Patients with cirrhosis have a reduced life expectancy. Anesthesia and surgery have been associated with clinical decompensation in patients with cirrhosis.
The authors retrospectively reviewed the records of all patients with the diagnosis of cirrhosis who underwent any surgical procedure under anesthesia at their institution between January 1980 and January 1991 (n = 733). Univariate and multivariate analyses were used to identify the variables associated with perioperative complications and short- and long-term survival.
The perioperative mortality rate (within 30 days of surgery) was 11.6%. The perioperative complication rate was 30.1%. Postoperative pneumonia was the most frequent complication. Multivariate factors that were associated with perioperative complications and mortality included male gender, a high Child-Pugh score, the presence of ascites, a diagnosis of cirrhosis other than primary biliary cirrhosis (especially cryptogenic cirrhosis), an elevated creatinine concentration, the diagnosis of chronic obstructive pulmonary disease, preoperative infection, preoperative upper gastrointestinal bleeding, a high American Society of Anesthesiologists physical status rating, a high surgical severity score, surgery on the respiratory system, and the presence of intraoperative hypotension.
Risk factors have been identified for patients with cirrhosis who undergo anesthesia and surgery.
PATIENTS with cirrhosis have a reduced life expectancy. Gines et al. reported a median survival time of 8.9 yr for patients (mean age, 50.2 years) with newly diagnosed cirrhosis. The median survival time decreased to 1.6 yr in patients after the onset of the first major complication of cirrhosis (ascites, jaundice, encephalopathy, or gastrointestinal hemorrhage). Anesthesia and surgery are known to have decompensatory effects on patients with cirrhosis. Aranha et al. reported a 25% perioperative mortality rate for those patients with cirrhosis who underwent open cholecystectomy. Those patients who died frequently experienced postoperative bleeding, renal failure, and sepsis. We did this retrospective investigation to identify those factors that might be predictive of perioperative complications and death in patients with cirrhosis who were undergoing anesthesia and surgery.
Materials and Methods
After we received approval from our institutional review board, we reviewed the records of all adult patients (age >or= to 18 yr) who had a diagnosis of cirrhosis and underwent any surgical procedure with anesthesia between January 1980 and January 1991 at the Mayo Clinic. January 1991 was selected as an end point for this review to allow sufficient time to follow those patients who survived the perioperative period. The minimum follow-up period was 2.7 yr, and the maximum follow-up time was 13.7 yr. Some patients had more than one operating room stay during the study. For these patients, data were collected and analyzed for only the first perioperative experience. Patients who underwent orthotopic liver transplantation before or during the study were excluded.
Cirrhosis was diagnosed primarily by liver biopsy. In patients in whom liver biopsy was not performed, the diagnosis was made by a history of liver disease with impaired liver function tests, plus one or more of the following: liver ultrasound or computed tomography scan suggesting the diagnosis of cirrhosis, or direct examination of a cirrhotic liver during abdominal surgery.
Preoperative data included age; gender; coexisting medical disease; year of surgery; type of cirrhosis; and the presence of ascites, esophageal varices, and encephalopathy. Ascites was graded as severe if it contained more than 2 1, as determined by imaging data, intraoperative findings, or clinical impression. Esophageal varices were graded on a scale of 1 to 4 according to upper gastrointestinal studies or direct esophagoscopy. Hepatic encephalopathy was graded on a scale of 1 to 4. Preoperative serum albumin, total bilirubin, creatinine concentration, and prothrombin time were recorded. Child's classification as modified by Pugh was used as a measure of liver dysfunction. 
Chronic obstructive pulmonary disease, hypertension, ischemic heart disease, congestive heart failure, and endocrine disease were considered present if they were listed in the patient's Mayo Clinic medical records and the patient was receiving medical therapy for the condition at the time of operation. Chronic renal failure was defined as a preoperative serum creatinine concentration more than 2 mg/dl, without evidence of recent deterioration in renal function. Each patient was categorized by an anesthesiologist just before surgery into one of five classes in the physical status classification established by the American Society of Anesthesiologists (ASA). The surgical setting was noted to be elective or emergent.
Surgical procedures were classified according to the International Classification of Diseases, 9th revision, Clinical Modification (Table 1). A surgical severity score (low, moderate, high) was assigned to each operative procedure. The surgical severity score was based on an estimation by the authors about the magnitude of tissue injury encountered with each of the surgical procedures (Table 1). These determinations occurred during a meeting in which the authors discussed each procedure individually. The anesthetic management was categorized as being general, regional, or monitored anesthesia care. Intraoperative transfusion requirements (erythrocyte units) and cardiovascular parameters were recorded.
Hypotension and hypertension were defined as a 20% decrease or increase of baseline blood pressure for 10 min or more, respectively. Cardiac arrhythmia was defined as any change in heart rate or rhythm that was treated. The occurrence of cardiopulmonary resuscitation was noted. Postoperative mortality was recorded during the same hospitalization and after discharge (see Data Collection). Postoperative morbidity was defined as outcomes that occurred during the same hospitalization or as long as 30 days after operation. Ventilation dependence was noted when mechanical ventilation lasted for more than 24 h after operation in patients not intubated before surgery. Ventilation dependence was also recorded when postoperative respiratory insufficiency required repeated intubation of the trachea and mechanical ventilation in patients whose tracheas were extubated before. Pneumonia, pulmonary embolism, adult respiratory distress syndrome, congestive heart failure, myocardial infarction, and arrhythmia were considered postoperative complications if they represented new findings, and they required that the patient receive additional medical therapy. Acute renal failure was defined as an increase of 2 mg/dl or more in the serum creatinine concentration compared with preoperative values, or the need for dialysis in a patient with previously adequate renal function. Hepatorenal syndrome was diagnosed when acute renal failure occurred in a setting of decreasing liver function with no other clinical, laboratory, or anatomic cause for the renal failure. Ascites was recorded as a complication if it was a new finding or if ascitic volume increased from mild to severe after operation. Infection was registered as being present if a patient had fever, leukocytosis, there was a clinical impression of infection, and antimicrobial medications were administered. Positive bacterial or fungal cultures supported the diagnosis of bacteremia or fungemia, respectively. Recorded surgical wound complications included dehiscence and infection. They were considered present if they were noted on the postoperative follow-up record and the patient received either medical or surgical therapy for the condition. Postoperative bleeding was defined as hemorrhage treated with transfusion of three or more units of packed erythrocytes. Intraluminal gastrointestinal and surgical site bleeding were distinguished from each other based on a review of the medical records. Coagulopathy (whether on the basis of reduced hepatic synthesis, disseminated intravascular coagulation, normal consumption, or dilutional mechanisms) was noted in those patients with decreased platelet count, fibrinogen concentration, and increased activated partial thromboplastin time. New onset of hepatic encephalopathy or a worsening of preoperative encephalopathy were registered as postoperative complications, as were repeated operations within the same hospital stay and within 30 days after the primary surgery. Patients were classified by the presence or absence of perioperative complications, including death within 30 days of surgery.
For the purpose of constructing the Kaplan-Meier survival curve, a written questionnaire was sent to all patients who were alive at the last follow-up according to their medical records. If no response was received, a second questionnaire was forwarded. Telephone interviews were conducted when no response was received in response to two written requests. The contents of the written questionnaire or telephone interview included questions regarding the patients' general health and asked specifically whether they had undergone liver transplantation. Those patients who had undergone liver transplantation were excluded from the investigation.
A chi-squared univariate analysis was performed to identify risk factors for complications within 30 days of operation. Logistic regression with backward elimination of nonsignificant variables was used to identify a set of multivariate predictors of complications within 30 days of operation and 30 days of survival. Kaplan-Meier survival curves were constructed to estimate survival and death over time. The log-rank test was used to compare survival curves across groups of patients. The Cox proportional hazards model was used to identify multivariate predictors of long-term survival, again using backward elimination of nonsignificant variables. Unless otherwise stated, two-tailed P values <or= to 0.05 were used to reflect findings not attributable to chance. Those variables deemed significant by multivariate analysis at P <or= to 0.001 were considered risk factors for perioperative complications and death.
Seven hundred thirty-three patients with the diagnosis of cirrhosis who underwent surgical procedures and anesthesia care at the Mayo Clinic between January 1980 and January 1991 qualified for the study. There were 338 women and 395 men with a mean (+/− standard deviation) age of 59.4 +/− 13.1 yr (range, 18-87 yr). Biopsy proof of cirrhosis was obtained in 563 patients (76.8%). The mean follow-up time after surgery was 4.12 +/− 3.8 yr (range, 2.7 to 13.7 yr) overall and 7.7 +/− 3.3 yr (range, 0 to 13.7 yr) for patients who were alive at the last follow-up. Three patients (0.4%) were lost to follow-up. For all patients, the time from surgery to hospital discharge was 10.3 +/− 14 days (range, 0-276 days). For patients who were discharged alive (n = 650), the mean hospital stay was 9.7 +/− 13.9 days (range, 0-276 days). Table 1summarizes the surgical procedures.
(Table 2) shows postoperative complications. Two hundred twenty-two patients (30.1%) had one or more postoperative complications. Pneumonia was diagnosed in 59 patients (8%) and was the most common postoperative complication. Other common postoperative complications included ventilatory dependence in 57 patients (7.8%), infection in 55 (7.5%), new-onset or worsening ascites in 49 patients (6.7%), and cardiac arrhythmia in 37 patients (5%).
Univariate analysis identified male gender, a high Child-Pugh score, an elevated prothrombin time, a low serum albumin level, the presence of ascites, the presence of varices, a diagnosis of cirrhosis other than primary biliary cirrhosis, and an elevated serum creatinine concentration as being associated with the occurrence of perioperative complications (Table 3, Table 5). Preoperative coexisting diseases, including chronic renal failure, chronic obstructive pulmonary disease, pneumonia, congestive heart failure, ischemic heart disease, insulin-dependent diabetes mellitus, and the presence of preoperative infection or upper gastrointestinal bleeding were associated with the occurrence of complications (Table 3). Univariate analysis identified factors concerning anesthesia management, including a high ASA classification, emergency surgery, and general anesthesia as being associated with the occurrence of complications (Table 3). Those surgical procedures that were univariately associated with the presence of complications included cardiovascular operations, surgery for portal-systemic shunt, splenectomy, digestive tract procedures, and hip and pelvic surgery (Table 3). A high surgical severity score and the presence of intraoperative hypotension were also statistically associated by univariate analysis with the occurrence of complications.
Multivariate analysis identified 10 variables that were independently associated with the occurrence of perioperative complications (P < 0.001). These multivariate associations included a high Child-Pugh score, the presence of ascites, a diagnosis of cirrhosis other than primary biliary cirrhosis, an elevated creatinine concentration, the diagnosis of chronic obstructive pulmonary disease, preoperative infection, preoperative upper gastrointestinal bleeding, a high ASA physical status classification, intraoperative hypotension, a high surgical severity score (Table 3), but not the type of anesthetic management. These variables with such strong independent associations with the occurrence of perioperative complications can be considered risk factors for anesthesia and surgery in patients with cirrhosis. This statement is supported by the findings shown in Figure 1, which illustrates the correlation between the number of risk factors identified by multivariate analysis and the rate of perioperative complications.
Within 30 days of surgery, 72 patients (9.8%) died in the hospital and 13 (1.8%) died outside the hospital. Eighty-three patients (11.3%) died at some time during the hospitalization. Figure 2shows that the most precipitous decrease in patient survival time occurred during the immediate postoperative period. Univariate analysis identified male gender, a high Child-Pugh score, an elevated prothrombin time, an elevated total bilirubin level, a low serum albumin level, the presence of ascites, the diagnosis of cryptogenic cirrhosis, and an elevated serum creatinine concentration as being associated with an increased mortality rate (Table 4, Table 6). The preoperative coexisting diseases that were associated with an increased mortality rate included chronic renal failure, chronic obstructive pulmonary disease, pneumonia, congestive heart failure, ischemic heart disease, insulin-dependent diabetes mellitus, and the presence of preoperative infection (Table 4). Factors concerning anesthesia management that were univariately associated with an increase in the mortality rate included a high ASA classification and emergency surgery. Patients who underwent surgery under general anesthesia had a higher 30-day mortality rate compared with those who had regional or monitored anesthetic care. However, those patients who had their surgery under monitored anesthetic had the highest 6-month and 1-, 5-, and 10-yr mortality rates (Table 4). Those surgical procedures that were univariately associated with higher mortality rates included respiratory procedures, those cardiovascular procedures that required extracorporeal circulation, and biliary tract and liver procedures (Table 4). The occurrence of intraoperative hypotension was also statistically associated by univariate analysis with increased mortality rates.
Multivariate analysis identified eight variables that were independently associated with a high short- and long-term mortality rates (P < 0.001). These multivariate associations included male gender, a high Child-Pugh score, the presence of ascites, a diagnosis of cryptogenic cirrhosis, an elevated creatinine concentration, preoperative infection, a high ASA physical status, and surgery on the respiratory system (Table 4), but not the type of anesthetic management. These variables with such strong independent associations with short- and long-term can be considered risk factors for anesthesia and surgery in patients with cirrhosis. This statement is supported by the findings shown in Figure 3, which shows how the number of risk factors that a patient had before anesthesia and surgery increased the short- and long-term mortality rates.
This retrospective investigation again documents a high perioperative mortality risk for patients with cirrhosis undergoing anesthesia and surgery. Aranha et al. reported an overall 25% perioperative mortality rate in patients with cirrhosis who underwent open cholecystectomy. They compared the mortality rates of three groups of patients having open cholecystectomy: non-cirrhotic patients (1.1%), patients with cirrhosis with a prothrombin time less than 2.5 s greater than the control value (9.3%), and patients with cirrhosis with a prothrombin time more than 2.5 s greater than the control value (83%). Leonetti et al. reported a mortality rate of 8.3% for patients with cirrhosis who underwent umbilical hernia repair. They contrasted this rate of mortality with a mortality rate of 1.8% in noncirrhotic patients who had the same operation. Recently, Rice et al. reported a 28% perioperative mortality rate for patients with chronic liver failure who underwent nonhepatic surgery. Our investigation shows that 11.6% of patients with cirrhosis who received an anesthetic agent for any type of surgery (except liver transplantation) died within 30 days of surgery.
