factor, loge[blood loss volume], was significantly associated with 90-day mortality (hazard ratio [HR] = 2.45; 95% confidence interval [CI] = 1.66-3.61). Model for End-Stage Liver Disease (MELD) score ≥25 (HR = 2.64; CI = 1.25-5.58) was significantly associated with short-term survival. The preoperative inflammatory status was also associated with a poor outcome. Loge[CRP] (HR = 1.70; CI = 1.21-2.40) was associated with a decreased survival rate. Regarding the past histories, a positive history of upper abdominal surgery (HR = 2.21; CI = 1.02-4.78) was statistically significant. Among immunological interactions between the recipient and the donor, 2 mismatches of human leukocyte antigen (HLA) DR (HR = 3.02; CI = 1.40-6.53), 2 mismatches of HLA B (HR = 2.19; CI = 1.01-4.74), and total mismatch numbers of HLA A, B, and DR ≥ 4 were found to be risk factors. Operations between 1998 and 2001 had poorer outcomes than those between 2002 and 2004 (HR = 2.25; CI = 1.02-4.97).

Table 3. Risk Factors for Short-Term Mortality after ALDLT by Univariate Analysis (Cohort A)
VariableCasesDeathsHR95% CIP
  • Abbreviations: CRP, C-reactive protein; GRWR, graft/recipient weight ratio; HLA, human leukocyte antigen; HR, hazard ratio; MELD, Model for End-Stage Liver Disease; PT/INR, prothrombin time/international normalized ratio; 95% CI, 95% confidence interval.

  • *

    Upper abdominal surgery excluded laparoscopic surgery undergone before the ALDLT.

Recipient age  1.000.97–1.030.91
Donor age  1.010.98–1.040.57
Donor age     
 ≤303751  
 31–404550.840.24–2.890.78
 41–505691.250.42–3.730.69
 >505391.250.42–3.720.69
Recipient gender (vs. female)106150.900.43–1.900.79
Donor gender (vs. female)107171.230.58–2.630.59
Period of surgery     
 2003–20049691  
 1999–200284172.271.01–5.090.0047
 -19981122.110.46–9.760.34
MELD score (6–40)  1.051.00–1.100.068
 MELD ≤ 188491  
 18 < MELD < 256271.030.38–2.790.96
 25 ≤ MELD < 303392.661.04–6.780.041
 MELD > 301232.580.70–9.560.16
Loge[blood loss volume]  2.451.66–3.61<0.0001
Loge[cold ischemic time]  1.650.97–2.830.066
Loge[warm ischemic time]  0.410.12–1.380.15
Loge[serum albumin]  1.840.24–14.20.56
Loge[PT/INR]  1.780.67–4.740.25
Loge[serum creatinine]  1.330.62–2.840.47
Loge[serum bilirubin]  1.330.95–1.850.094
Loge[serum CRP]  1.701.21–2.400.0025
Encephalopathy     
 None129161  
 Moderate5391.390.61–3.140.43
 Severe733.851.12–13.20.032
Ascites     
 None7771  
 Moderate75131.990.79–4.990.14
 Severe3882.490.90–6.880.078
History of spontaneous bacterial peritonitis     
 Negative166221  
 Positive2462.060.84–5.080.12
History of upper abdominal surgery*     
 Negative151181  
 Positive40102.211.02–4.780.045
HLA A     
 0 or 1 mismatch162251  
 2 mismatches1910.320.044–2.370.27
HLA B     
 0 or 1 mismatch127141  
 2 mismatches54122.191.01–4.740.046
HLA DR     
 0 or 1 mismatch137141  
 2 mismatches44123.021.40–6.530.0050
HLA     
 0–3 mismatches127121  
 4- mismatches53132.921.33–6.390.0075
Relative donor124141  
Non-blood relative donor66142.020.96–4.240.063
ABO-compatible158201  
ABO-incompatible3382.100.93–4.770.076
GRWR* ≥ 1.0120201  
GRWR < 1.05180.950.42–2.150.90
Graft with middle hepatic vein2731  
Graft without middle hepatic vein158251.460.44–4.840.53

Multivariate Analysis

The 8 potential risk factors derived from the univariate analyses were further assessed by multivariate analysis. The multivariate analysis revealed 4 variables (P < 0.05) with independent prognostic significance (Table 4). These were 2 mismatches of HLA DR (HR = 4.45; CI = 1.96-10.1), loge[blood loss volume] (HR = 2.43; CI = 1.64-3.60), period of surgery (1998-2001 vs. 2002-2004) (HR = 2.41; CI = 1.04-5.57), and loge[serum CRP] (HR = 1.64, CI = 1.13-2.38). These variables were incorporated into model 1.

