INTRODUCTION — Patients with liver disease who require surgery are at greater risk for surgical and anesthesia related complications than those with a healthy liver [1-4]. The magnitude of the risk depends upon the type of liver disease and its severity, the surgical procedure, and the type of anesthesia.

The assessment of surgical risk in patients with liver disease will be reviewed here. Patients with liver disease may have concomitant disorders (such as cardiovascular disease) that influence surgical outcomes; these issues are discussed separately. (See "Preoperative medical evaluation of the healthy patient" and "Estimation of cardiac risk prior to noncardiac surgery".)

SCREENING FOR LIVER DISEASE BEFORE SURGERY — All patients undergoing surgery should undergo a careful history and physical examination to exclude findings or risk factors for liver disease. This should include asking specifically about prior blood transfusions, tattoos, illicit drug use, sexual promiscuity, a family history of jaundice or liver disease, a history of jaundice or fever following anesthesia, alcohol use (current, prior and quantity), and a complete review of current medications. Clinical features suggestive of liver disease (such as fatigue, pruritus, increased abdominal girth, jaundice, palmar erythema, spider telangiectasias, splenomegaly, and gynecomastia and testicular atrophy in men) should be evaluated.

Whether otherwise healthy surgical candidates should undergo biochemical screening for liver disease is controversial. The vast majority of patients found to have abnormal liver biochemical test results do not have advanced liver disease. Thus, it is unlikely that routinely obtaining a liver biochemical profile in otherwise healthy patients without risk factors for liver disease would lead to improved outcomes; thus, such testing is not recommended. (See "Preoperative medical evaluation of the healthy patient".)

EFFECTS OF ANESTHESIA AND SURGERY ON THE LIVER — The effects of anesthesia and surgery on the liver depend upon the type of anesthesia used, the specific surgical procedures, and the severity of liver disease. In addition, perioperative events, such as hypotension, sepsis, or the administration of hepatotoxic drugs, can compound injury to the liver occurring during the procedure. Although mild elevations in serum aminotransferase, alkaline phosphatase, or bilirubin concentrations are common following surgical procedures, regardless of the type of anesthesia used, they usually resolve without further consequences [5,6].

Clinically important hepatic dysfunction is uncommon, but is much more likely in patients with preexisting liver disease. A review of 733 patients with cirrhosis who underwent surgery at the Mayo Clinic found a perioperative mortality rate of 11.6 percent and a complication rate of 30 percent, which is much higher than would be expected for a population without liver disease [7]. A number of risk factors for mortality or complications were identified including a high Child-Pugh score, the presence of ascites, an elevated serum creatinine concentration, preoperative upper gastrointestinal bleeding, and a high American Society of Anesthesiologists physical status rating (table 1).

Type of surgery — The type of surgery is probably the most important determinant of postoperative hepatic dysfunction in patients with liver disease [1]. The following are general principles:

• Laparotomy is associated with greater reduction in hepatic arterial blood flow than extraabdominal surgery, in part because traction on the abdominal viscera can cause reflex systemic hypotension due to dilatation of capacitance vessels [8,9].
• Procedures associated with a large amount of blood loss increase the risk of ischemic hepatic injury [10].
• Patients with cirrhosis who have had prior abdominal surgery can have highly vascular adhesions surrounding the liver, which can bleed excessively during surgery.
• Emergency surgery is associated with higher morbidity and mortality than elective surgery.
• Cholecystectomy, gastric surgery, and colectomy have been associated with particularly high mortality rates in patients with decompensated cirrhosis [11-20]. More recent data suggest that the outcome of cholecystectomy has improved. A MELD score (see below) of ≥8 was associated with increased 30- and 90-day morbidity rates in one study [21]. In selected cases of patients with compensated cirrhosis, a laparoscopic approach may be feasible [22,23]. Nevertheless, less invasive options, such as cholecystostomy, should be considered, especially in patients with advanced cirrhosis. (See "Treatment of acute cholecystitis".)

• Cardiac surgery in patients with cirrhosis is also associated with a high mortality (see 'Cardiac surgery' below [24].

