548093 research-article2014
SJS0010.1177/1457496914548093Predictors in liver surgeryC. Riediger, et al.
Original Article
Scandinavian Journal of Surgery 0: 1–9, 2014
Preoperative Serum Bilirubin And Lactate Levels Predict Postoperative Morbidity And Mortality In Liver Surgery: A Single-Center Evaluation C. Riediger1,2, M. W. Mueller1,3, A. Hapfelmeier4, J. Bachmann1, H. Friess1, J. Kleeff1 1
Department of Surgery, Technische Universität München, Munich, Germany Department of General, Thoracic and Vascular Surgery, University Hospital Carl Gustav Carus, Technische Universität Dresden, Dresden, Germany 3 Department of Surgery, Krankenhaus Bad Cannstatt, Klinikum Stuttgart, Stuttgart, Germany 4 Institute of Medical Statistics and Epidemiology, Technische Universität München, Munich, Germany 2
Abstract
Background and Aims: In spite of huge developments in liver surgery during the last decades, morbidity and mortality continue to pose problems in this field. The aim of this study was to identify preoperative predictors for postoperative mortality and morbidity in liver surgery. Material and Methods: In a single-center study, an extensive analysis of a prospective database, including clinical criteria and laboratory tests of patients undergoing liver surgery between July 2007 and July 2012 was performed. Cutoff values of selected laboratory tests were calculated. Results: In all, 337 patients were included in the study. Univariate analysis showed a statistically significant association of preoperative bilirubin, lactate, hemoglobin levels, platelet count, and prothrombin time with postoperative morbidity and mortality. Multivariate analysis revealed preoperatively elevated serum bilirubin and lactate levels as independent predictors for increased postoperative morbidity and mortality after liver surgery. Conclusions: The identified laboratory values showed a statistically significant association with postoperative morbidity and mortality in liver surgery and might be helpful in preoperative patient selection. Key words: Liver surgery; preoperative predictors; postoperative morbidity; postoperative mortality; bilirubin; lactate
C.R. and M.W.M contributed equally to the writing of this article. Correspondence: Carina Riediger, M.D. Department of General, Thoracic and Vascular Surgery University Hospital Carl Gustav Carus Technische Universität Dresden Fetscherstrasse 74 01307 Dresden Germany Email: [email protected]
2
C. Riediger, et al.
Introduction Improvements in liver surgery have resulted in remarkable reductions in mortality and morbidity over the last few decades. Due to the rapid development of liver dissection techniques, as well as improvements in anesthetics, the reported mortality rate of 50% for major liver resection in the 1970s has decreased dramatically (1). Even though the mortality rate is nowadays less than 5%, morbidity rates of up to 40% are still reported (2). Primarily, bleeding and biliary complications, as well as infections, are clinically the most serious problems (3). The high morbidity rates are explained by the rising numbers of liver resections due to the higher incidence of liver tumors (especially metastasis) and the increased percentage of extended resections (4). Moreover, patients undergoing liver surgery are increasingly older and present with more co-morbidity (5). Different studies in the literature discuss possible risk factors with regard to the outcome after liver surgery. Some authors analyzed factors influencing short-term outcome, while others focused on long-term survival, especially in cancer patients (6, 7). It has been shown that postoperative morbidity has an additional impact on the long-term and disease-free survival in cancer patients, and that the reduction of perioperative morbidity can also improve the oncological outcome of curative resection (7, 8). According to the literature, some risk factors for postoperative morbidity and mortality are defined, for example, higher ASA (American Society of Anesthesiologists) classification, steatohepatitis, liver cirrhosis, extent of resection, operating time, need for transfusions, and selected laboratory tests (9–13). Many groups have analyzed defined subgroups of patients with specific malignancies (hepatocellular carcinoma, colorectal liver metastases), or underlying liver disease like liver cirrhosis (14). The majority of studies focus on intra- and postoperative risk factors of liver surgery. Only a few groups have analyzed preoperative parameters to predict postoperative mortality and morbidity (11). The aim of this single-center study was to define predictors that help to identify preoperatively patients at risk for increased postoperative morbidity and mortality in liver surgery. Methods Patients