Table 4. Four Significant Independent Variables
VariableHR (95% CI)P
  • Abbreviations: CRP, C-reactive protein; HLA, human leukocyte antigen; HR, hazard ratio; 95% CI, 95% confidence interval.

  • *

    Period of surgery: 1998–2002 vs. 2003–2004.

Loge[blood loss volume]2.43 (1.64–3.60)<0.0001
2 mismatches of HLA DR4.45 (1.96–10.1)0.0003
Loge[serum CRP]1.64 (1.13–2.38)0.0087
Period of surgery*2.41 (1.04–5.57)0.040

Mortality risk scores were calculated for individual patients among the 95 patients and used to stratify the patients into 3 risk groups. The risk score cutoff values for dividing the patients were chosen so that the groups contained approximately equal numbers of deceased patients. These groups were designated low (R ≤ 8.61), intermediate (8.61 < R ≤ 9.99), and high (R > 9.99) risk groups. The actual patient survival estimates based on the mortality scores and observations by Kaplan-Meier survival analysis for the different groups are shown in Figure 1. The observed survival rates were essentially the same as the predicted values. The model can accurately predict the survival of patients with high risk scores. To further validate this model, we performed receiver operating characteristic curve analysis for 90-day survival for cohort A, and found a c-statistic of 0.63.

thumbnail image

Figure 1. Survival model (model 1) using all variables. Predicted (cohort A: n = 191) and actual (cohort B: n = 95) survival curves for the 3 risk groups. The actual survival curves are essentially the same as the predicted curves for each group.

Download figure to PowerPoint

Predictive Survival Model Using Only the Preoperative Variables

The final goal of the present study was to obtain a predictive quantitative survival model using only the preoperative variables. Since blood loss is an operative value, we tried to replace this variable with preoperative variables. Blood loss volume showed a strong association with the Model for End-Stage Liver Disease (MELD) score (Table 5). Although past history of upper abdominal surgery showed no significant association, its linkage to blood loss volume based on experience and its identification as a significant risk factor by univariate analysis urged us to incorporate this variable. Therefore, the analysis was carried out by incorporating past history of upper abdominal surgery and Model for End-Stage Liver Disease (MELD) score ≥ 25 into model 2 in place of blood loss volume (Table 6).

Table 5. Associations of Blood Loss Volume with Preoperative Potential Risk Factors
 Blood loss volume <4000 mLBlood loss volume >4000 mLTotalP*
  • Abbreviations: MELD, Model for End-Stage Liver Disease.

  • *

    Proportions were analyzed by the chi-squared test.

  • Dichotomized into almost equal numbers.

  • Upper abdominal surgery excluded laparoscopic surgery undergone before the ALDLT.

History of upper abdominal surgery    
 Negative78731510.19
 Positive162440 
MELD < 2581651460.0018
MELD ≥ 25133245 
Table 6. Replacement of Blood Loss with Past History of Upper Abdominal Surgery and MELD ≥ 25
VariableHR (95% CI)P
  • Abbreviations: CRP, C-reactive protein; HLA, human leukocyte antigen; HR, hazard ratio; MELD, Model for End-Stage Liver Disease; 95% CI, 95% confidence interval.

  • *

    Upper abdominal surgery excluded laparoscopic surgery undergone before the ALDLT.

  • Period of surgery: 1998–2001 vs. 2002–2004.

2 mismatches of HLA DR5.07 (2.22–11.6)0.0001
Loge[serum CRP]1.56 (1.07–2.27)0.020
Upper abdominal surgery*2.33 (1.02–5.32)0.045
Period of surgery2.07 (0.87–4.91)0.098
MELD ≥ 252.12 (0.84–5.35)0.11

The risk score cutoff values for dividing the patients were chosen in a similar manner to model 1. Patients were grouped into low (R ≤ 1.34), intermediate (1.34 < R ≤ 2.40), and high (R>2.40) risk groups. The actual patient survival estimates are shown in Figure 2. Although the observed survival rates were less similar to the predicted values than those of model 1, the low risk group, which contained the majority (65.3%) of the patients, showed approximately the same survival rate. To further validate this model, we performed receiver operating characteristic curve analysis for 90-day survival for cohort B, and found a c-statistic of 0.62.

thumbnail image

Figure 2. Predictive survival model (model 2) using only the preoperative variables. Predicted (cohort A: n = 191) and actual (cohort B: n = 95) survival curves. The actual survival curve of the patients with low risk scores fits the predicted curve.

Download figure to PowerPoint

DISCUSSION

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES

The present study was conducted to identify risk factors for 90-day survival in a cohort of consecutive ALDLT patients who received a right lobe graft at a single institution.