Hemodynamic effects of anesthesia on the liver — Studies on healthy patients have demonstrated that the effects of general anesthesia on the hepatic circulation are variable. As a general rule, most inhalation anesthetics are associated with reductions in portal blood flow and cardiac output [25]. The effect on hepatic arterial blood flow depends upon the specific anesthetic used and the type of operation. As examples, the volatile anesthetics halothane and enflurane reduce hepatic arterial blood flow because of systemic vasodilatation and a slightly negative inotropic effect [5,26,27], while isoflurane increases or has a neutral effect on hepatic arterial blood flow, and has therefore been preferred in patients with liver disease [28]. Because most of the blood supply to the liver is provided by the portal vein, the net effect of these hemodynamic changes is usually to decrease hepatic blood flow and oxygen delivery.

The degree to which these changes take place in patients with advanced liver disease has not been well studied in humans. Such patients have a number of baseline circulatory abnormalities, including an increased cardiac index and decreased systemic vascular resistance, which could influence hepatic blood flow during anesthesia [29-31]. In an animal model of cirrhosis, the increased cardiac index was maintained during general anesthesia and was associated with increased portal venous inflow compared to controls, suggesting that the hyperdynamic circulation in cirrhosis might protect against hepatic hypoperfusion during general anesthesia [32]. On the other hand, patients with cirrhosis may be more sensitive to hepatic hypoperfusion relative to baseline conditions. As a result, they may be more susceptible to liver injury, particularly if they are faced with additional insults to the liver (such as administration of a hepatotoxic drug or the development of infection).

Hepatic metabolism of inhaled anesthetics — Anesthetic agents are capable of producing hepatotoxicity by a variety of mechanisms. (See "Drugs and the liver: Patterns of drug-induced liver injury".) As a general rule, the degree of hepatic metabolism correlates with the likelihood of a toxic reaction. Isoflurane and the newer haloalkanes (ie, desflurane and sevoflurane), undergo less hepatic metabolism than halothane or enflurane (0.2 percent for isoflurane, 20 percent for halothane, and 2 to 4 percent for enflurane), and have been associated with the lowest risk of hepatitis [5,33,34].

Of the currently used haloalkanes, halothane has been associated with the greatest risk of hepatotoxicity. The incidence of fulminant hepatitis following halothane exposure is approximately 1 in 6000 to 1 in 35,000 following a single exposure, and increases to about 1 in 3000 following multiple exposures [35,36]. Other risk factors for severe hepatotoxicity include female gender, age greater than 50, and obesity [35]. (See "Halothane hepatitis".)

Concern about hepatotoxicity has virtually eliminated the use of halothane in adults in the United States. However, halothane toxicity has not been reported in children. Because it is highly potent and has a tolerable odor, halothane continues to be used widely for the induction of anesthesia in children in the United States. Furthermore, halothane is possibly the most commonly used inhalational anesthetic in all patients worldwide.

Mechanical ventilation — Patients with liver disease commonly have respiratory compromise due to a number of causes either directly related to the liver disease (such as hepatopulmonary syndrome, pleural effusions, pulmonary hypertension) or because of associated conditions such as COPD [7,37]. Severe hypoxemia (p02 <60 mmHg) is usually considered to be a relative contraindication to surgery (except for liver transplantation in patients with hepatopulmonary syndrome), but is fortunately uncommon [38]. (See "Hepatopulmonary syndrome".)

Hypercarbia should be avoided in patients with liver disease during mechanical ventilation since it initiates sympathetic stimulation of the splanchnic vasculature, thereby decreasing portal blood flow. Patients with advanced liver disease usually have a mild compensated respiratory alkalosis [39]. Thus, the pCO2 should be maintained in the range of 35 to 40 mmHg during surgery [28]. A higher-than-usual concentration of oxygen during anesthesia has also been recommended for patients with cirrhosis [28].

Neuromuscular blocking agents — Neuromuscular blocking agents may contribute to the changes in liver blood flow during general anesthesia discussed above [40]. Furthermore, their effects may be prolonged in patients with liver disease because of decreased biliary excretion, reduced plasma pseudocholinesterase activity, and an increase in volume of distribution [41,42].

Atracurium and cisatracurium are favored for patients with liver disease or biliary obstruction because their metabolism does not depend upon the liver or kidneys [28,43]. Doxacurium (a long-acting, nondepolarizing muscle relaxant) is recommended for prolonged surgical procedures such as liver transplantation [44].