In this single-center study, patients’ data from July 2007 to July 2012 were prospectively recorded in an electronic database. Patients who received liver resections simultaneously with another operation such as colorectal resection, gastrectomy, esophagectomy, or pancreatectomy were excluded from the analysis, as were atypical liver resections during explorative laparotomy/laparoscopy and “secondary” liver resections. In all, 337 patients met inclusion criteria and were finally included in the database analysis. Patient characteristics, pre- and postoperative treatment, laboratory values, operative procedures, intraand perioperative data, complications, hospital stay,
and others were recorded. Analysis of surgical procedures included anatomical segmentectomies, nonanatomical segmentectomies, right and extended right hemihepatectomies as well as left and extended left hemihepatectomies. It is of note that only non-laparoscopic liver resections were included in this analysis. The study was performed along the principles of the Declaration of Helsinki and approved by the local ethical committee. A written informed consent of patients was obtained. Statistical Analysis
Statistical analysis was performed using SPSS software, version 20 (SPSS Inc., Chicago, IL, USA) and R for Statistical Computing (R Foundation for Statistical Computing, Vienna, Austria). Descriptive statistics were given by median and range or interquartile range (IQR) for continuous data. The distribution of categorical data was presented in contingency tables showing absolute and relative frequencies. All statistical tests were conducted in an explorative manner on a twosided 5% (α = 0.05) significance level. Univariate analysis Univariate analysis was performed to identify risk factors associated with rates of mortality, re-operation, morbidity, and surgical or medical complications. Mortality was defined as in-hospital mortality or 30-day mortality. Morbidity included surgical and medical complications. Surgical complications were defined as abdominal wall dehiscence, wound infection, intra-abdominal abscess, liver abscess, liver necrosis, liver hematoma, bilioma, icterus, biliary fistula, insufficiency of bile duct anastomosis, ascites, pneumothorax, ischemia, intra-abdominal bleeding, perforation of the gut, and mechanical ileus. Medical complications were defined as aspiration, upper or lower gastro-intestinal (GI)-bleeding, perforation of stomach ulceration, venous thrombosis, lung embolism, pleural effusion, pneumonia, other pulmonary complications, cardiac ischemia, cardiac arrhythmia, other cardiac complications, urinary infection, temporary dialysis, permanent dialysis, cholangitis, sepsis, liver failure, multi-organ failure, seizures, and damage to the central nervous system. Gender and age were analyzed, as well as patients’ ASA score. Moreover, preoperative clinical symptoms (pain, jaundice, pruritus, bleeding complications, loss of weight, and reduced general condition); medical history, including co-morbidities (such as cardiac, pulmonary or renal diseases, diabetes mellitus, viral hepatitis, autoimmune hepatitis or liver cirrhosis with portal hypertension or cholelithiasis); and long-term medication (diuretics, aspirin, coumarines, vitamin K, oral contraceptives, or immunosuppressants) as well as specific oncological pre-treatments (chemotherapy, radiation, transarterial chemoembolization (TACE), radiofrequency-assisted ablation (RFA), and others) were analyzed in univariate analyses. Univariate analysis of the patient’s blood tests on day 1/0 preoperative to surgery, as well as analysis of intraoperative data (operation time, need of blood transfusion, and
Predictors in liver surgery
extent of liver resections) and postoperative data (need of transfusions, intensive care unit (ICU) stay, and days of mechanical ventilation) was performed. To investigate the relationship between these potential risk factors and outcome, Mann–Whitney U tests were performed for continuous data and χ2 tests or Fisher’s exact test, depending on the cell counts of the respective contingency tables, for categorical data. Multivariate analysis Parameters that showed statistically significant differences in regard to mortality, overall morbidity, surgical complications, medical complications, and re-operations in the univariate analyses were included in a multivariate analysis. The multivariate analysis was performed by logistic regression models designed with stepwise forward variable selection