Ascites and Spontaneous Bacterial Peritonitis

Noncirrhotic Ascites

The mechanism of fluid retention in patients with malignancy-related ascites depends on the location of the tumor. Peritoneal carcinomatosis appears to cause ascites through the production of proteinaceous fluid by tumor cells lining the peritoneum. Extracellular fluid enters the peritoneal cavity to reestablish oncotic balance. Fluid accumulates in patients with massive liver metastases because of portal hypertension caused by stenosis or occlusion of portal veins by tumor nodules or tumor emboli.² In patients with hepatocellular carcinoma, ascites arises because of the underlying cirrhosis-related portal hypertension, tumor-induced portal vein thrombosis, or both. Chylous ascites in patients with malignant lymphoma appears to be caused by lymph node obstruction by tumor and rupture of chyle-containing lymphatics.

Ascites can complicate high-output or low-output heart failure or nephrotic syndrome. As in cirrhosis, effective arterial blood volume appears to be decreased, and the vasopressin, renin-aldosterone, and sympathetic nervous systems are activated.³ These changes lead to renal vasoconstriction and sodium and water retention. Fluid then “weeps” from the congested hepatic sinusoids as lymph, as in cirrhosis. TB, Chlamydia infection, and coccidioidomycosis probably cause ascites through the production of proteinaceous fluid, as in peritoneal carcinomatosis. SBP does not appear to cause fluid to accumulate; infection develops only in preexisting ascites.

In patients with pancreatic or biliary ascites, fluid accumulates by leakage of pancreatic juice or bile into the peritoneal cavity or forms secondary to a “chemical burn” of the peritoneum. After abdominal surgery, especially extensive retroperitoneal dissection, lymphatics may be transected and may leak lymph for varying amounts of time. The mechanism of development of ascites in this condition is similar to that for malignant chylous ascites, namely, lymphatic leak.

Physical Examination

On the basis of the history and the appearance of the abdomen, the diagnosis of ascites is readily suspected and usually confirmed easily on physical examination. The presence of a full, bulging abdomen should lead to percussion of the flanks. If the degree of flank dullness is greater than usual (i.e., if the percussed air-fluid level is higher than that normally found on the lateral aspect of the abdomen with the patient supine), the examiner should check for “shifting.” If flank dullness is absent, checking for shifting is unnecessary. Approximately 1500 mL of fluid must be present before dullness is detected.⁸ If flank dullness is not present, the chance that the patient has ascites is less than 10%.⁸ A fluid wave is not worth testing for.⁸

Gaseous distention of the bowel, a thick panniculus, and an ovarian mass can mimic ascites. Gaseous distention should be readily apparent on percussion. Ovarian masses usually cause tympanitic flanks with central dullness. The speed of increase in abdominal girth also can be helpful: ascites develops in days to weeks, whereas thickening of the omentum and panniculus takes months to years. An obese abdomen may be diffusely dull to percussion, and abdominal US may be required to determine if fluid is present. US can detect as little as 100 mL of fluid in the abdomen.

The presence of palmar erythema, large pulsatile spider telangiectasias, large abdominal wall collateral veins, or fetor hepaticus is suggestive of parenchymal liver disease and portal hypertension (see Chapter 74). The presence of large veins on the patient's back suggests inferior vena cava blockage. An immobile mass in the umbilicus, the Sister Mary Joseph nodule, is suggestive of peritoneal carcinomatosis.

The neck veins of patients with ascites should always be examined. Alcoholic cardiomyopathy with cardiac ascites can mimic cirrhosis with ascites; an elevated jugular venous pressure helps with this aspect of the differential diagnosis. Constrictive pericarditis is one of the few curable causes of ascites. Most patients with cardiac ascites have impressive jugular venous distention. Some have no visible jugular venous distention but such high central venous pressures that their bulging forehead veins rise to the top of their skulls. When present, peripheral edema in patients with liver disease is usually found in the lower extremities and occasionally may involve the abdominal wall. Patients with nephrotic syndrome or cardiac failure may have total body edema (anasarca).

