摘要
This guideline has been approved by the AASLD and represents the position of the Association. These recommendations provide a data-supported approach. They are based on the following: (1) formal review and analysis of the recently-published world literature on the topic (Medline search); (2) American College of Physicians Manual for Assessing Health Practices and Designing Practice Guidelines1; (3) guideline policies, including the AASLD Policy on the Development and Use of Practice Guidelines and the American Gastroenterological Association Policy Statement on Guidelines2; and (4) the author's decades of experience caring for patients with cirrhosis and ascites. Intended for use by physicians, these recommendations suggest preferred approaches to the diagnostic, therapeutic, and preventive aspects of care. They are intended to be flexible, in contrast to standards of care, which are inflexible policies to be followed in every case. Specific recommendations are based on relevant published information. To more fully characterize the quality of evidence supporting recommendations, the Practice Guidelines Committee of the AASLD requires a Class (reflecting benefit versus risk) and Level (assessing strength or certainty) of Evidence to be assigned and reported with each recommendation (Table 1, adapted from the American College of Cardiology and the American Heart Association Practice Guidelines3).4 These guidelines were developed for the care of adult patients with clinically detectable ascites. Although the general approach may be applicable to children, the pediatric database is much smaller and there may be unanticipated differences between adults and children. Patients with ascites detected only by imaging modalities but not yet clinically evident are excluded because of the lack of published information regarding the natural history of this entity. A Medline search from 1966 through 2007 was performed; search terms included ascites, hepatorenal syndrome, diet therapy, drug therapy, radiotherapy, surgery, and therapy. The search involved only articles published in English and involving humans. A manual search of the author's files and recent abstracts was also performed. The search yielded 2115 articles including 153 published since a similar search was performed in 2002 in preparation for writing the previous guideline on ascites. AASLD, American Association for the Study of Liver Diseases; LDH, lactate dehydrogenase; PMN, polymorphonuclear leukocyte; SAAG, serum-ascites albumin gradient; SBP, spontaneous bacterial peritonitis; TIPS, transjugular intrahepatic portasystemic stent-shunt. Cirrhosis was the twelfth leading cause of death in the United States, according to a 2006 Vital Statistics Report in which data were collected through 2004.5 Ascites is the most common of the three major complications of cirrhosis; the other complications are hepatic encephalopathy and variceal hemorrhage.6 Approximately 50% of patients with “compensated” cirrhosis, i.e., without having developed one of these complications, develop ascites during 10 years of observation.6 Ascites is the most common complication of cirrhosis that leads to hospital admission.7 The pathophysiology of ascites and hepatorenal syndrome have been reviewed elsewhere.8 Development of fluid retention in the setting of cirrhosis is an important landmark in the natural history of chronic liver disease: approximately 15% of patients with ascites succumb in 1 year and 44% succumb in 5 years.9 Many patients are referred for liver transplantation after development of ascites. Most patients (approximately 85%) with ascites in the United States have cirrhosis (Table 2).10 In about 15% of patients with ascites, there is a nonhepatic cause of fluid retention. Successful treatment is dependent on an accurate diagnosis of the cause of ascites; e.g., peritoneal carcinomatosis does not respond to diuretic therapy. Patients with ascites should be questioned about risk factors for liver disease. Those who lack an apparent cause for cirrhosis should also be questioned about lifetime body weight; nonalcoholic steatohepatitis has been concluded to be causative in many of these patients.11 Past history of cancer, heart failure, renal disease, or tuberculosis is also relevant. Hemophagocytic syndrome can masquerade as cirrhosis with ascites.12 These patients have fever, jaundice, and hepatosplenomegaly, usually in the setting of lymphoma or leukemia.12 The presence of a full, bulging abdomen should lead to percussion of the flanks. If the amount of flank dullness is greater than usual (i.e., if the percussed tympany-dullness interface is higher than normally found on the lateral aspect of the abdomen with the patient supine), one should test for “shifting”. The presence of shifting dullness has 83% sensitivity and 56% specificity in detecting ascites.13 Approximately 1500 mL of fluid must be present before flank dullness is detected.13 If no flank dullness is present, the patient has less than a 10% chance of having ascites.13 The fluid wave and puddle sign are cumbersome and perform less well when compared to shifting dullness.13 Ascites due to alcoholic cardiomyopathy can mimic that due to alcoholic cirrhosis. Jugular venous distension is present in the former but not in the latter. Also measurement of a blood concentration of brain natriuretic peptide or pro–brain natriuretic peptide can help distinguish ascites due to heart failure from ascites due to cirrhosis.14 The median pro–brain natriuretic peptide concentration is 6100 pg/mL in the former and only 166 pg/mL in the latter.14 Giant cysts or pseudocysts can rarely mimic ascites. Paracentesis may produce fluid with unusual characteristics. Imaging usually provides the correct diagnosis.15 The physical examination for detecting ascites in the obese patient is problematic. An abdominal ultrasound may be required to determine with certainty if fluid is present. Ascites usually is present for only a few weeks before the patient seeks medical attention. In contrast, a slowly enlarging abdomen over months to years is most likely due to obesity not ascites. The diagnosis of new-onset ascites is suspected on the basis of the history and physical examination and usually confirmed by successful abdominal paracentesis and/or ultrasound. The diagnosis of the etiology of ascites formation is based on the results of the history, physical examination, and ascitic fluid analysis. In general, few other tests are required. However, the liver is commonly imaged to screen for hepatocellular carcinoma, portal vein thrombosis, and hepatic vein thrombosis. Abdominal paracentesis with appropriate ascitic fluid analysis is probably the most rapid and cost-effective method of diagnosing the cause of ascites.16, 17 Fluid due to portal hypertension can be readily differentiated from fluid due to other causes.10 Also, in view of the high prevalence of ascitic fluid infection at the time of admission to the hospital, an admission surveillance tap may detect unexpected infection.18 Although older published series reported a relatively high morbidity, and even mortality, when trocars were used for paracentesis, more recent studies regarding paracentesis complications in patients with ascites documented no deaths or infections caused by the paracentesis.19 Complications were reported in only about 1% of patients (abdominal wall hematomas), despite the fact that 71% of the patients had an abnormal prothrombin time.19 Although more serious complications (hemoperitoneum or bowel entry by the paracentesis needle) occur,20 they are sufficiently unusual (<1/1000 paracenteses) that they should not deter performance of this procedure. In a study of 4729 paracenteses, investigators reported that eight of nine bleeding complications occurred in patients with renal failure; perhaps the qualitative platelet abnormality in this setting predisposes to more bleeding.21 Although some physicians give blood products (fresh frozen plasma and/or platelets) routinely before paracentesis in patients with cirrhosis and coagulopathy, this policy is not data-supported.19, 22 Routine tests of coagulation also do not reflect bleeding risk in patients with cirrhosis; these patients regularly have normal global coagulation because of a balanced deficiency of procoagulants and anticoagulants.23 In a recent survey of the use of blood products in relation to paracentesis, 50% of approximately 100 hepatologists attending a conference on coagulopathy in liver disease indicated that they either never used plasma before procedure or used it only if the international normalized ratio was >2.5.24 The risks and costs of prophylactic transfusions may exceed the benefit. Coagulopathy should preclude paracentesis only when there is clinically evident hyperfibrinolysis (three-dimensional ecchymosis/hematoma) or clinically evident disseminated intravascular coagulation. A shortened (<120 minutes) euglobulin clot lysis time documents hyperfibrinolysis.25 However, this test may not be routinely available. Epsilon aminocaproic acid can be used to treat hyperfibrinolysis; paracentesis can be performed after the lysis time has normalized on treatment.26 Bleeding conditions occur in less than 1/1000 patients who require paracentesis. There is no data-supported cutoff of coagulation parameters beyond which paracentesis should be avoided.19 In a study of 1100 large-volume paracenteses, there were no hemorrhagic complications despite (1) no prophylactic transfusions, (2) platelet counts as low as 19,000 cells/mm3 (19 × 106/L) (54% <50,000), and (3) international normalized ratios for prothrombin time as high as 8.7 (75% >1.5 and 26.5% >2.0).22 In the past, the avascular midline, midway between the pubis and the umbilicus, was usually chosen as the site for paracentesis. Now, because many paracenteses are performed to remove a large volume of fluid and abdominal obesity increases the midline wall thickness, the left lower quadrant is the preferred location (Fig. 1). The abdominal wall in the left lower quadrant, 2 finger breadths (3 cm) cephalad and 2 finger breadths medial to the anterior superior iliac spine, has been shown to be thinner and with a larger pool of fluid than the midline and is usually a good choice for needle insertion for performance of a therapeutic paracentesis.27 The right lower quadrant may be a suboptimal choice in the setting of a dilated cecum (due to lactulose) or an appendectomy scar. The area of the inferior epigastric arteries should be avoided; these vessels are located midway between the pubis and anterior superior iliac spines and then run cephalad in the rectus sheath. Visible collaterals should also be avoided. A laparoscopic study found that collaterals can be present in the midline and thus present a risk for rupture during paracentesis.28 Diagram of the abdomen showing the three usual sites for abdominal paracentesis. The author prefers the left lower quadrant site. Reproduced from Thomsen TW, Shaffer RW, White B, Setnik GS. Paracentesis. N Engl J Med 2006;355:e21, with permission from the Massachusetts Medical Society. Copyright (2006) Massachusetts Medical Society. All rights reserved. If the fluid is difficult to localize by examination because of obesity, ultrasonography can be a useful adjunct in locating fluid and visualizing the spleen and other structures to be avoided. There are few contraindications to paracentesis. The procedure should be performed by a provider who has been trained in its performance. Recommendations: 1. Abdominal paracentesis should be performed and ascitic fluid should be obtained from inpatients and outpatients with clinically apparent new-onset ascites. (Class I, Level C) 2. Because bleeding is sufficiently uncommon, the routine prophylactic use of fresh frozen plasma or platelets before paracentesis is not recommended. (Class III, Level C) An algorithm approach seems preferable to ordering a large number of tests on most specimens (Table 3). If uncomplicated ascites due to cirrhosis is suspected, only screening tests (e.g., cell count and differential, albumin and total protein concentration) are performed on the initial specimen. If the results of these tests are unexpectedly abnormal, further testing can be performed on another ascitic fluid sample. Also, many laboratories save an aliquot of fluid for a few days; this fluid can be tested if the specimen has been handled properly. However, because most specimens are consistent with uncomplicated cirrhotic ascites, no further testing will be needed in the majority of patients. If ascitic fluid infection is suspected (fever, abdominal pain, unexplained encephalopathy, acidosis, azotemia, hypotension, or hypothermia), bacterial culture of the fluid in blood culture bottles inoculated at the bedside should be performed. Use of a urine dipstick to detect neutrophils in ascitic fluid takes only 90 seconds to 2 minutes.29, 30 However, the largest study of a urine dipstick (2123 paracenteses) demonstrated a sensitivity of only 45%.31 Development of an ascites-specific dipstick in contrast to a urine dipstick is needed. Automated cell counting has been shown to be accurate in one study; the result is rapidly available and could replace the manual cell count if it is further validated.32 Additional testing, e.g., for total protein, lactate dehydrogenase (LDH), and glucose to assist in differentiating spontaneous from secondary bacterial peritonitis, can be performed on the initial specimen based on clinical judgment.33 An ascitic fluid carcinoembryonic antigen >5 ng/mL or ascitic fluid alkaline phosphatase >240 U/L has also been shown to be accurate in detecting gut perforation into ascitic fluid.34 The serum-ascites albumin gradient (SAAG) has been proved in prospective studies to categorize ascites better than the total-protein–based exudate/transudate concept and better than modified pleural fluid exudate/transudate criteria.11, 35 Calculating the SAAG involves measuring the albumin concentration of serum and ascitic fluid specimens obtained on the same day and subtracting the ascitic fluid value from the serum value. If the SAAG is ≥1.1 g/dL (11 g/L), the patient has portal hypertension, with approximately 97% accuracy.11 Patients who have portal hypertension plus a second cause for ascites formation also have a SAAG ≥1.1 g/dL. Patients undergoing serial outpatient therapeutic paracenteses probably should be tested only for cell count and differential36, 37 (the author has detected eight episodes of spontaneous bacterial peritonitis [SBP] in approximately 400 paracenteses in a paracentesis clinic in 2 years [unpublished observations]). Bacterial culture is not necessary in asymptomatic patients undergoing serial large-volume paracenteses. The most expensive tests are the cytology and smear and culture for mycobacteria; these tests should probably be ordered only when there is a high pretest probability of occurrence of the disease under consideration. The ascitic fluid cytology is positive only in the setting of peritoneal carcinomatosis.38 The sensitivity of cytology in detecting peritoneal carcinomatosis is 96.7% if three samples are sent and processed promptly; the first sample is positive in 82.8% and at least one of two samples is positive in 93.3%.38 In this study, 50 mL of fresh warm ascitic fluid were hand-carried to the laboratory for immediate processing. Use of DNA cytometry or magnetic enrichment may improve the sensitivity of cytology further.39, 40 Patients with peritoneal carcinomatosis usually have a history of a breast, colon, gastric, or pancreatic primary carcinoma. The sensitivity of smear for mycobacteria is approximately 0%; the sensitivity of fluid culture for mycobacteria is approximately 50%.41 Only patients at high risk for tuberculous peritonitis (e.g., recent immigration from an endemic area or acquired immunodeficiency syndrome)42 should have testing for mycobacteria on the first ascitic fluid specimen. Laparoscopy with biopsy and mycobacterial culture of tubercles are the most rapid and accurate methods of diagnosing tuberculous peritonitis. Multiple prospective trials have shown that bacterial growth occurs in only about 50% of instances when ascitic fluid with a polymorphonuclear leukocyte (PMN) count ≥250 cells/mm3 (0.25 × 109/L) is cultured by older methods, i.e., sending a syringe or tube of fluid to the laboratory, as compared to approximately 80% if the fluid is inoculated into blood culture bottles at the bedside and prior to administration of antibiotics.43, 44 Although cirrhosis is the cause of ascites formation in most patients, approximately 15% have a cause other than liver disease, including cancer, heart failure, tuberculosis, or nephrotic syndrome (Table 3).10 Approximately 5% of patients with ascites have two or more causes of ascites formation, i.e., “mixed” ascites.10 Usually, these patients have cirrhosis plus one other cause, e.g., peritoneal carcinomatosis or peritoneal tuberculosis. Many patients with enigmatic ascites are eventually found to have two or even three causes for ascites formation (e.g., heart failure, diabetic nephropathy, and cirrhosis due to nonalcoholic steatohepatitis). In this setting, the sum of predisposing factors leads to sodium and water retention when each individual factor might not be severe enough to cause fluid overload. The cancer antigen 125 (CA125) warrants mention. Essentially all patients, including men with ascites or pleural fluid of any cause, have an elevated serum CA125; when ascites is controlled, the CA125 level decreases dramatically.45, 46 This test is elevated when mesothelial cells are under pressure from the presence of fluid; it is very nonspecific. When this test is found to be abnormal, the female patient may be unnecessarily referred for gynecologic surgery even if the ovaries were removed decades earlier; cirrhosis is regularly detected at laparotomy as the cause for ascites formation (because it is most common cause) rather than ovarian cancer, and the patient may die postoperatively. Patients with ascites should not have serum tested for CA125. Recommendations: 3. The initial laboratory investigation of ascitic fluid should include an ascitic fluid cell count and differential, ascitic fluid total protein, and SAAG. (Class I, Level B) 4. If ascitic fluid infection is suspected, ascitic fluid should be cultured at the bedside in blood culture bottles prior to initiation of antibiotics. (Class I, Level B) 5. Other studies of ascitic fluid can be ordered based on pretest probability of disease (Table 3). (Class IIa, Level C) 6. Testing serum for CA125 is not helpful in the differential diagnosis of ascites. Its use is not recommended in patients with ascites of any type. (Class III, Level B) Appropriate treatment of patients with ascites depends on the cause of fluid retention. SAAG can be helpful diagnostically as well as in decision-making regarding treatment. Patients with low SAAG (<1.1 g/dL) ascites usually do not have portal hypertension and, with the exception of nephrotic syndrome, do not respond to salt restriction and diuretics.17 In contrast, patients with a high SAAG (≥1.1 g/dL) have portal hypertension and usually are responsive to these measures.17 The remainder of this guideline is applicable only to patients with cirrhosis as the cause of their ascites. Improvement in the outcome of patients with nonportal-hypertension–related ascites depends on successful treatment of the underlying disorder. Alcohol-induced liver injury is one of the most reversible causes of liver disease that leads to high SAAG ascites.17 One of the most important steps in treating ascites in this setting is to treat the underlying liver disease by ceasing alcohol consumption. In a period of months, abstinence can result in dramatic improvement in the reversible component of alcoholic liver disease. One recent study demonstrates that patients who have Child-Pugh class C cirrhosis due to alcohol and who stop drinking have an approximately 75% 3-year survival, but all those who continue to drink die in 3 years.47 Ascites may resolve or become more responsive to medical therapy with abstinence and time. Decompensated hepatitis B cirrhosis can also have a dramatic response to antiviral treatment.48 Liver diseases other than those that are related to alcohol, hepatitis B, and autoimmune hepatitis are less reversible; by the time ascites is present, these patients may be best served by referral for liver transplantation evaluation rather than protracted medical therapy. The mainstays of treatment of patients with cirrhosis and ascites include (1) education regarding dietary sodium restriction (2000 mg/day [88 mmol/day]) and (2) oral diuretics.16, 17 More stringent dietary sodium restriction can speed mobilization of ascites, but is not recommended because it is less palatable and may further worsen the malnutrition that is usually present in these patients. Fluid loss and weight change are directly related to sodium balance in patients with portal hypertension–related ascites. It is sodium restriction, not fluid restriction, which results in weight loss, as fluid follows sodium passively.49 Measurement of urinary sodium excretion is a helpful parameter to follow when rapidity of weight loss is less than desired.16, 17 Random urinary sodium concentrations are of value when they are 0 mmol/L or >100 mmol/L but are much less helpful when they are intermediate because of lack of uniformity of sodium excretion during the day and lack of knowledge of total urine volume, which may vary from 300 mL to greater than 3000 mL. Twenty-four-hour collections of urine for determination of sodium excretion are much more informative than random specimens; however, full-day collections are cumbersome. Providing patients with verbal and written instructions, a container, and a lab order slip to turn in with the completed specimen helps insure compliance. Completeness of collection of the 24-hour specimen can be assessed by measurement of urinary creatinine. Men with cirrhosis should excrete more than 15 mg creatinine per kilogram of body weight per day, and women with cirrhosis should excrete more than 10 mg/kg/day. Less creatinine is indicative of an incomplete collection. Total nonurinary sodium excretion is less than 10 mmol/day in afebrile patients with cirrhosis without diarrhea.50 One of the goals of treatment is to increase urinary excretion of sodium so that it exceeds 78 mmol/day (88 mmol intake/day − 10 mmol nonurinary excretion per day). Only the 10%–15% of patients who have spontaneous natriuresis >78 mmol/day can be considered for dietary sodium restriction alone (i.e., without diuretics). However, when given a choice, most patients would prefer to take some diuretics and have a more liberal sodium intake than take no pills and have a more severe sodium restriction. A random “spot” urine sodium concentration that is greater than the potassium concentration correlates with a 24-hour sodium excretion greater than 78 mmol/day with approximately 90% accuracy.51 This urine sodium/potassium ratio may replace the cumbersome 24-hour collection. Fluid restriction is not necessary in treating most patients with cirrhosis and ascites. The chronic hyponatremia usually seen in patients with cirrhosis and ascites is seldom morbid unless it is rapidly corrected in the operating room at the time of liver transplantation.52 A study of 997 patients with cirrhosis and ascites demonstrates that the serum sodium is ≤120 mmol/L in only 1.2% of patients and ≤125 mmol/L in only 5.7%.53 Attempts to rapidly correct hyponatremia in this setting with hypertonic saline can lead to more complications than the hyponatremia itself.54 Preliminary data suggest that aquaretic drugs have the promise of correcting hyponatremia. The intravenous aquaretic agent conivaptan has been studied in patients with cirrhosis and is approved for use for treatment of “euvolemic and hypervolemic hyponatremia in hospitalized patients”.55 Caution is advised by the manufacturer in the use of this drug in patients with cirrhosis. An oral preparation tolvaptan increases serum sodium in patients who have pretreatment values <130 mmol/L.56 However, whether these agents will be effective without side effects in the subset of patients with cirrhosis who are more in need of correction of hyponatremia (serum sodium ≤120 mmol/L) remains unproven. Cost-effectiveness also warrants investigation. Unfortunately, many drugs that have theoretical promise in treating ascites, e.g., angiotensin-converting enzyme inhibitors, have been shown to aggravate hypotension and have not been clinically useful. Severe hyponatremia does warrant fluid restriction in the patient with cirrhosis and ascites; however, there is no data-supported specific threshold for initiating fluid restriction. A serum sodium <120–125 mmol/L is a reasonable threshold. Patients with cirrhosis do not usually have symptoms from hyponatremia until the sodium is <110 mmol/L or unless the decline in sodium is very rapid. Although it is traditional to recommend bed rest (based on extrapolation from heart failure), this is impractical and there are no controlled trials to support this practice. Upright posture may aggravate the plasma renin elevation found in patients with cirrhosis with ascites. Theoretically, this may increase sodium avidity. This theoretical concern would have to translate into clinically relevant outcomes before bed rest could be advocated. The usual diuretic regimen consists of single morning doses of oral spironolactone and furosemide, beginning with 100 mg of the former and 40 mg of the latter.16, 17 Previously, single-agent spironolactone was advocated, but hyperkalemia and the long half-life of this drug have resulted in its use as a single agent only in patients with minimal fluid overload.57 Single-agent furosemide has been shown in a randomized controlled trial to be less efficacious than spironolactone.58 The good oral bioavailability of furosemide in the patient with cirrhosis, together with the acute reductions in glomerular filtration rate associated with intravenous furosemide, favor use of the oral route of administration.59, 60 A randomized trial purports to demonstrate that spironolactone should be used as a single agent, with furosemide added only for refractory patients.61 Diuresis was slower in the single-agent spironolactone group with a lesser need for dose adjustments; thus, this approach may be useful for outpatients.61 However, another randomized trial indicates that initial combination treatment shortens the time to mobilization of moderate ascites.62 Most patients require combination treatment eventually. The largest study ever performed (involving 3860 patients with cirrhosis and ascites) used combination therapy from the beginning.63 Starting with both drugs appears to be the preferred approach in achieving rapid natriuresis and maintaining normokalemia. An alternative approach would be to start with single-agent spironolactone, in particular in the outpatient setting. The doses of both oral diuretics can be increased simultaneously every 3–5 days (maintaining the 100 mg:40 mg ratio) if weight loss and natriuresis are inadequate. In general, this ratio maintains normokalemia. Usual maximum doses are 400 mg/day of spironolactone and 160 mg/day of furosemide.16, 17 Furosemide can be temporarily withheld in patients presenting with hypokalemia; this is very common in the setting of alcoholic hepatitis. Patients with parenchymal renal disease (e.g., diabetic nephropathy or immunoglobulin A nephropathy or those having undergone liver transplantation) may tolerate less spironolactone than usual because of hyperkalemia. Single morning dosing maximizes compliance. Amiloride (10–40 mg/day) can be substituted for spironolactone in patients with tender gynecomastia. However, amiloride is more expensive and has been shown to be less effective than an active metabolite of spironolactone in a randomized controlled trial.64 Triamterene, metolazone, and hydrochlorothiazide have also been used to treat ascites.65-67 Hydrochlorothiazide can also cause rapid development of hyponatremia when added to the combination of spironolactone and furosemide.67 Eplenerone is a new aldosterone antagonist that has been used in heart failure.68 It has not been studied in the setting of cirrhosis and ascites. Newer loop diuretics must be proven to be superior to current drugs before their expense can be justified. Although an intravenous dose of 80 mg furosemide can cause an acute reduction in renal perfusion and subsequent azotemia in patients with cirrhosis and ascites, this same dose has been shown in one study to separate diuretic-resistant (<50 mmol urine sodium in 8 hours) from diuretic-sensitive patients (>50 mmol).69 Another study has confirmed this observation.70 This intravenous furosemide “test” may help speed detection of diuretic-resistant patients so that they can more rapidly be given second-line treatment options.69 However, intravenous furosemide can cause azotemia (see below), and its repeated use should probably be minimized until its safety and efficacy are evaluated in randomized trials. In the largest, multicenter, randomized controlled trial performed in patients with ascites, dietary sodium restriction and a dual diuretic regimen with spironolactone and furosemide has been shown to be effective in more than 90% of patients in achieving a reduction in the volume of ascites to acceptable levels.63 An unblinded randomized controlled trial in patients with new-onset ascites demonstrates that weekly 25 g infusions of albumin for 1 year followed by infusions every 2 weeks improved survival compared to diuretics alone.71 However, further studies including cost-effectiveness analysis in the Unit