based on Akaike’s information criterion (AIC). The odds ratio (OR) and corresponding 95% confidence intervals (CIs) are presented for each predictor to quantify its relation to the investigated risks. Definition of cutoff values Cutoff values of selected preoperative laboratory tests (hemoglobin, hematocrit, lactate, bilirubin, platelet count) were calculated to define risk groups with maximally divergent risks for surgical or medical complications, morbidity, mortality, or re-operation. Statistical significance of such maximally selected test statistics was assessed in a permutation test framework (15). A global cutoff value for each of the following laboratory tests was performed: serum hemoglobin, hematocrit, lactate, bilirubin, and platelet count. The global cutoff value was determined for a combined outcome that accounts for any event resulting from mortality, morbidity, surgical complications, medical complications, and re-operation. Decision trees for preoperative blood tests Decision trees perform successive binary splits in the selected preoperative laboratory results of lactate and bilirubin to produce risk groups for the outcomes mortality and morbidity (including surgical and medical complications) (15). They can be seen as a multivariate extension of the univariate cutoff determination described in the preceding steps. After the first split is found to separate the patient cohort into a high-risk and a low-risk group, it might be reasonable to divide these subsets even further. If this is the case, additional cutoff values are searched for within each subgroup. The procedure stops when there are no meaningful separations left. Results Patients’ Characteristics
Between July 2007 and July 2012, a total of 337 liver resections were finally included in our analysis. Detailed information about the analyzed cohort is
3
shown in Table 1. The overall mortality of the cohort was 3.7%, whereas the overall morbidity was 32.7% with an overall re-operation rate of 8.3%. Univariate Analysis
Age and gender The median age was 64.5 years (range: 20–87 years). The cohort consists of 189 male (56%) and 148 female (44%) patients. There were no statistical significant differences in gender and age with regard to mortality, re-operation, and morbidity, surgical or medical complications. Nevertheless, some tendencies were observed: patients with medical complications showed a higher median age of 67.1 years compared to a median age of 63.8 years for patients without medical complications (p = 0.179). Similar results were seen with regard to in-hospital mortality: patients who died within 30 days after surgery showed a median age of 70.7 years (range: 31.5–85.9 years), whereas patients without mortality had a median age of 64.3 years (range: 19.9– 86.7 years) (p = 0.198). The group of patients who died consisted of 69% male and 31% female patients. Of note, the mortality rate within the female cohort was 2.7% and within the male cohort 4.8%. Medical history, long-term medication, pre-treatment, and ASA score Patients presenting with increased postoperative morbidity had statistically significant more preoperative weight loss (p = 0.049) and blood loss (p = 0.046) had an increased rate of diabetes mellitus (p = 0.034), cardiac disease (p = 0.016), and long-term diuretic medication (p = 0.006), or were pre-treated with TACE (p = 0.019). Cardiac disease (p = 0.001) and long-term medication with aspirin (p = 0.010) and diuretics (p = 0.035) were associated with higher rates of surgical complications, whereas cardiac disease, diabetes mellitus (p = 0.036), liver cirrhosis (p = 0.033), and long-term medication with diuretics (p = 0.001) and preoperative chemotherapy (p = 0.024) were associated with medical complications. The group of patients in need of early re-operation showed a significant higher number of patients with long-term diuretic medication (p = 0.029). Neither medical history and long-term medication nor pre-treatment showed a statistically significant relation to in-hospital mortality. However, higher ASA score was statistically significant associated with higher postoperative morbidity (p = 0.002) and mortality (p = 0.001) and need for operative revision (p = 0.008). Preoperative laboratory results The univariate analysis revealed that preoperative elevated serum lactate and bilirubin levels, as well as preoperative reduced prothrombin time, hemoglobin levels, and platelet count are associated with significantly higher rates of postoperative morbidity (including surgical and medical complications), re-operation, and in-hospital mortality. Detailed results are shown in Table 2.
4
C. Riediger, et al. Table 1 Patients’ characteristics.
Number of liver resections Number of patients Age (years) Gender Liver cirrhosis Hepatitis B C Pretreatment Chemotherapy TACE RFA Resection technique Extent of liver resections Right hemihepatectomy Left hemihepatectomy Extended right hemihepatectomy Extended left hemihepatectomy Cyst deroofing Segmentectomy atypical Segmentectomy anatomical MELD score ASA score Diagnosis
N = 337 N = 337 Median: 64.5 Range: 20–87 Male: 189 (56%) Female: 148 (44%) 48 (15%) 12 (4%) 16 (5%) 104 (32%) 79 (25%) 18 (6%) 19 (6%) CUSA: 6 (2%) Electrocautery: 52 (15%) Stapler: 144 (43%) Habib: 16 (5%) Water jet: 53 (16%) Others or combination: 66 (19%) 50 (15%) 16 (5%) 24 (7%) 10 (3%) 23 (7%) 127 (37%) 87 (26%) Median: 7.5 Range: 6–23 I: 27 (8%) II: 191 (58%) III: 103 (31%) IV: 5 (2%) V: 1 (0.3%) Missing: 1 (0.3%) I. Malignancies, n = 262 (77.7%) HCC: 68 CCC: 40 Metastases: 143 Others: 11 II. Benign tumors, n = 70 (20.8%) Cysts: 23 FNH: 8 Adenoma: 10 Hemangioma: 6 Others: 23 III. Liver injury, n = 5 (1.5%)
TACE: transarterial chemoembolization; RFA: radio-frequencyassisted ablation; MELD: model for end-stage liver disease; CUSA: cavitron ultrasonic surgical aspirator; HABIB: bipolar resection device; ASA: American Society of Anesthesiologists; HCC: hepatocellular carcinoma; CCC: cholangiocarcinoma; FNH: focal nodular hyperplasia.
Intra- and postoperative data Univariate analysis showed statistically significant differences for morbidity, re-operation, and in-hospital mortality for elevated intra- and postoperative transfusions of packed red blood cells (PRBC) and fresh frozen plasma (FFP), longer operating time, as well as longer ICU stay and longer duration of mechanical ventilation (Table 3). Noteworthy is the fact that postoperative admission to the ICU was a former standard procedure for most patients after liver resections at our hospital. Multivariate Analysis
In the multivariate analysis of preoperative parameters, independent risk factors for in-hospital mortality were preoperatively elevated serum bilirubin (OR: 1.441; 95% CI: 1.097–1.894; p = 0.009) and lactate levels (OR: 1.474; 95% CI: 1.145–1.898; p = 0.004). Regarding independent risk factors for short-term re-operation, the multivariate analysis revealed only preoperatively elevated lactate level (OR: 1.294; 95% CI: 1.116–1.5; p = 0.001), whereas independent risk factors for postoperative morbidity were preoperatively elevated bilirubin (OR: 1.43; 95% CI: 1.107–1.849; p = 0.008) level, preoperatively reduced hematocrit (OR 0.926; 95%CI 0.881–0.974; p = 0.003), and long-term medication with diuretics (OR: 2.054; 95% CI: 1.147–3.676; p = 0.016). More detailed analysis of postoperative morbidity provided preoperatively elevated bilirubin and lactate levels as well as long-term medication with diuretics as independent risk factors, whereas preoperatively reduced hematocrit and sodium levels and long-term medication with diuretics were provided as independent risk factors for non-surgical complications. Calculated Cutoff Values Of Selected Blood Tests As Preoperative Predictors
Preoperative serum bilirubin levels >3.2 mg/dL (normal range up to 1.2 mg/dL) were associated with a mortality rate of 22.6%, whereas patients with preoperative serum bilirubin levels ⩽3.2 mg/dL had a mortality rate of 2.3% (p 4.5 mmol/L (normal range up to 1.8 mmol/L) as compared to 3.0% in patients with serum lactate levels ⩽4.5 mmol/L (p = 0.005). Another important predictor was the preoperative hemoglobin level; there was a mortality rate of 10.7% in patients with hemoglobin levels ⩽10.3 g/dL (normal range 12–16 g/dL) as compared to 2.1% in patients with preoperative hemoglobin level >10.3 g/dL (p = 0.030) (Table 4). Global cutoff values for preoperative serum levels of bilirubin >3.3 mg/dL (p 1.2 mmol/L (p
SJS0010.1177/1457496914548093Predictors in liver surgeryC. Riediger, et al.