Abdominal paracentesis⁹ with appropriate ascitic fluid analysis is probably the most rapid and cost-effective method of diagnosing the cause of ascites. Also, because of the possibility of ascitic fluid infection in a patient with cirrhosis admitted to the hospital, a surveillance paracentesis performed on admission may detect unexpected infection.¹⁰ Not all patients with ascitic fluid infection are symptomatic; many have subtle symptoms, such as mild confusion noticed only by the family. An emergency department study has demonstrated that a physician's impression regarding the presence or absence of ascitic fluid infection is inaccurate.¹¹

Detection of infection at an early asymptomatic stage may reduce mortality. Therefore, ascitic fluid should be sampled in all inpatients and outpatients with new-onset ascites and in all patients with ascites who are admitted to the hospital. Paracentesis should be repeated in patients (whether hospitalized or not) in whom symptoms, signs, or laboratory abnormalities suggestive of infection develop (e.g., abdominal pain or tenderness, fever, encephalopathy, hypotension, renal failure, acidosis, peripheral leukocytosis).

Non-neutrocytic (i.e., ascitic fluid polymorphonuclear neutrophil [PMN] count less than 250/mm³ [0.25 × 10⁹/L]) ascitic fluid is transparent and usually slightly yellow (Fig. 93-2). Ascitic fluid with a very low protein concentration may have no pigment and look like water. The opacity of many cloudy ascitic fluid specimens is caused by neutrophils. The presence of neutrophils leads to a shimmering effect when a glass tube containing the fluid is rocked back and forth in front of a light. Fluid with an absolute neutrophil count less than 1000/mm³ (1.0 × 10⁹/L) may be nearly clear. Fluid with a count greater than 5000/mm³ (5.0 × 10⁹/L) is quite cloudy, and fluid with a count greater than 50,000/mm³ (50.0 × 10⁹/L) resembles mayonnaise.

Ascitic fluid specimens frequently are blood-tinged or frankly bloody. An RBC count of 10,000/mm³ (10.0 × 10⁹/L) is the threshold for a pink appearance; lower concentrations result in clear or turbid fluid. Ascitic fluid with an RBC count greater than 20,000/mm³ (20.0 × 10⁹/L) is distinctly red. Many ascitic fluid specimens are bloody because of a traumatic tap; these specimens are blood-streaked and frequently clot unless the fluid is transferred immediately to the anticoagulant-containing tube for the cell count. By contrast, nontraumatic or remotely traumatic blood-tinged ascitic fluid is homogeneous and does not clot because it has already clotted and the clot has lysed. Some patients with portal hypertension have bloody hepatic lymph, resulting in bloody ascitic fluid—perhaps because of rupture of lymphatics that are under high pressure. Samples from patients with hepatocellular carcinoma are regularly bloody, but only about 10% of samples from patients with peritoneal carcinomatosis are red.² Although many physicians have the impression that TB results in bloody ascitic fluid, less than 5% of tuberculous samples are hemorrhagic in my experience.

Ascitic fluid frequently is lipid-laden. Lipid opacifies the fluid. The degree of opalescence of ascitic fluid ranges from slightly cloudy to completely opaque and chylous. Most opaque, milky fluid samples have a TG concentration greater than 200 mg/dL (2.26 mmol/L) and usually greater than 1000 mg/dL (11.30 mmol/L). Fluid that has the appearance of dilute skim milk has a TG concentration between 100 mg/dL (1.13 mmol/L) and 200 mg/dL (2.26 mmol/L). A substantial minority of cirrhotic ascitic fluid samples are neither transparent nor frankly milky. These opalescent samples have slightly elevated TG concentrations ranging from 50 mg/dL (0.56 mmol/L) to 200 mg/dL (2.26 mmol/L).¹⁶ The opacity of these fluids does not have the shimmering characteristics of ascitic fluid with an elevated WBC count. The lipid usually layers out when a tube of ascitic fluid is placed in a refrigerator for 48 to 72 hours. In contrast to findings in older published reports, most patients with chylous or opalescent ascites have cirrhosis.^16,17

Dark-brown fluid with a bilirubin concentration greater than that of serum usually indicates biliary perforation.¹⁸ Deeply jaundiced patients have bile-stained ascitic fluid, but the bilirubin level and the degree of pigmentation are visually less than those of the corresponding serum. Pancreatic ascites may be pigmented because of the effect of pancreatic enzymes on RBCs. The RBCs may have to be centrifuged before the discolored supernatant is revealed. The degree of pigmentation ranges from tea-colored to jet black, as in pancreatic necrosis. Black ascitic fluid also may be found in patients with malignant melanoma.