Hepatotoxicity related to anesthetic agents can occur on an immunologic basis, but the incidence is low and there is no evidence that cirrhosis is a predisposing factor. [14-18]In our investigation, we could not identify a patient with cirrhosis in whom postoperative liver failure was thought to have developed based on this criterion.
Direct traumatic effects of the surgical procedure on the liver can predispose patients with cirrhosis to postoperative liver insufficiency and death. The most extreme example of operative trauma is liver resection. However, less obvious surgical trauma (e.g., perihepatic surgical retraction and so forth) may result in postoperative liver dysfunction. In our investigation, surgery for liver and biliary disease was univariately associated with increased morbidity and mortality rates, but multivariate analysis eliminated it as an independent variable. In measuring the simultaneous effect of multiple variables, multivariate analysis did not show that surgery for liver and biliary disease was associated with reduced patient survival.
Because patients with cirrhosis have a reduced life expectancy, it is reasonable to consider that some patients may undergo clinical deterioration while they are undergoing anesthesia and surgery without the need to postulate intervening factors. In support of this, our investigation identified by multivariate analysis several risk factors indicative of poor and possibly declining physical status (high Child-Pugh score, elevated creatinine concentration, high ASA physical status, and so on). Our retrospective analysis could not determine whether these risk factors can be altered in such a way to result in improved perioperative outcome.
During anesthesia and surgery, liver blood flow is altered. The liver anatomically and physiologically adapts to altered perfusion pressure and blood flow. The organ is endowed with a dual blood supply (hepatic artery and portal vein). Under normal conditions, each blood source provides the liver with approximately 50% of its oxygen supply. The hepatic artery can provide more than 50% of the oxygen supply by vasodilating during periods of reduced portal inflow. This “reciprocity of flow” of the hepatic artery was described by Hanson and Johnson. The portal venous system is essentially a passive vascular bed. Thus intraoperative decreases in blood pressure and cardiac output can result in a decrease in portal blood flow. Similarly, surgical manipulation in the splanchnic bed may also reduce portal blood flow. The hepatic artery's ability to respond by vasodilatation to these portal blood flow alterations are blunted, particularly by certain anesthetics (such as halothane) and by high anesthetic concentrations. Even patients with healthy livers can experience alterations in hepatic blood flow and cellular ischemia, as evidenced by asymptomatic liver enzyme elevations. With cirrhosis, the reciprocal flow relation between the hepatic artery and the portal vein is not well maintained, and the liver architecture is disrupted by fibrosis and regenerative nodules. These pathologic alterations of the normal patterns of liver blood flow may make the cirrhotic liver more prone to ischemia. Hepatic ischemia and necrosis can cause the release of a multitude of inflammatory mediators that can result in the development of multiorgan system failure. [23-31]In our investigation, intraoperative hypotension was associated with increased morbidity and mortality rates by univariate and multivariate analysis.
The hepatic enzymes that can become elevated in patient's serum after they receive anesthetics do not necessarily indicate severe hepatic injury. [32-34]
In this retrospective analysis, many variables were found to be associated with perioperative complications and death. Table 3and Table 4list many variables that were univariately, but not multivariately, associated with patient outcome. In measuring the simultaneous effect of multiple variables, multivariate analysis identified those variables that have an effect on outcome independent of the other measured variables. When two variables covary (as age and weight do in children), both may contribute to a univariate analysis, but only one does so in a multivariate analysis. This does not mean that only one of the covariables matters, but only that there is insufficient evidence to incorporate both into the multivariate model. The patients in this investigation who underwent extracorporeal bypass illustrate the importance of a variable that is associated with outcome by univariate but not by multivariate methods. Univariate analysis determined that postoperative complications occurred more frequently in patients undergoing cardiovascular procedures that required extracorporeal bypass. This statistical association was lost when multivariate analysis was applied. However, this may have been the result of the type 2 error that occurred from the low number of patients (n = 12) in this category. Thus it is possible that the use of extracorporeal support may be associated with increased risk in the patient with cirrhosis.
This retrospective investigation can only help to identify those variables that are associated with perioperative complications and death. These statistical associations, even if strong, do not necessarily indicate a cause-and-effect relation. However, investigators may wish to be mindful of these when they design future studies. In addition, clinicians may wish to consider these factors when making decisions regarding surgical selection, preoperative preparation, and perioperative management.
This investigation confirms high rates of morbidity and mortality for patients with cirrhosis who require anesthesia and surgery. Several factors associated with increased perioperative morbidity and mortality rates are identified.