Original Article
Scandinavian Journal of Surgery 0: 1–9, 2014
Preoperative Serum Bilirubin And Lactate Levels Predict Postoperative Morbidity And Mortality In Liver Surgery: A Single-Center Evaluation C. Riediger1,2, M. W. Mueller1,3, A. Hapfelmeier4, J. Bachmann1, H. Friess1, J. Kleeff1 1
Department of Surgery, Technische Universität München, Munich, Germany Department of General, Thoracic and Vascular Surgery, University Hospital Carl Gustav Carus, Technische Universität Dresden, Dresden, Germany 3 Department of Surgery, Krankenhaus Bad Cannstatt, Klinikum Stuttgart, Stuttgart, Germany 4 Institute of Medical Statistics and Epidemiology, Technische Universität München, Munich, Germany 2
Abstract
Background and Aims: In spite of huge developments in liver surgery during the last decades, morbidity and mortality continue to pose problems in this field. The aim of this study was to identify preoperative predictors for postoperative mortality and morbidity in liver surgery. Material and Methods: In a single-center study, an extensive analysis of a prospective database, including clinical criteria and laboratory tests of patients undergoing liver surgery between July 2007 and July 2012 was performed. Cutoff values of selected laboratory tests were calculated. Results: In all, 337 patients were included in the study. Univariate analysis showed a statistically significant association of preoperative bilirubin, lactate, hemoglobin levels, platelet count, and prothrombin time with postoperative morbidity and mortality. Multivariate analysis revealed preoperatively elevated serum bilirubin and lactate levels as independent predictors for increased postoperative morbidity and mortality after liver surgery. Conclusions: The identified laboratory values showed a statistically significant association with postoperative morbidity and mortality in liver surgery and might be helpful in preoperative patient selection. Key words: Liver surgery; preoperative predictors; postoperative morbidity; postoperative mortality; bilirubin; lactate
C.R. and M.W.M contributed equally to the writing of this article. Correspondence: Carina Riediger, M.D. Department of General, Thoracic and Vascular Surgery University Hospital Carl Gustav Carus Technische Universität Dresden Fetscherstrasse 74 01307 Dresden Germany Email: [email protected]
2
C. Riediger, et al.
Introduction Improvements in liver surgery have resulted in remarkable reductions in mortality and morbidity over the last few decades. Due to the rapid development of liver dissection techniques, as well as improvements in anesthetics, the reported mortality rate of 50% for major liver resection in the 1970s has decreased dramatically (1). Even though the mortality rate is nowadays less than 5%, morbidity rates of up to 40% are still reported (2). Primarily, bleeding and biliary complications, as well as infections, are clinically the most serious problems (3). The high morbidity rates are explained by the rising numbers of liver resections due to the higher incidence of liver tumors (especially metastasis) and the increased percentage of extended resections (4). Moreover, patients undergoing liver surgery are increasingly older and present with more co-morbidity (5). Different studies in the literature discuss possible risk factors with regard to the outcome after liver surgery. Some authors analyzed factors influencing short-term outcome, while others focused on long-term survival, especially in cancer patients (6, 7). It has been shown that postoperative morbidity has an additional impact on the long-term and disease-free survival in cancer patients, and that the reduction of perioperative morbidity can also improve the oncological outcome of curative resection (7, 8). According to the literature, some risk factors for postoperative morbidity and mortality are defined, for example, higher ASA (American Society of Anesthesiologists) classification, steatohepatitis, liver cirrhosis, extent of resection, operating time, need for transfusions, and selected laboratory tests (9–13). Many groups have analyzed defined subgroups of patients with specific malignancies (hepatocellular carcinoma, colorectal liver metastases), or underlying liver disease like liver cirrhosis (14). The majority of studies focus on intra- and postoperative risk factors of liver surgery. Only a few groups have analyzed preoperative parameters to predict postoperative mortality and morbidity (11). The aim of this single-center study was to define predictors that help to identify preoperatively patients at risk for increased postoperative morbidity and mortality in liver surgery. Methods Patients