Tests

The practice of ordering every available body fluid test on every ascitic fluid specimen is expensive and can be more confusing than helpful, especially when unexpectedly abnormal results are encountered. An algorithm for the analysis of ascitic fluid is shown in Figure 93-2. The basic concept is that screening tests are performed on the initial specimen; additional testing is performed only when necessary as indicated by the results of the screening tests. Further testing may require another paracentesis, but because most specimens consist of ascitic fluid resulting from uncomplicated cirrhosis, no further testing is needed in a majority of cases. Also, because laboratories frequently store the fluid for a few days, additional testing can often be ordered on the stored fluid.

On the basis of cost analysis, tests can be classified as routine, optional, unusual, and unhelpful (Table 93-2).¹⁰ The cell count is the single most helpful ascitic fluid test. Only approximately 10 µL of fluid is required for a standard manual hemocytometer count. Therefore, if only 1 drop of fluid can be obtained, it should be sent for cell count. More fluid is almost always obtainable, however. The fluid should be submitted in an anticoagulant-containing tube (i.e., EDTA) to prevent clotting. Because the decision to begin empirical antibiotic treatment of suspected ascitic fluid infection is based largely on the absolute neutrophil count (which should have a turnaround time of a few minutes) rather than the culture (which takes 12 to 48 hours to demonstrate growth), the cell count is more important than the culture in the early detection and treatment of ascitic fluid infection.

TABLE 93-2

Ascitic Fluid Laboratory Tests

Routine	Optional	Unusual	Unhelpful
Cell count Albumin Total protein	Amylase Culture in blood culture bottles Glucose Gram stain LDH	Bilirubin Cytology TB smear, culture, and PCR test TG	Cholesterol Fibronectin Lactate pH

Even samples from asymptomatic outpatients undergoing therapeutic paracentesis should be sent for a cell and differential count; the information obtained can lead to early, life-saving treatment of bacterial infection.

Cell Count

Surprisingly, ascitic fluid cell counts have not been standardized. Some laboratories count mesothelial cells in addition to WBCs and label the sum as “nucleated cells.” The usefulness of mesothelial cell counts is not clear. The WBC count in uncomplicated ascites in the setting of cirrhosis is usually less than 500/mm³ (0.5 × 10⁹/L) (see Fig. 93-2).^10,19 During diuresis in patients with cirrhosis and ascites, the WBC count can concentrate to more than 1000/mm³ (1.0 × 10⁹/L).¹⁹ A diagnosis of diuresis-related elevation of the ascitic fluid WBC count, however, requires that a prediuresis count be available, that normal lymphocytes predominate in the fluid, and that unexplained clinical symptoms or signs (e.g., fever or abdominal pain) be absent.

The upper limit of normal for the absolute PMN count in uncomplicated ascitic fluid in cirrhosis is usually stated to be lower than 250/mm³ (0.25 × 10⁹/L).^10,19 The short survival of PMNs results in relative stability of the absolute PMN count during diuresis.¹⁹ Therefore, the 250/mm³ (0.25 × 10⁹/L) cutoff value remains reliable even after diuresis.

New methods have been developed to estimate the number of ascitic fluid cells.²⁰ Dipsticks can detect an ascitic fluid PMN count greater than 250/mm³ (0.25 × 10⁹/L) in 90 to 120 seconds. Urine-specific dipsticks have been used to date but are not very sensitive.²¹ An ascitic fluid–specific dipstick has been developed and calibrated to a PMN count of 250/mm³ (0.25 × 10⁹/L).²²

Any inflammatory process can result in an elevated ascitic fluid WBC count. SBP is the most common cause of inflammation of ascitic fluid and the most common cause of an elevated ascitic WBC count (see later). The total WBC count, as well as the absolute PMN count, is elevated in SBP, and PMNs usually account for more than 70% of the total WBC count. Also, in tuberculous peritonitis and peritoneal carcinomatosis, the total ascitic WBC count is frequently elevated, but usually with a predominance of lymphocytes.²

In most instances, bloody ascitic fluid is the result of a slightly traumatic tap. Leakage of blood into the peritoneal cavity leads to an elevated ascitic fluid WBC count. Because neutrophils predominate in blood, the ascitic fluid differential count may be altered by contamination of ascitic fluid with blood. To correct for this, 1 PMN is subtracted from the absolute ascitic fluid PMN count for every 250 RBCs¹⁹ (see Fig. 93-2). If the leakage of blood occurred at a remote time, the PMNs will have lysed, and the corrected PMN count will be a negative number. If the corrected PMN count in a bloody specimen is 250/mm³ (0.25 × 10⁹/L) or greater, the patient must be assumed to be infected.

Serum-Ascites Albumin Gradient

Before the 1980s, the ascitic fluid total protein concentration was used to classify ascites as either exudative (at least 2.5 g/dL [25 g/L]) or transudative (<2.5 g/dL [25 g/L]). Unfortunately, this classification does not work well in ascitic fluid, and these terms as applied to ascitic fluid were never carefully defined or validated. Attempts at using combinations of LDH and serum-to-ascitic fluid ratios of LDH and protein also have not been shown to classify ascitic fluid accurately into exudates and transudates.²³

The serum-ascites albumin gradient (SAAG) has been proved to categorize ascites better than the total protein concentration or other parameters²⁴ (Box 93-1). The SAAG is based on oncotic-hydrostatic balance. Portal hypertension results in an abnormally high hydrostatic pressure gradient between the portal bed and ascitic fluid. A similarly large difference must exist between ascitic fluid and intravascular oncotic forces. Albumin exerts greater oncotic force per gram than that exerted by other proteins. Therefore, the difference between the serum and ascitic fluid albumin concentrations correlates directly with portal pressure.

Box 93-1

Classification of Ascites by the Serum-Ascites Albumin Gradient

High Gradient ≥ 1.1 g/dL (11 g/L)

Alcoholic hepatitis

Budd-Chiari syndrome

Cardiac ascites

Cirrhosis

Fatty liver of pregnancy

Fulminant hepatic failure

Massive liver metastases

“Mixed” ascites

Myxedema

Portal vein thrombosis

Sinusoidal obstruction syndrome

Low Gradient < 1.1 g/dL (11 g/L)

Biliary ascites

Bowel obstruction or infarction

Nephrotic syndrome

Pancreatic ascites

Peritoneal carcinomatosis

Postoperative lymphatic leak

Serositis in connective tissue diseases

Tuberculous peritonitis

Calculating the SAAG involves measuring the albumin concentration of serum and ascitic fluid specimens and simply subtracting the ascitic fluid value from the serum value. Unless a laboratory error has been made, the serum albumin concentration is always the larger value. The gradient is calculated by subtraction and is not a ratio. If the SAAG is 1.1 g/dL (11 g/L) or greater, the patient can be considered to have portal hypertension with an accuracy of approximately 97%.²⁴ Also, if the serum albumin minus ascitic fluid total protein gradient is 1.1 g/dL (11 g/L) or greater, the patient has portal hypertension because the ascitic fluid albumin concentration cannot be greater than the ascitic fluid total protein concentration. Conversely, if the SAAG is less than 1.1 g/dL (11 g/L), the patient is unlikely to have portal hypertension. The SAAG does not explain the pathogenesis of ascites formation, nor does it explain where the albumin came from (i.e., liver or bowel). It simply gives the physician an indirect but accurate index of portal pressure. The accuracy of the test is excellent, even with ascitic fluid infection, diuresis, therapeutic paracentesis, IV infusions of albumin, and various causes of liver disease.²⁴

Measurement of the ascitic fluid albumin concentration has been routine in some laboratories since the 1980s. The accuracy of the albumin assay at low albumin concentrations (e.g., <1 g/dL [10 g/L]) should be confirmed because many patients with ascites have a serum albumin concentration in the range of 2.0 g/dL (20 g/L) and an ascitic fluid albumin concentration in the range of 0 to 1.0 g/dL (0 to 10 g/L). If a patient with cirrhosis has a serum albumin level of less than 1.1 g/dL (11 g/L), as occurs in less than 1% of patients with cirrhosis and ascites, the SAAG will be falsely low.