Analgesics and sedatives — Shorter acting analgesics and sedatives are preferred in patients with liver disease since longer acting agents increase the risk of postoperative delirium and hepatic encephalopathy. The effects of liver disease on the metabolism of these drugs is variable. As an example, although the metabolism of thiopental is decreased in patients with cirrhosis, plasma protein binding is also decreased, so that total body clearance is unaltered [45].

Patients with liver disease have an increased risk for rare hepatotoxic reactions to benzodiazepines or barbiturates [46]. The pattern of injury is usually cholestatic, and sometimes includes a syndrome of fever, hepatitis, lymphadenopathy, eosinophilia, and dermatitis.

ESTIMATING SURGICAL RISK — Assessment of surgical risk in patients with liver disease includes an appraisal of the severity of liver disease, the urgency of surgery (and alternatives to surgery), and coexisting medical illness. Surgical risk assessment is less relevant if immediate surgery is required to prevent death. On the other hand, the vast majority of decisions are made in the setting of semi-urgent or elective procedures for which there is time for risk assessment, optimization of the patient's medical status, and consideration of alternative approaches.

The majority of studies examining the risk of surgery in patients with liver disease have focused on patients with cirrhosis from which a number of risk factors have been identified (table 2) [7,11-19,47-49]. Much less information has been published on the risk of surgery in patients with milder forms of liver disease. The available evidence is derived mostly from small retrospective studies and clinical experience. Furthermore, many of the studies were published prior to the availability of a number of serologic tests for specific types of liver disease, and modern hepatobiliary imaging. Thus, there is relatively little information on the risk of surgery in patients with specific types of liver disease.

Patients in whom surgery is contraindicated — A number of settings have been identified that are associated with unacceptable surgical mortality. As a result, these conditions are usually considered to be contraindications to elective surgery (table 3).

Acute or fulminant hepatitis — Acute hepatitis is a contraindication to elective surgery. This recommendation is based upon older studies, in which operative mortality rates of 10 to 13 percent were reported among icteric patients who underwent laparotomy as part of a diagnostic evaluation that ultimately led to a diagnosis of acute viral hepatitis [50].

Similarly, patients with fulminant hepatitis are gravely ill and are unlikely to withstand surgery other than liver transplantation. (See "Acute liver failure: Definition; etiology; and prognostic indicators".)

Alcoholic hepatitis — Elective surgery is contraindicated in patients with histologic evidence of alcoholic hepatitis. Mortality rates as high as 55 to 100 percent have been observed in such patients undergoing open liver biopsy [51], portosystemic shunt surgery [52-54], or exploratory laparotomy [55]. (See "Clinical manifestations and diagnosis of alcoholic liver disease".)

However, it is possible that advances in surgical technique and postoperative care may have improved the outcome in such patients compared to the above studies, some of which were conducted more than 20 to 30 years ago. This was illustrated in a report from 1984, in which operative liver biopsy findings were reviewed in 164 patients with alcoholic cirrhosis and bleeding varices who underwent emergency portacaval shunt surgery [56]. Of these patients, 49 (30 percent) had histologic evidence of alcoholic hepatitis, but had survival rates similar to those without alcoholic hepatitis. These results have not been duplicated.

We recommend that elective surgery should be delayed for at least 12 weeks, or that a repeat liver biopsy should be performed to confirm resolution. The severity of underlying liver disease should be reassessed prior to making a final recommendation.

Severe chronic hepatitis — Surgical risk in patients with chronic hepatitis correlates with the clinical, biochemical, and histologic severity of disease. Patients with symptomatic and histologically severe chronic hepatitis have increased surgical risk, particularly in those with impaired hepatic synthetic or excretory function, portal hypertension, or bridging or multilobular necrosis on liver biopsy [57,58].

Patients at variable increased risk — The risk of surgery in patients with cirrhosis depends upon the severity of disease, the clinical setting and type of surgical procedure. For over 30 years, the principal predictor of operative risk in patients with cirrhosis has been the Child's classification, but newer studies suggest that the Model for End-Stage Liver Disease (MELD) score may be superior [59].