There is a previous report regarding risk factors (e.g., donor age and intensive care unit status) for allograft loss after living donor liver transplantation from the United Network for Organ Sharing data set.17 We believe that the present study represents the first exhaustive search for risk factors for 90-day survival after ALDLT at a single center. A systemic search focusing on 39 variables revealed various new risk factors: HLA DR mismatching, blood loss volume, and serum CRP. We constructed a survival model using these variables that can accurately predict the 90-day survival after ALDLT.

Among these risk factors, HLA DR mismatches of 2 loci unexpectedly elevated the mortality rates by about 5-fold compared to mismatches at 0 or 1 locus. It is widely accepted that HLA mismatching has little effect on the outcome after liver transplantation.18 However, the present result may indicate that a recipient receiving a graft from a blood relative donor has some immunological merits compared to a graft from a non-blood relative donor as well as deceased donor grafts. HLA DR mismatching was reported to be associated with acute cellular rejection after living donor liver transplantation.19 Patients with cholestatic disease comprised the majority (44%) of the above cohort. In our study, cholestatic disease was more common (n = 6; 37.5%) than other diseases among the 16 patients who had 2 mismatches of HLA DR among the deceased cases. Although these results suggest an interaction between HLA DR mismatches and liver disease for short-term survival after ALDLT, this observation remains premature due to the small number of cases in the cohort. Further investigations are required to confirm the present finding.

Elevation of serum CRP, which is known to be associated with systemic inflammation, can represent occult bacterial infection, damage to certain organs, or immunological system activity before transplantation. Serum CRP was found to be an independent risk factor after ALDLT. An occult bacterial infection before transplantation may emerge directly as an apparent infection under immunosuppression after ALDLT. On the other hand, occult systemic inflammation, as manifested by an increased CRP level, was reported to be associated with thickening and stenosis after arterial allograft transplantation20 and decreased graft survival after kidney or heart transplantation.21, 22 The mechanism by which serum CRP decreases the survival of patients after ALDLT remains unclear.

Blood loss volume, which has been reported to be an independent risk factor among preoperative recipient factors for deceased donor liver transplantation,23 was also confirmed as a risk factor after ALDLT in the present study. Since this represents an operative variable, we tried to replace blood loss volume with 2 associated preoperative variables, namely past history of upper abdominal surgery and Model for End-Stage Liver Disease (MELD) score ≥25. This model could predict the 90-day mortality for patients with a high c-statistic. However, this model could not accurately predict the success of ALDLT for patients with higher risk scores compared to model 1. Despite such a limitation, this model could be a useful tool for explaining the risks to recipients and their families prior to surgery.

Small-for-size grafts and ABO blood group–incompatible grafts, which were previously reported to be risk factors, are no longer associated with a poor outcome in our institution due to new strategies for addressing these problems (e.g., grafts with middle hepatic vein24 and arterial infusion therapy25) improved the outcome.

The spectrum of risk factors for short-term mortality after ALDLT clearly differs from that for mortality after deceased donor liver transplantation,26–29 indicating the need for a specific clinical management protocol for ALDLT to be established.

Various prognostic models have been reported for deceased donor liver transplantation using preoperative variables, with the aim of maximizing the use of limited organ resources for waiting patients with severe illnesses. In contrast, our study was aimed at improving the short-term outcome after ALDLT. Therefore, we investigated both operative and preoperative factors together with improvements in ALDLT technology. Among the operative factors, blood loss volume was found to be significant, and this may associated with the technical difficulties encountered. The clear message from the present study is that operative factors should also be taken into account for improving the short-term mortality.

This study has several limitations. First, since we collected a dataset at a single institution, the ability to generalize its conclusions is questionable. However, in this cohort, we could exclude heterogeneity in the surgical procedure and postoperative therapeutic regimens, including immunosuppression. Thus, we consider that the latter merit outweighs the former demerit of the study. However, we understand that this model may not be readily applicable to other centers with different patient populations and surgical procedures. Second, in the present model, we only predicted the 90-day mortality. Since the majority (>53.8%) of mortality and graft failures occur within 90 days, improvement of the short-term survival is essential to achieve better long-term survival. Nevertheless, the importance of critical evaluation of long-term survival after ALDLT can not be ignored, and we are currently analyzing long-term survival using the same cohort. Third, due to the small size of study population, meaningful statistical analyses were limited. As such, the present models underestimated the survivals: the discrepancy was more obvious in higher risk groups (Figs. 1 and 2). To obtain better predictions, we should primarily increase the statistical power to enable us to overlook important parameters in univariate analysis. In addition, increasing the size of population will enable us to investigate interactions among parameters. Those fine adaptations to obtain better model will be expected in future.

In summary, we found several risk factors unique to ALDLT and successfully developed a quantitative survival model for 90-day survival after ALDLT. Improving these risk factors may bring about a better short-term prognosis and promise a better long-term outcome.

REFERENCES

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES