There is no financial support to declare, and the authors have no competing interests to declare.
Address reprint requests to Moritz Kleine, M.D., Department of General, Visceral, and Transplant Surgery, Hannover Medical School, Carl-Neuberg Strasse 1, 30625 Hannover, Germany. Telephone: +49-511-532-6534; FAX: +49-511-532-4010; E-mail: firstname.lastname@example.org
area under the receiver operating characteristic curve
fresh frozen plasma
fraction of inspired oxygen
intensive care unit
Model for End-Stage Liver Disease
partial pressure of arterial oxygen
receiver operating characteristic
respiratory risk score
Prolonged mechanical ventilation is known to have a significant negative impact on the survival of critically ill patients.[1, 2] After liver transplantation (LT), the survival of patients requiring prolonged mechanical ventilation has been shown to be significantly inferior (5-year survival: 65.1% for patients with prolonged mechanical ventilation versus 84.4% for patient who were immediately extubated). A primary diagnosis of acute liver failure, retransplantation, the intraoperative transfusion of more than 15 fresh frozen plasma (FFP) concentrates, mechanical ventilation before LT, a partial pressure of arterial oxygen (PaO2)/fraction of inspired oxygen (FiO2) ratio less than 300 mm Hg, and the trend of GOT and GPT in the first 24 hours have been shown to be linked to a prolonged duration of mechanical ventilation after LT.[3, 4]
Typical pulmonary complications following LT include pleural effusion, pulmonary atelectasis, pneumonia, and pulmonary edema.[5-7] In this context, the Model for End-Stage Liver Disease (MELD) score, impaired hepatic function, reperfusion injury, and the intraoperative application of FFP and fluids have been identified as risk factors for the development of lung edema after LT.[7, 8] Pulmonary infections after LT have been shown to appear in 13.7% of organ recipients with a mortality rate of 36.6%. Other studies have reported risk factors for pulmonary infections after LT, including a poor preoperative patient state, the duration of the transplantation procedure, and retransplantation.[4, 5] It can be demonstrated that early tracheostomy within the first 4 days after surgery decreases the ventilation time. Definitions of prolonged ventilation differ widely in the published literature.
Although controversy exists for very early extubation[11, 12] there is a widespread consensus that prolonged ventilation in LT patients should be avoided in order to enable superior outcomes.[1, 2] Prolonged ventilation increases the risk for postoperative pulmonary complications, which tend to prolong ventilation further. The aims of this study were to identify significant risk factors for long-term ventilation and 3-month mortality and to develop and validate a respiratory risk score (RRS) for the prediction of these outcomes. So far, clinicians have no validated predictive tools that may be helpful for decisions on the timing of tracheostomy and the prognosis of 3-month mortality and prolonged mechanical ventilation after LT. A further goal of this study was to find a rationale for the definition of prolonged ventilation.
PATIENTS AND METHODS
The setting was a tertiary referral university hospital within the Eurotransplant area.
All adults (>18 years) who underwent LT between January 2006 and December 2010 with follow-up until April 2011 and were treated in the intensive care unit (ICU) of our department were included. Table 1 shows the indications for LT and the most probable causes of death in the study population.
Table 1. Indications for LT and Most Probable Causes of Death in the Study Population
Indications for LT
Retransplantation: chronic rejection
Primary sclerosing cholangitis
Hepatitis C virus–related cirrhosis
Glycogen storage disease
Hepatitis B virus/hepatitis C virus–related cirrhosis
Acute liver failure
Intrahepatic cholangiocellular carcinoma
Hepatitis B virus–related cirrhosis
Retransplantation: chronic graft failure
Primary biliary cirrhosis
Retransplantation: chronic biliary complication
Secondary biliary cirrhosis
Retransplantation: recurrent viral hepatitis
Familial amyloidotic polyneuropathy
Most Probable Causes of Death
Liver graft: initial nonfunction
Liver graft: biliary tract complication
Lungs: acute respiratory distress syndrome
Data not available
De novo malignancy
Patients with acute retransplantation (defined as retransplantation during the first 4 weeks after initial transplantation) were excluded from the analysis in order to avoid bias caused by the significantly worse outcomes of acute retransplantation versus primary transplantation. Patients who died before postoperative day 8 were excluded from the analysis of prolonged mechanical ventilation (defined in this study as ventilation > 7.5 days). This definition of prolonged ventilation was determined here with the best Youden index, with the duration of mechanical ventilation in days used as a prognostic model to predict 3-month mortality.
As an observational, retrospective study, neither informed consent nor approval by an ethics committee was needed for this study according to the professional code of the German Medical Association (article B.III. §15.1).
Clinical Data Collection
Clinical data were collected retrospectively.
Three-month mortality was defined as the primary study endpoint. The secondary endpoints were prolonged mechanical ventilation (>7.5 days) and the duration of mechanical ventilation in days. The duration of mechanical ventilation included all periods of mechanical ventilation after transplantation, even if patients were extubated in the meantime.
The whole cohort of 254 LT patients with completely available preoperative and postoperative data was retrospectively randomized into a training group (n = 135) and a validation group (n = 119). The training group was used to provide data for a statistical playing field guided by clinical expertise in order to enable creative score development. The validation group was used to validate developed model scores. Three additional random samples were drawn from the complete cohort for the purpose of validating the final model score. Successful validation was defined as an area under the receiver operating characteristic curve (AUROC) > 0.7, a Brier score < 0.25, and a P value > 0.05 according to the Hosmer-Lemeshow test. A Cox regression analysis was used to determine hazard ratios (HRs) for 3-month mortality. These HRs were used as coefficients for the calculation and weighing of variables that were used for the design of scores used as prognostic models to predict 3-month mortality. Further methods included receiver operating characteristic (ROC) curve analysis, c statistics, the Hosmer-Lemeshow test, and Brier scores in order to test and compare scores as prognostic models and determine their model calibration and model accuracy as well as their sensitivity, specificity, and overall model correctness for the prediction of study endpoints.[14-18] The best Youden index was used to determine cutoff values. Kaplan-Meier analysis and log-rank tests were applied when they were appropriate. For all statistical tests, the level of significance was defined as P < 0.05. Variables and follow-up data are presented as medians and ranges. SPSS statistics software (version 20.0, IBM, Somers, NY) was used to perform the statistical analysis.
All inclusion and exclusion criteria were fulfilled by 302 patients. The median length of time on a respirator was 3 days (mean = 12.68 days, range = <1 to 130 days) with a median follow-up period of 1182 days (mean = 1100 days, range = 124-1910 days). The median time that patients spent in the ICU and in the hospital were 9 days (mean = 22.36 days, range = <1 to 354 days) and 30 days (mean = 44.32 days, range = 1-409 days), respectively. The mean age of the whole cohort at the time of transplantation was 51 years (range = 18-69 years). Fifty-nine percent of the patients were male. The overall survival of all 302 patients was 82.1%, with 248 patients still alive at the last follow-up. Complete preoperative data were available for 254 of the 302 patients. Before transplantation, 32 of these 254 patients were treated in the ICU, 16 were on mechanical ventilation, 27 needed dialysis, and 77 were diagnosed with encephalopathy. Thirty of the 254 patients received an organ with significant histologically confirmed steatosis hepatis (>40%). The donor risk index and the extended criteria donor score for transplanted donor livers were available for 199 of the 254 patients. Seventy-two patients underwent transplantation with donor livers with a donor risk index > 1.7, which is identical to the cutoff value that we published before for the prediction of 3-month mortality after LT (Table 2). Seventy-three patients underwent transplantation with donor livers with an extended criteria donor score > 2.5 points (the cutoff level determined with the best Youden index for the prediction of 3-month mortality; Table 2), which is markedly lower than our previously published cutoff for the prediction of 3-month mortality (3.5 points) in a slightly different study population.
Table 2. Variables With Significant Influence on Prolonged Mechanical Ventilation (>7.5 Days on a Respirator) and 3-Month Mortality According to the Cutoff Values
Log Regression: Variable and Prolonged Mechanical Ventilation
Cox Regression: Variable and 3-Month Mortality
NOTE: Log regression and Cox regression analyses are provided with P values, ORs or HRs, and 95% CIs.
Preoperative MELD score
FFP within 24 hours (U)
Packed red blood cells within 24 hours (U)
PaO2/FiO2 ratio on day 1
Preoperative ICU stay
Preoperative mechanical ventilation
Donor steatosis hepatis > 40%
Donor ICU time (days)
Donor risk index
Extended criteria donor score
Influence of the Time on Mechanical Ventilation on Posttransplant Survival
The survival of patients declined significantly with each additional day on mechanical ventilation with a significant HR [P < 0.001, HR = 1.015, 95% confidence interval (CI) = 1.01-1.02]. A Cox regression analysis revealed significantly superior survival during follow-up for patients with <7.5 days on respirator therapy versus >7.5 days (93.2% versus 67.3%, P < 0.001, HR = 5.98, 95% CI = 3.14-11.41), and this was confirmed by a Kaplan-Meier survival analysis with the log-rank test (P < 0.001; Fig. 1A). As described previously, the cutoff value for the duration of mechanical ventilation (>7.5 days) was determined with the best Youden index as the duration in days on respirator therapy with the best sensitivity and specificity for the prediction of 3-month mortality and, therefore, defined prolonged mechanical ventilation for this study.
Calculation of the Cutoff Values for Variables and Their Influence on 3-Month Mortality and/or the Time on Mechanical Ventilation
The cutoff values for continuous variables such as the preoperative MELD score, the number of transfused FFP and packed red blood cell units within the first 24 hours after transplantation, the PaO2/FiO2 ratio on day 1, the donor risk index, and the extended criteria donor score were determined with the best Youden index for the prediction of 3-month mortality and for the prediction of prolonged ventilation (>7.5 days). These cutoff values as well as additional categorical variables and their statistical influence on 3-month mortality and prolonged ventilation > 7.5 days [pretransplant ICU stay (yes/no), preoperative mechanical ventilation (yes/no), preoperative dialysis (yes/no), preoperative encephalopathy (yes/no), histological donor steatosis hepatis > 40% (yes/no), and donor ICU time > 7.5 days (yes/no)] are summarized in Table 2.
Development and Validation of the RRS
Variables with significant influence on outcomes were used to design 4 scores as prognostic models with the use of training group data. ROC curve analysis results, Hosmer-Lemeshow test results, and Brier scores as well as the sensitivities and specificities for the cutoff values of these 4 scores for the prediction of 3-month mortality are summarized in Table 3.
Table 3. Results of the 4 Model Scores for the Prediction of 3-Month Mortality in the Training Group of 135 Patients and Results of the RRS for the Prediction of 3-Month Mortality and Prolonged Mechanical Ventilation (>7.5 Days) in the Complete Cohort of 254 Patients
AUROC (> 0.7)
Brier Score (P < 0.25)
Hosmer-Lemeshow Test (P > 0.05)
NOTE: AUROC results, calculated cutoff values, corresponding Brier scores, Hosmer-Lemeshow test results, and sensitivity and specificity values are presented.
Training Group (n = 135)
Prediction of 3-month mortality with model score 1
Prediction of 3-month mortality with model score 2
Prediction of 3-month mortality with model score 3
Prediction of 3-month mortality with model score 4
Complete Cohort (n = 135)
Prediction of 3-month mortality with RRS
Prediction of prolonged mechanical ventilation (>7.5 days) with RRS
All parameters with significant influence on 3-month mortality or prolonged mechanical ventilation were used for the design of model score 1 (Tables 2 and 4). The HRs of these variables for 3-month mortality were used to weigh parameters in model score 1, which was calculated by the simple addition of the HRs of these variables if they were positive. The Brier score of model score 1 indicated inadequate model accuracy (Brier score = 0.28). Therefore, model score 2 was designed.
Table 4. Four Different Model Scores With Their Variables and Corresponding Points
Model Score 1
Model Score 2
Model Score 3
Model Score 4
Preoperative MELD score > 30
FFP within 24 hours > 13.5 U
Packed red blood cells within 24 hours > 10.5 U
PaO2/FiO2 ratio on day 1 < 200
Preoperative ICU stay: yes/no
Preoperative mechanical ventilation: yes/no
Preoperative dialysis: yes/no
Preoperative encephalopathy: yes/no
Donor steatosis hepatis > 40%: yes/no
Donor ICU time > 7.5 days
Model score 2 contained all 8 variables with significant influence on 3-month mortality (Tables 2 and 4). The HRs of these variables for 3-month mortality were used to weigh parameters in model score 2, which was calculated by the simple addition of the HRs of all 8 variables if they were positive. The Brier score of model score 2 indicated inadequate model accuracy (Brier score = 0.28). Therefore, model score 3 was designed.
Model score 3 was designed as model score 2 was but without the variable preoperative encephalopathy, which displayed only a marginally significant influence on 3-month mortality after LT (Tables 2 and 4). The Brier score of model score 3 indicated good model accuracy (Brier score = 0.23). Model score 3 was able to predict 3-month mortality with marginally acceptable sensitivity (61%) and high specificity (82%; Table 3).
Because the final RRS should be a clinically applicable predictive score for all LT cases, model score 4 was designed without the variable steatosis hepatis > 40%, which was used in model score 3. Steatosis hepatis > 40% in donor livers was determined by the histological examination of donor liver biopsy samples, which might not be available for all LT cases. Unfortunately, the Brier score of model score 4 demonstrated inadequate model accuracy (Brier score = 0.45).
Altogether, model score 3 displayed the best ROC analysis results with the highest sensitivity and specificity for the prediction of 3-month mortality after LT, and it had a Brier score that indicated good model accuracy in the training group. Therefore, the validation of model score 3 in the validation group was tested.
The application of this score in the validation group confirmed an AUROC greater than 0.7 (0.913) for the prediction of 3-month mortality after transplantation. The sensitivity (88%) and specificity (87%) for the prediction of 3-month mortality after LT in the validation group were markedly higher in comparison with the training group results with a Brier score that confirmed good model accuracy (0.151). Table 5 shows the characteristics of all studied sets and demonstrates that the score parameters differed only slightly between the studied groups, with a higher 3-month mortality rate in the training group. In order to rule out the idea that this difference was a major influence on the validation of model score 3 in the first validation, we analyzed 3 additional random samples drawn from the complete cohort for the purpose of soundly validating proposed model score 3. The application of model score 3 to the 3 different random samples confirmed an AUROC value greater than 0.7 (range = 0.761-0.805) for the prediction of 3-month mortality after transplantation. The sensitivity (range = 63.6%-84.6%) and specificity (range = 67.6%-85.3%) for the prediction of 3-month mortality after LT in the 3 random samples were comparable to the training and validation group results, and the Brier scores confirmed good model accuracy in every randomized sample (range = 0.160-0.186). Good model calibration was also demonstrated in these 3 additional random samples with Hosmer-Lemeshow test results, with P values greater than 0.05 (range = 0.44-0.59). Because the best results for model calibration and model accuracy were obtained in the training and validation groups and the 3 random samples with model score 3, this model was applied to the entire cohort of 254 LT recipients with completely available data and was finally termed the RRS. The RRS is calculated as follows: (lMELD score > 30 = 2.36 points) + (FFP>13.5 U = 2.70 points)+(PaO2/FiO2 ratio < 200 mmHg = 2.23 points) + (packed red blood cells >10.5 U = 3.50 points) + (preoperative mechanical ventilation = 3.87 points) + (preoperative dialysis = 2.83 points)+(donor steatosis hepatis >40%=2.95 points).
Table 5. Continuous and Categorical RRS Variables and 3-Month Mortality Rates for the Retrospectively Randomized Training and Validation Groups and 3 Random Samples
Application of the RRS to the Prediction of 3-Month Mortality in the Complete Cohort
The RRS (formerly termed model score 3) demonstrated in the complete cohort a sensitivity of 69%, a specificity of 83% and an overall model correctness of 76% (AUROC = 0.794) for the prediction of 3-month mortality after LT. The P value of the Hosmer-Lemeshow goodness-of-fit test was 0.16, and the Brier score was 0.18. These results confirm the good model discrimination, model calibration, and model accuracy of the RRS for the prediction of 3-month mortality after LT (Table 3).
Survival after LT was significantly worse for LT patients with an RRS above the cutoff level of 6.64 points (60.7%) versus patients with a lower RRS (88.8%, log-rank P < 0.001; Fig. 1B). One hundred thirty-two patients with an RRS of 0 to 3.69 points displayed a very low 3-month mortality rate (1.5%), whereas patients with an RRS > 6.64 points (n = 50) had a high 3-month mortality rate of 24% (Fig. 2). An RRS > 6.64 points was associated with a significant HR for 3-month mortality (Cox regression: P < 0.001, HR = 10.61, 95% CI = 4.43-25.41).
Application of the RRS to the Prediction of Prolonged Ventilation in the Complete Cohort
As shown previously, the duration of mechanical ventilation had a significant influence on 3-month mortality with a cutoff value of 7.5 days (the best Youden index). The RRS displayed an AUROC > 0.7 (0.798) for the prediction of prolonged ventilation > 7.5 days. The cutoff value for this prediction was 3.69 RRS points with a sensitivity of 81%, a specificity of 73% and an overall model correctness of 77%. The Hosmer-Lemeshow test confirmed good model discrimination and model calibration with a P value > 0.05 (P = 0.17). The Brier score confirmed the good model accuracy for this prediction (Brier score = 0.24; Table 3).
Forty-six percent of the LT recipients (n = 114) had an RRS greater than 3.69 points with a significant odds ratio (OR) for prolonged ventilation (log regression analysis: P < 0.001, OR = 11.53, 95% CI = 6.05-21.97). Sixty-one percent of these patients (n = 70) experienced prolonged ventilation (>7.5 days), whereas 12% (n = 16) did in the group of LT recipients with an RRS below 3.69 points (n = 132).
LT recipients were subgrouped into 3 risk categories by the RRS cutoff values for the prediction of prolonged ventilation (3.69) and 3-month mortality (6.64). The probability of long-term ventilation longer than 7.5 days was 12% for patients in the low-risk group and rose to 74% for individuals in the high-risk group (Fig. 2). Patients with an RRS between 3.70 and 6.64 points were at an intermediate risk for prolonged ventilation (52% of the patients were at risk). These findings are underlined by the median time that patients spent on respirator therapy in the low-risk group (1 day), the intermediate-risk group (8 days), and the high-risk group (19 days; Fig. 2).
This study provides the first validated prognostic model for the prediction of 3-month mortality and prolonged ventilation after LT with high sensitivity and specificity, good model discrimination, good model accuracy, and good model calibration. The proposed RRS might provide a useful clinical tool for intensive care resource planning and prognostic decisions in the early phase after LT.
The identification of patients at high risk for prolonged mechanical ventilation might also help to establish indications for nonstandard therapies aimed at improvements in lung function after LT (eg, kinetic bed therapy and noninvasive positive pressure ventilation). Clinical trials in this field, especially after major surgery, are frequently underpowered because of the heterogeneity of the investigated populations and the widely differing definitions of prolonged ventilation. The categorization of LT patients with the RRS could help to improve study designs.
Many patients after LT are candidates for early weaning and extubation (n = 117 or 39% for a respirator time < 24 hours in the present study). Reduced costs have been demonstrated for early extubation after LT in single-center studies.[22, 23] Nevertheless, Glanemann et al. could not find a significant survival difference between patients extubated immediately after transplantation and patients who were extubated within the first 24 hours after surgery.
Certainly, the extubation of LT patients should not be delayed because of the risk of ventilator-associated pneumonia, which significantly increases with each day of mechanical ventilation. This view is underlined by the data of the present study, which demonstrate that the survival of patients declined significantly with each additional day on mechanical ventilation with a significant HR (Cox regression analysis: P < 0.001, HR = 1.015, 95% CI = 1.01-1.02). At our center, the extubation of LT patients is not performed until the patients have stable vital signs and conditions that render the need for surgical interventions unlikely. Furthermore, standard extubation criteria are expected to be fulfilled; these include adequate patient cooperation with the ability to follow simple commands, a body temperature greater than 36°C, and an adequate spontaneous tidal volume. This approach is based on a retrospective study demonstrating that the main indications for reintubation after LT are pulmonary and surgical complications, which in turn have been associated with significantly worse outcomes. The results of the present study highlight the advantages of early extubation after LT and show variables with significant influence on 3-month mortality and prolonged ventilation (>7.5 days) after LT.
It should be taken into account that the time on a respirator in the present study also included the time on a respirator after reintubation or tracheostomy in cases of failed weaning. Therefore, the median overall time on mechanical ventilation in the whole cohort of this present study (3 days) was longer than the time in other studies in which the time on respirator treatment was calculated only until initial tracheal extubation with no regard to possibly necessary reintubation and further mechanical ventilation (eg, due to failed weaning).
We believe that the RRS is not designed for the identification of LT candidates who may be suitable for a fast-track extubation program. The RRS is a score for the identification of patients with a high risk of 3-month mortality and a high risk of prolonged mechanical ventilation (>7.5 days). The RRS might be useful for decisions on early preventive measures against common complications associated with long-term mechanical ventilation (eg, the indication for kinetic bed therapy). We believe that the RRS is probably more useful for identifying patients in need of early tracheostomy, kinetic bed therapy, or noninvasive positive pressure ventilation than for identifying patients who may be suitable candidates for early extubation. We believe that the current study warrants validation of the usefulness of the RRS in a transplant center with a fast-track extubation program.
In the present study, patients who were extubated within the first 24 hours after LT stayed for a comparatively long time in our ICU (median = 6 days) with the aim of preventing readmission to the ICU because this would likely have resulted in significantly lower patient and graft survival, as has been reported before. Published data on the length of ICU stays after LT vary greatly, and this is largely dependent on logistical prerequisites and local conditions. Huang et al. reported a comparatively long median stay in the ICU (9 days) for patients without postoperative respiratory failure, whereas Levy et al. reported a comparatively short median ICU stay (3 days) for uncomplicated LT recipients.
The MELD score is used to predict mortality on the waiting list and predicts the pretransplant need for mechanical ventilation. Creatinine clearance before LT has a significant impact on graft and recipient survival, and impaired renal function has been identified as an independent risk factor for respiratory failure after LT. Our study clearly shows that a pretransplant laboratory MELD score > 30 points is an independent risk factor for long-term mechanical ventilation (Table 2). The calculated laboratory MELD scores have increased significantly (by approximately 30%) since the introduction of MELD-based allocation at our center (median pre–MELD era score = 14.35, median MELD era score = 18), and this has resulted in a worsened clinical pretransplant state of liver recipients. In the same time period, the calculated RRS has increased from a median of 2.14 points in the pre-MELD era to a median of 6.84 points in the MELD era. This increase in the median RRS correlates with an increase in the median time on mechanical ventilation from 2 days in the pre-MELD era to 3 days in the MELD era in our cohort. In our population, the median ICU stay has likewise increased correspondingly from 8 to 10 days.
Pretransplant patient morbidity varies significantly between different transplant centers, as observed in a recently published multicenter study on early extubation with a calculated median MELD score of 16 points (range = 11-22, mean = 16.9). Therefore, the validation of our proposed RRS in transplant centers with lower median laboratory MELD scores would be desirable. Ideally, the validation of the RRS should be carried out prospectively in a multicenter setting.
Taner et al. published a series of 1045 LT recipients, 523 of whom were fast-tracked to the surgical ward after transplantation. For the group of LT patients with less than 24 hours of mechanical ventilation in our cohort, this procedure might be adequate. Unfortunately, logistical prerequisites at our institution are not appropriate for such a fast-track approach. For example, we do not have a 24-hour postanesthesia care unit with a 1:1 nurse-to-patient ratio as proposed by Taner et al. A multivariate logistic regression analysis in the study by Taner et al. revealed that the laboratory MELD score was a significant predictor for the length of the ICU stay. The mean laboratory MELD score was only 15.50 points for fast-tracked patients but 18.19 points for patients who were treated in the ICU instead of the postanesthesia care unit, and both values were markedly lower than the mean laboratory MELD score of 19.78 points for the cohort of our study. The laboratory MELD score in the present cohort showed a significant influence on survival and the time on mechanical ventilation. We conclude, therefore, that the LT recipient populations of our study and the recent study by Taner et al. differed significantly, and this may provide a further explanation for the longer respirator therapy times and ICU stays in the present study.
Methodological limitations of the present study were likely caused by center bias, the lack of a uniform decision process regarding extubation criteria, and significant transplant center–specific differences in patient cohorts as well as logistical prerequisites and local intensive care resources. It can be expected that these factors will require a reassessment of the HRs of the investigated variables under different local conditions.
It cannot be ruled out that potentially significant variables that were not available to us in this retrospective study might have enabled the design of a superior predictive model. Despite the strengths of our score, it might be a matter of debate whether the RRS has been built in an optimal way. However, an alternative model score design using a multivariate logistic regression methodology created a model score with suboptimal model calibration (data not shown).
The severity of postoperative lung dysfunction, as quantified by the PaO2/FiO2 ratio on the first day after transplantation, was a risk factor for long-term mechanical ventilation and short-term mortality (Table 2). A PaO2/FiO2 ratio less than 200 mm Hg indicates severe deterioration of lung function early after transplantation and is frequently due to pulmonary edema. A similar finding was published before.
It has been shown before that the number of transfused units of FFP influences the risks of severe pulmonary edema, readmission to the ICU, respiratory complications and long-term mechanical ventilation, and inferior 1- and 5-year survival significantly.[3, 8, 27, 35] The data of the present study confirm these findings to a large degree. The management of FFP transfusions differs widely between transplant centers, partly because of the significant impact of a patient's status as measured by the MELD score and because of multiple other preexisting factors that influence coagulopathy (eg, donor organ function). So far, there are no evidence-based transfusion triggers available for the transfusion of FFP or packed red blood cells. The median number of transfused FFP units in the first 24 hours was higher in our collective experience in comparison with previously published data (12 versus 9). We believe that this observation may be due to a higher median laboratory MELD score in our cohort (17 versus 15.9). As shown by Xia et al., the preoperative laboratory MELD score has an influence on transfusion rates. The link between these 2 variables and other possible influences of the variables used in the RRS might be a methodological problem for the generalizability of the RRS to other LT populations.
In conclusion, our data demonstrate the ability of the newly introduced RRS to discriminate short-term survivors and nonsurvivors as well as prolonged ventilation (>7.5 days). Future studies with an independent external cohort are still needed for the purpose of soundly validating the proposed RRS and validating the used cutoff values. After external verification, the RRS might be useful to clinicians for optimizing ICU resource utilization and improving the design of clinical studies in the field of LT in the near future.