Child's classification — A number of retrospective studies have demonstrated that perioperative mortality and morbidity in patients with cirrhosis correlate well with the Child-Turcotte (table 4) [60] or Child-Pugh [7,61] classification of cirrhosis (table 5). In one study, for example, perioperative mortality rates of 10, 31, and 76 percent were observed in 100 patients with predominantly alcoholic cirrhosis undergoing abdominal surgery who were Child-Pugh class A, B, and C, respectively [49]. On multivariate analysis, the Child-Pugh classification was the best predictor of surgical mortality and morbidity. Nearly identical results were observed in a similarly designed study of 92 patients with cirrhosis (approximately 50 percent alcoholic) undergoing abdominal surgery (mortality rates of 10, 30, and 82 percent in patients with Child-Pugh class A, B, and C, respectively) [62].

Patients with Child's class A cirrhosis and portal hypertension are at increased risk of postoperative ascites, jaundice, and encephalopathy [63]. Limited observations suggest that postoperative morbidity may be reduced by preoperative placement of a transjugular intrahepatic portosystemic shunt [64].

Measures of hepatic function and the APACHE score — A number of measures of hepatic function have been proposed as predictors of perioperative morbidity and mortality in patients with cirrhosis. Examples include quantitative assessment of liver function with dynamic tests such as galactose elimination capacity, aminopyrine breath testing, indocyanine green clearance, and the rate of metabolism of lidocaine to monoethylglycinexylidide (MEGX). (See "Tests of the liver's biosynthetic capacity".) However, none has been shown convincingly to provide additional prognostic information compared to the Child-Pugh classification, and, as a result, they are not used widely [65].

The Acute Physiology, Age, and Chronic Health Evaluation System (APACHE III) score can predict survival in cirrhotic patients admitted to an intensive care unit [66]. However, it has not been studied specifically in cirrhotic patients undergoing surgery. (See "Predictive scoring systems for the severity of illness in the intensive care unit".)

MELD score — The MELD score is a statistical model predicting survival in patients with cirrhosis. It has been evaluated principally for selecting patients for liver transplantation. Use of this model for predicting surgical risk in the nontransplant setting has been promising and thus it may ultimately supplant the Child's classification as the principal method for determining surgical risk [21,67-71]. However, more studies involving diverse groups of surgical patients with a wide range of MELD scores are needed to understand its performance as a predictive model for surgery. The following summarizes representative studies. (See "Model for End-stage Liver Disease (MELD)".)

• The MELD score, American Society of Anesthesiologists class, and age predicted mortality in a study of 772 patients with cirrhosis who underwent major digestive, orthopedic, or cardiovascular surgery [67]. The MELD score was the best predictor of 30 and 90 day mortality. Mortality at 30 days ranged from 6 percent (MELD score, <8) to more than 50 percent (MELD score, >20) and correlated linearly with the MELD score.
• Another study compared the MELD score with the Child-Turcotte-Pugh classification in 40 cirrhotic patients who required either elective or emergency surgery with general anesthesia [69]. Emergency surgery was associated with significantly higher one- and three-month mortality rates. There was good correlation between the Child-Turcotte-Pugh classification and the MELD scores in predicting mortality, especially in the emergency surgery group.
• In other reports in selected settings, a MELD score ≥8 was useful for predicting morbidity after cholecystectomy [21]; a MELD score >14 was a better predictor of poor outcomes than the Child-Pugh classification in a series of 53 cirrhotic patients undergoing abdominal surgery [70], and a MELD score ≥11 predicted a high risk for postoperative liver failure in patients with cirrhosis undergoing hepatectomy for hepatocellular carcinoma [71].

It has been suggested that patients with a MELD score below 10 can undergo elective surgery, those with a MELD score of 10 to 15 may undergo elective surgery with caution, and those with a MELD score >15 should not undergo elective surgery [72]. A web site (www.mayoclinic.org/meld/mayomodel9.html) can be used to calculate the estimated 7-day, 30-day, 90-day, 1-year, and 5-year mortality rates after surgery based on the patient's age, ASA class, INR, serum bilirubin and creatinine. The model is based on the original MELD score, not the one currently being used for organ transplantation.