In this single-center study, patients’ data from July 2007 to July 2012 were prospectively recorded in an electronic database. Patients who received liver resections simultaneously with another operation such as colorectal resection, gastrectomy, esophagectomy, or pancreatectomy were excluded from the analysis, as were atypical liver resections during explorative laparotomy/laparoscopy and “secondary” liver resections. In all, 337 patients met inclusion criteria and were finally included in the database analysis. Patient characteristics, pre- and postoperative treatment, laboratory values, operative procedures, intraand perioperative data, complications, hospital stay,
and others were recorded. Analysis of surgical procedures included anatomical segmentectomies, nonanatomical segmentectomies, right and extended right hemihepatectomies as well as left and extended left hemihepatectomies. It is of note that only non-laparoscopic liver resections were included in this analysis. The study was performed along the principles of the Declaration of Helsinki and approved by the local ethical committee. A written informed consent of patients was obtained. Statistical Analysis
Statistical analysis was performed using SPSS software, version 20 (SPSS Inc., Chicago, IL, USA) and R for Statistical Computing (R Foundation for Statistical Computing, Vienna, Austria). Descriptive statistics were given by median and range or interquartile range (IQR) for continuous data. The distribution of categorical data was presented in contingency tables showing absolute and relative frequencies. All statistical tests were conducted in an explorative manner on a twosided 5% (α = 0.05) significance level. Univariate analysis Univariate analysis was performed to identify risk factors associated with rates of mortality, re-operation, morbidity, and surgical or medical complications. Mortality was defined as in-hospital mortality or 30-day mortality. Morbidity included surgical and medical complications. Surgical complications were defined as abdominal wall dehiscence, wound infection, intra-abdominal abscess, liver abscess, liver necrosis, liver hematoma, bilioma, icterus, biliary fistula, insufficiency of bile duct anastomosis, ascites, pneumothorax, ischemia, intra-abdominal bleeding, perforation of the gut, and mechanical ileus. Medical complications were defined as aspiration, upper or lower gastro-intestinal (GI)-bleeding, perforation of stomach ulceration, venous thrombosis, lung embolism, pleural effusion, pneumonia, other pulmonary complications, cardiac ischemia, cardiac arrhythmia, other cardiac complications, urinary infection, temporary dialysis, permanent dialysis, cholangitis, sepsis, liver failure, multi-organ failure, seizures, and damage to the central nervous system. Gender and age were analyzed, as well as patients’ ASA score. Moreover, preoperative clinical symptoms (pain, jaundice, pruritus, bleeding complications, loss of weight, and reduced general condition); medical history, including co-morbidities (such as cardiac, pulmonary or renal diseases, diabetes mellitus, viral hepatitis, autoimmune hepatitis or liver cirrhosis with portal hypertension or cholelithiasis); and long-term medication (diuretics, aspirin, coumarines, vitamin K, oral contraceptives, or immunosuppressants) as well as specific oncological pre-treatments (chemotherapy, radiation, transarterial chemoembolization (TACE), radiofrequency-assisted ablation (RFA), and others) were analyzed in univariate analyses. Univariate analysis of the patient’s blood tests on day 1/0 preoperative to surgery, as well as analysis of intraoperative data (operation time, need of blood transfusion, and
Predictors in liver surgery
extent of liver resections) and postoperative data (need of transfusions, intensive care unit (ICU) stay, and days of mechanical ventilation) was performed. To investigate the relationship between these potential risk factors and outcome, Mann–Whitney U tests were performed for continuous data and χ2 tests or Fisher’s exact test, depending on the cell counts of the respective contingency tables, for categorical data. Multivariate analysis Parameters that showed statistically significant differences in regard to mortality, overall morbidity, surgical complications, medical complications, and re-operations in the univariate analyses were included in a multivariate analysis. The multivariate analysis was performed by logistic regression models designed with stepwise forward variable selection based on Akaike’s information criterion (AIC). The odds ratio (OR) and corresponding 95% confidence intervals (CIs) are presented for each predictor to quantify its relation to the investigated risks. Definition of cutoff values Cutoff values of selected preoperative laboratory tests (hemoglobin, hematocrit, lactate, bilirubin, platelet count) were calculated to define risk groups with maximally divergent risks for surgical or medical complications, morbidity, mortality, or re-operation. Statistical significance of such maximally selected test statistics was assessed in a permutation test framework (15). A global cutoff value for each of the following laboratory tests was performed: serum hemoglobin, hematocrit, lactate, bilirubin, and platelet count. The global cutoff value was determined for a combined outcome that accounts for any event resulting from mortality, morbidity, surgical complications, medical complications, and re-operation. Decision trees for preoperative blood tests Decision trees perform successive binary splits in the selected preoperative laboratory results of lactate and bilirubin to produce risk groups for the outcomes mortality and morbidity (including surgical and medical complications) (15). They can be seen as a multivariate extension of the univariate cutoff determination described in the preceding steps. After the first split is found to separate the patient cohort into a high-risk and a low-risk group, it might be reasonable to divide these subsets even further. If this is the case, additional cutoff values are searched for within each subgroup. The procedure stops when there are no meaningful separations left. Results Patients’ Characteristics
Between July 2007 and July 2012, a total of 337 liver resections were finally included in our analysis. Detailed information about the analyzed cohort is
3
shown in Table 1. The overall mortality of the cohort was 3.7%, whereas the overall morbidity was 32.7% with an overall re-operation rate of 8.3%. Univariate Analysis
Age and gender The median age was 64.5 years (range: 20–87 years). The cohort consists of 189 male (56%) and 148 female (44%) patients. There were no statistical significant differences in gender and age with regard to mortality, re-operation, and morbidity, surgical or medical complications. Nevertheless, some tendencies were observed: patients with medical complications showed a higher median age of 67.1 years compared to a median age of 63.8 years for patients without medical complications (p = 0.179). Similar results were seen with regard to in-hospital mortality: patients who died within 30 days after surgery showed a median age of 70.7 years (range: 31.5–85.9 years), whereas patients without mortality had a median age of 64.3 years (range: 19.9– 86.7 years) (p = 0.198). The group of patients who died consisted of 69% male and 31% female patients. Of note, the mortality rate within the female cohort was 2.7% and within the male cohort 4.8%. Medical history, long-term medication, pre-treatment, and ASA score Patients presenting with increased postoperative morbidity had statistically significant more preoperative weight loss (p = 0.049) and blood loss (p = 0.046) had an increased rate of diabetes mellitus (p = 0.034), cardiac disease (p = 0.016), and long-term diuretic medication (p = 0.006), or were pre-treated with TACE (p = 0.019). Cardiac disease (p = 0.001) and long-term medication with aspirin (p = 0.010) and diuretics (p = 0.035) were associated with higher rates of surgical complications, whereas cardiac disease, diabetes mellitus (p = 0.036), liver cirrhosis (p = 0.033), and long-term medication with diuretics (p = 0.001) and preoperative chemotherapy (p = 0.024) were associated with medical complications. The group of patients in need of early re-operation showed a significant higher number of patients with long-term diuretic medication (p = 0.029). Neither medical history and long-term medication nor pre-treatment showed a statistically significant relation to in-hospital mortality. However, higher ASA score was statistically significant associated with higher postoperative morbidity (p = 0.002) and mortality (p = 0.001) and need for operative revision (p = 0.008). Preoperative laboratory results The univariate analysis revealed that preoperative elevated serum lactate and bilirubin levels, as well as preoperative reduced prothrombin time, hemoglobin levels, and platelet count are associated with significantly higher rates of postoperative morbidity (including surgical and medical complications), re-operation, and in-hospital mortality. Detailed results are shown in Table 2.
4
C. Riediger, et al. Table 1 Patients’ characteristics.
Number of liver resections Number of patients Age (years) Gender Liver cirrhosis Hepatitis B C Pretreatment Chemotherapy TACE RFA Resection technique Extent of liver resections Right hemihepatectomy Left hemihepatectomy Extended right hemihepatectomy Extended left hemihepatectomy Cyst deroofing Segmentectomy atypical Segmentectomy anatomical MELD score ASA score Diagnosis
N = 337 N = 337 Median: 64.5 Range: 20–87 Male: 189 (56%) Female: 148 (44%) 48 (15%) 12 (4%) 16 (5%) 104 (32%) 79 (25%) 18 (6%) 19 (6%) CUSA: 6 (2%) Electrocautery: 52 (15%) Stapler: 144 (43%) Habib: 16 (5%) Water jet: 53 (16%) Others or combination: 66 (19%) 50 (15%) 16 (5%) 24 (7%) 10 (3%) 23 (7%) 127 (37%) 87 (26%) Median: 7.5 Range: 6–23 I: 27 (8%) II: 191 (58%) III: 103 (31%) IV: 5 (2%) V: 1 (0.3%) Missing: 1 (0.3%) I. Malignancies, n = 262 (77.7%) HCC: 68 CCC: 40 Metastases: 143 Others: 11 II. Benign tumors, n = 70 (20.8%) Cysts: 23 FNH: 8 Adenoma: 10 Hemangioma: 6 Others: 23 III. Liver injury, n = 5 (1.5%)
TACE: transarterial chemoembolization; RFA: radio-frequencyassisted ablation; MELD: model for end-stage liver disease; CUSA: cavitron ultrasonic surgical aspirator; HABIB: bipolar resection device; ASA: American Society of Anesthesiologists; HCC: hepatocellular carcinoma; CCC: cholangiocarcinoma; FNH: focal nodular hyperplasia.
Intra- and postoperative data Univariate analysis showed statistically significant differences for morbidity, re-operation, and in-hospital mortality for elevated intra- and postoperative transfusions of packed red blood cells (PRBC) and fresh frozen plasma (FFP), longer operating time, as well as longer ICU stay and longer duration of mechanical ventilation (Table 3). Noteworthy is the fact that postoperative admission to the ICU was a former standard procedure for most patients after liver resections at our hospital. Multivariate Analysis
In the multivariate analysis of preoperative parameters, independent risk factors for in-hospital mortality were preoperatively elevated serum bilirubin (OR: 1.441; 95% CI: 1.097–1.894; p = 0.009) and lactate levels (OR: 1.474; 95% CI: 1.145–1.898; p = 0.004). Regarding independent risk factors for short-term re-operation, the multivariate analysis revealed only preoperatively elevated lactate level (OR: 1.294; 95% CI: 1.116–1.5; p = 0.001), whereas independent risk factors for postoperative morbidity were preoperatively elevated bilirubin (OR: 1.43; 95% CI: 1.107–1.849; p = 0.008) level, preoperatively reduced hematocrit (OR 0.926; 95%CI 0.881–0.974; p = 0.003), and long-term medication with diuretics (OR: 2.054; 95% CI: 1.147–3.676; p = 0.016). More detailed analysis of postoperative morbidity provided preoperatively elevated bilirubin and lactate levels as well as long-term medication with diuretics as independent risk factors, whereas preoperatively reduced hematocrit and sodium levels and long-term medication with diuretics were provided as independent risk factors for non-surgical complications. Calculated Cutoff Values Of Selected Blood Tests As Preoperative Predictors
Preoperative serum bilirubin levels >3.2 mg/dL (normal range up to 1.2 mg/dL) were associated with a mortality rate of 22.6%, whereas patients with preoperative serum bilirubin levels ⩽3.2 mg/dL had a mortality rate of 2.3% (p 4.5 mmol/L (normal range up to 1.8 mmol/L) as compared to 3.0% in patients with serum lactate levels ⩽4.5 mmol/L (p = 0.005). Another important predictor was the preoperative hemoglobin level; there was a mortality rate of 10.7% in patients with hemoglobin levels ⩽10.3 g/dL (normal range 12–16 g/dL) as compared to 2.1% in patients with preoperative hemoglobin level >10.3 g/dL (p = 0.030) (Table 4). Global cutoff values for preoperative serum levels of bilirubin >3.3 mg/dL (p 1.2 mmol/L (p