Obstructive jaundice — Patients with obstructive jaundice are at increased risk for several perioperative complications including infections (which result in part from bacterial colonization of the biliary tree, impaired Kupffer cell function, defective neutrophil function, and a high rate of endotoxemia), stress ulceration, disseminated intravascular coagulation, wound dehiscence, and renal failure [73-76]. Perioperative mortality ranged from 8 to 28 percent in several reports [77-79]. As an example, an overall mortality rate of 9 percent was found in a large retrospective study that included 373 patients undergoing surgery for obstructive jaundice [78]. Multivariate analysis identified three predictors of postoperative mortality:

• An initial hematocrit value <30 percent
• An initial serum bilirubin level >11 mg/dL (200 µmol/L)
• A malignant cause of obstruction (eg, pancreatic carcinoma or cholangiocarcinoma)

When all three factors were present, mortality approached 60 percent; when none was present, mortality was only 5 percent. Several other preoperative predictors of poor surgical outcome have been observed in other studies including azotemia, hypoalbuminemia, and cholangitis (table 6) [77-82]. The presence of portal hypertension can also be presumed to increase the surgical risk.

A number of interventions have been attempted to reduce morbidity and mortality in these patients:

• Perioperative administration of broad-spectrum intravenous antibiotics reduces the frequency of postoperative infections but does not influence mortality [83].
• External biliary drainage via a transhepatic approach has not been proven to improve morbidity or mortality in controlled studies [84-87].
• Endoscopic biliary drainage has the advantage of restoring enterohepatic circulation of bile acids while avoiding the complications of percutaneous puncture. However, as for external biliary drainage, it also has not been shown to improve surgical mortality in patients with a malignant cause of biliary obstruction [88], although preoperative biliary drainage has been recommended in patients undergoing extended hepatic resection [89]. In patients with cholangitis and choledocholithiasis, broad spectrum intravenous antibiotics and endoscopic drainage have been associated with lower mortality and morbidity rates compared to surgical decompression [90-92]. Although endoscopic sphincterotomy is associated with an increased rate of complications in patients with cirrhosis [93], morbidity and mortality rates are low even in patients with Child's class C cirrhosis when biliary decompression can be achieved [94]. (See "Endoscopic management of bile duct stones: Standard techniques and mechanical lithotripsy".)

• A major cause of morbidity in patients with obstructive jaundice is postoperative renal failure, which is usually due to acute tubular necrosis; the average frequency was approximately 8 percent in several reports [75,95,96]. The high incidence may be related to the absorption of endotoxin from the gut [97]. In normal subjects, endotoxin absorption is limited by the detergent effect of bile salts on the lipopolysaccharide endotoxin molecule; this protection is lost with obstructive jaundice, since bile salt secretion is minimal. As a result, patients may develop exaggerated renal vasoconstriction. (See "Pathogenesis and etiology of postischemic acute tubular necrosis".)

Limited evidence suggests that the administration of bile salts or lactulose to patients with obstructive jaundice can prevent both the endotoxemia and the exaggerated renal vasoconstriction [83,97-99]. In one report, for example, 102 patients with obstructive jaundice who had a serum bilirubin concentration >5.8 mg/dL (100 µmol/liter) were randomly assigned to receive lactulose, sodium deoxycholate (a bile salt) or no specific treatment prior to surgery [98]. Postoperative deterioration in renal function in patients with normal preoperative function was significantly more common in patients who had received no specific treatment.

Another approach that has been attempted to reduce the incidence of renal failure is the postoperative administration of mannitol [83,100]. Despite its theoretical benefit, maintenance of intravascular volume, and the avoidance of nephrotoxic drugs, such as aminoglycosides, are probably more critical elements in management [101-103].

Prophylactic oral antibiotics, such as rifaximin, have also been proposed as a means to reduce adverse effects of endotoxemia but a benefit has not yet been demonstrated. Furthermore, it is possible that oral antibiotics could increase endotoxemia because they may lead to increased release of endotoxin caused by destruction gram-negative organisms. On the other hand, intravenous broad-spectrum antibiotics should generally be given perioperatively to reduce the incidence of postoperative infection, although a benefit on mortality has not been demonstrated [83].

Whether patients with cholestatic liver disease (such as primary biliary cirrhosis and primary sclerosing cholangitis) also have an increased risk of acute tubular necrosis following surgery has not been well studied. An interesting clinical observation is that patients with primary biliary cirrhosis appear to be at decreased risk for developing hepatorenal syndrome after surgery compared to patients with other forms of liver disease [104]. A possible explanation is the natriuretic and renal vasodilator actions of retained bile salts.

Cardiac surgery — Cardiac surgery is associated with increased mortality in patients with cirrhosis compared to other surgical procedures.

• One of the largest series included 44 patients with cirrhosis of whom 12 developed hepatic decompensation and 7 died [68]. The Child's class was a significant predictor of decompensation and mortality. For mortality, a CPT score of >7 had sensitivity, specificity, positive, and negative predictive values of 86, 92, 67, and 97 percent, respectively. The authors concluded that cardiopulmonary bypass can be conducted safely in patients with a CPT score of <7. However, in other reports, major complications have been described in patients with lower scores [105].
• Another series included 27 patients with cirrhosis (10 with Child class A, 11 with Child class B and 6 with Child class C, with an overall mean MELD score of 14.2) [106]. Operative mortality was 11, 18, and 67 percent for Child class A, B, and C, respectively (26 percent overall), and the Child classification was a better predictor or mortality than the MELD score (using an arbitrary MELD classification of 0-12 and ≥13).
• A third report included 13 patients with predominantly alcoholic cirrhosis (eight with Child class A, and five with Child class B) who required coronary artery bypass grafting, valve replacement or both [24]. All patients who were Child class B experienced major complications and only one patient survived. All patients who were Child class A survived despite a complication rate of 25 percent. The high mortality rate in patients with Child class B was attributed to postoperative infections and bleeding, rather than cardiac dysfunction.

A number of risk factors for hepatic decompensation following cardiac surgery have been identified including the total time of cardiopulmonary bypass, use of nonpulsatile as opposed to pulsatile cardiopulmonary bypass, and need for perioperative pressor support [107]. Cardiopulmonary bypass can exacerbate underlying coagulopathy by inducing platelet dysfunction, fibrinolysis, and hypocalcemia [108].

Thus, the least invasive options, such as angioplasty, valvuloplasty, or minimally invasive revascularization techniques, should be considered in patients with advanced cirrhosis who require invasive intervention for cardiac disease [109]. (See "Minimally invasive coronary artery bypass graft surgery: Clinical efficacy of beating heart surgery".) Cardiac surgery followed by liver transplantation has been performed in rare instances [107,110]. Even more rarely, liver transplantation has been undertaken before cardiac surgery in patients with left ventricular dysfunction [108]. This approach is hazardous because of the risk of hemodynamic instability resulting from reduced venous return and reperfusion of the graft during liver transplantation [107].

Hepatic resection — Patients with cirrhosis undergoing resection for hepatocellular carcinoma or other liver tumors are at increased risk for hepatic decompensation compared to those undergoing other types of operations [111]. In addition to having severe underlying disease, a significant portion of functional hepatocellular mass may be removed in a setting in which patients already have severely compromised hepatic reserve. In the past, cirrhosis was considered to be a contraindication to resection of hepatic tumors since mortality rates exceeded 50 percent.

More recently, the perioperative mortality rate for hepatic resection has decreased to 3 to 16 percent, although postoperative morbidity rates are still as high as 60 percent [112-120]. The improvement in outcomes has been attributed to better patient selection (including earlier detection of tumors), meticulous preoperative preparation, intensive intra- and postoperative monitoring, and improved surgical techniques. Options for treating hepatocellular carcinoma including surgical resection are discussed elsewhere. (See "Surgical resection for hepatocellular carcinoma".)

Several systems for risk stratification of patients undergoing hepatic resection have been proposed, although none has been validated extensively. A database study of 587 patients who underwent hepatic resection concluded that the Child-Pugh Score and American Society of Anesthesiologists (ASA) physical status classification were better predictors of morbidity and mortality than the MELD score [121]. The ASA score was the only significant predictor of 30-day mortality (area under the receiver operating curve {ROC} of 0.63) while the ASA and Charlson Index of Comorbidity were the only significant predictors of morbidity (ROC of .56 and .40, respectively). However, the low ROC areas indicate that none of these models was an accurate predictor of outcomes. Moderate to severe hepatic steatosis (>30 percent of liver volume) is a risk factor for postoperative complications after major hepatectomy [122].

Trauma — Trauma patients found to have cirrhosis at laparotomy are at increased risk for morbidity and mortality. In one study, the overall mortality rate was 45 percent, significantly higher than of a matched control population (24 percent) [123]. Mortality and morbidity rates were increased even for patients considered to have relatively minor trauma. The authors recommended that trauma patients found to have cirrhosis at laparotomy be admitted to the intensive care unit for close monitoring and aggressive management irrespective of the severity of their injuries.

Patients with minimally increased risk — Patients with mild to moderate chronic liver disease without cirrhosis usually tolerate surgery well. However, medical therapy should be optimized prior to surgery.

Mild chronic hepatitis — Asymptomatic patients with mild chronic hepatitis are at low risk for complications [124]. In one report, for example, no major complications were noted during 34 surgical procedures in 24 patients with mild to moderate chronic hepatitis [124]. Two patients developed sustained hyperbilirubinemia, both of whom had preoperative bilirubin levels of 2.5 mg/dL (35.91 µmol/liter) or more.

Fatty liver and nonalcoholic steatohepatitis — Although the histologic appearance of nonalcoholic steatohepatitis (NASH) is similar to alcoholic hepatitis, patients with NASH do not appear to have excessive mortality following elective surgery. However, a trend toward increased mortality following hepatic resection has been observed in those with moderate to severe steatosis (>30 percent of hepatocytes containing fat) [125].

NASH is relatively common in patients with morbid obesity who undergo gastric bypass surgery. Cirrhosis, due presumably to NASH, has been found unexpectedly in up to 6 percent of such patients, in whom a perioperative mortality rate of 4 percent has been observed [48].

It may be difficult to distinguish NASH from alcoholic hepatitis since the histologic features can be identical, and patients do not always admit to alcohol ingestion. (See "Nonalcoholic steatohepatitis" and "Screening for and diagnosis of alcohol problems".) Thus, recommending a period of abstinence from alcohol prior to surgery is advisable for all patients with the histologic appearance of steatohepatitis, or those who are suspected of excessive alcohol consumption, since alcoholics are at increased risk for perioperative complications, such as alcohol withdrawal and hepatotoxicity with therapeutic doses of acetaminophen (often used for analgesia in the postoperative period) [126], even if they do not have liver disease. Furthermore, alcohol may potentiate the toxicity of halothane [127,128].

Autoimmune hepatitis — Elective surgery is usually well-tolerated in patients with autoimmune hepatitis who have compensated liver disease. Perioperative "stress" doses of hydrocortisone should be given to patients taking prednisone.

Hemochromatosis — Patients with hemochromatosis should be evaluated for complications such as diabetes and cardiomyopathy, which could influence perioperative care. (See "Clinical manifestations of hereditary hemochromatosis".) In the past, a relatively poor outcome of liver transplantation in these patients compared to other types of liver disease was attributed to underlying cardiomyopathy [129], but outcomes have improved with careful patient selection.

Wilson's disease — Patients with Wilson's disease who have neuropsychiatric involvement may not be able to provide informed consent. Furthermore, surgery can precipitate or aggravate neurologic symptoms. (See "Diagnosis of Wilson's disease".) D-penicillamine (a copper chelator commonly used for treatment), interferes with the crosslinking of collagen and may impair wound healing [130,131]. As a result, the dose should be decreased prior to surgery and during the first one to two postoperative weeks. (See "Treatment of Wilson's disease".)

OPTIMIZING MEDICAL THERAPY — In addition to assessing surgical risk, all patients with known liver disease should be assessed for the presence of jaundice, coagulopathy, ascites, electrolyte abnormalities, renal dysfunction and encephalopathy, all of which may require specific treatment prior to surgery. The basic principles involved in the evaluation of patients with specific forms of liver disease are discussed in detail separately. (See "Diagnostic approach to the patient with cirrhosis", and other appropriate topic reviews).

• In patients with an elevated prothrombin time, a reasonable goal is to attempt correction with vitamin K and fresh frozen plasma to achieve a prothrombin time within three seconds of normal prior to surgery. Experience is also accumulating with recombinant factor VIIA, which can temporarily correct the prothrombin time. (See "Gastroenterologic procedures in patients with disorders of hemostasis".)

• A prolonged bleeding time can be treated with diamino-8-D-arginine vasopressin (DDAVP). (See "Treatment of von Willebrand disease".)

Desmopressin).

• Optimal surgical technique and maintenance of low central venous pressure may reduce blood loss [132].
• Ascites should be treated aggressively to reduce the chance of wound dehiscence and abdominal wall herniation. This can be achieved safely with diuretics in patients who also have peripheral edema. In patients without edema or those in whom there is not enough time for a course of diuretic therapy, ascites can be drained completely during laparotomy. (See "Initial therapy of ascites in patients with cirrhosis".)

• Electrolyte abnormalities, particularly hypokalemia and metabolic alkalosis, should be corrected to reduce the chance of cardiac arrhythmias and hepatic encephalopathy.
• Conditions known to exacerbate hepatic encephalopathy should be corrected. (See "Pathogenesis of hepatic encephalopathy".) However, there is no evidence that prophylactic therapy can prevent encephalopathy after surgery.
• Renal function should be evaluated. For most patients, assessment of the blood urea nitrogen and creatinine is sufficient. However, these measures often overestimate renal function in patients with advanced liver disease because of the reduction in urea and creatinine synthesis. (See "Diagnosis and treatment of hepatorenal syndrome", section on 'Estimation of renal function'.)

• Patients with known gastroesophageal varices should receive the appropriate prophylactic treatment. (See "Primary prophylaxis against variceal hemorrhage in patients with cirrhosis".) Although surgery has not been associated with an increased risk of variceal bleeding, fluid overload should be avoided postoperatively.
• TIPS may be considered before surgery in patients with portal hypertension [133-136], but the role of preoperative TIPS has not been well-studied.
• Patients with cirrhosis are often malnourished. Perioperative nutritional support can reduce the frequency of postoperative complications and short-term mortality; its benefit on long-term survival is uncertain [137-141]. A reasonable approach is to provide total calories equal to 1.2 times the estimated resting energy expenditure and a 1 g/kg per day of protein. Approximately 30 to 35 percent of total energy should be given as fat and the remainder (typically 50 to 55 percent) as carbohydrates. Supplementation of the fat soluble vitamins A, D, E, and K, may also be necessary [140].

Percutaneous gastrostomy (PEG) is contraindicated in patients with ascites, and should usually be avoided in patients with portal hypertension due to the possibility of lacerating an abdominal wall varix during PEG insertion.

Following surgery, patients with liver disease should be observed closely for hepatic decompensation, which often presents with worsening jaundice, encephalopathy, and ascites. The best biochemical measures of liver function are probably the prothrombin time and serum bilirubin concentration. However, the serum bilirubin concentration usually rises, particularly after complicated surgery, multiple blood transfusions, excessive bleeding, hemodynamic instability, or systemic infection. Renal function, serum electrolytes, and glucose should also be monitored carefully.

SUMMARY AND RECOMMENDATIONS — Considering the above data and clinical experience, guidelines for assessing the risk of elective or semi-urgent surgery in patients with liver disease can be suggested:

• Medical therapy should be optimized in all patients as described above. (See 'Optimizing medical therapy' above.)

• Operative mortality can be estimated based upon the Child classification and the MELD score taking into consideration other factors such as the patient's age, ASA score, and additional comorbidities.
• We recommend elective or semi-urgent surgery not be performed in patients with acute or fulminant hepatitis, alcoholic hepatitis, severe chronic hepatitis, Child class C or MELD >15 cirrhosis, severe coagulopathy, or severe extrahepatic manifestations of liver disease (such as hypoxia, cardiomyopathy, or acute renal failure) (Grade 1B).

For other patients, the risk of surgery should be considered individually depending upon the clinical setting and the type of procedure:

• It is well-tolerated in patients with Child's class A or MELD <10 cirrhosis and those with mild chronic liver disease without cirrhosis.
• It is permissible in patients with Child's class B or MELD 10-15 cirrhosis (except those undergoing extensive hepatic resection or cardiac surgery) who have undergone thorough preoperative preparation.

Consideration should also be given as to whether surgery can be deferred until after liver transplantation (either orthotopic or live donor) in appropriate candidates.