The accuracy of the SAAG is also reduced when specimens of serum and ascites are not obtained nearly simultaneously. The specimens should be obtained on the same day, preferably within the same hour. Both serum and ascitic fluid albumin concentrations change over time; however, these values change in parallel, so the difference is stable. Arterial hypotension may result in a decrease in the portal pressure and a narrowing of the SAAG. Lipid interferes with the assay for albumin, and chylous ascites may result in a falsely high SAAG.

Serum hyperglobulinemia (serum globulin level > 5 g/dL [50 g/L]) leads to a high ascitic fluid globulin concentration and can narrow the albumin gradient by contributing to the oncotic forces. A narrowed gradient caused by high serum globulin levels occurs in only approximately 1% of ascitic fluid specimens. To correct the SAAG in the setting of a high serum globulin level, the following formula is used²⁵:

Corrected SAAG=uncorrected SAAG×0.16 × (serum globulin [g/dL]+2.5

Approximately 5% of patients with ascites have “mixed” ascites (i.e., 2 causes of ascites) (see Table 93-1). Most of these patients have portal hypertension from cirrhosis as well as another cause of ascites, such as tuberculous peritonitis or peritoneal carcinomatosis.²⁴ The albumin gradient is high (≥1.1 g/dL [11 g/L]) in mixed ascites, as a reflection of the underlying portal hypertension.²⁴

The presence of a high SAAG does not confirm a diagnosis of cirrhosis; it simply indicates the presence of portal hypertension. Many causes of portal hypertension other than cirrhosis are recognized (Table 93-3; also see Table 93-1 and Chapter 92). A low SAAG does not confirm a diagnosis of peritoneal carcinomatosis. Although peritoneal carcinomatosis is the most common cause of a low SAAG, other causes exist (see Box 93-1). The SAAG needs to be determined only on the first paracentesis specimen in a given patient; it does not need to be repeated on subsequent specimens, if the first value is definitive. If the first result is borderline (e.g., 1.0 or 1.1 g/dL [10 or 11 g/L]), repeating the paracentesis and analysis usually provides a definitive result. High-albumin-gradient and low-albumin-gradient have replaced the modifiers “transudative” and “exudative” in the classification of ascites.²⁴

TABLE 93-3

Symptoms and Signs of Ascitic Fluid Infection

Symptom or Sign	Frequency (%)*
	SBP	Bacterascites	CNNA	Secondary Peritonitis	Polymicrobial Bacterascites
Fever	68	57	50	33	10
Abdominal pain	49	32	72	67	10
Abdominal tenderness	39	32	44	50	10
Rebound tenderness	10	5	0	17	0
Altered mental status	54	50	61	33	0

Culture

The most common bacterial infection of ascitic fluid, SBP, is monomicrobial, with a low bacterial concentration (median colony count of only 1 organism/mL).²⁶ In the past, culture methodology for ascitic fluid was based on the notion that most episodes of ascitic fluid infection were polymicrobial with high colony counts, as in surgical peritonitis. The older method of culture consisted of inoculation (in the microbiology laboratory) of each of 3 agar plates and some broth with a few drops of fluid. This method of culturing ascitic fluid, as is used for urine or stool, is predictably insensitive for detecting monomicrobial infections with a low colony count. SBP is more like bacteremia in terms of the number of bacteria present; culturing ascitic fluid as if it were blood has a high yield.²⁶ In fact, the sensitivity of culture in detecting bacterial growth in neutrocytic ascites (i.e., ascitic fluid with a PMN count of ≥ 250/mm³ [0.25 × 10⁹/L]) depends on the method of culture used. The older method of culture was found to detect bacterial growth in approximately 50% of neutrocytic samples, whereas bedside inoculation of blood culture bottles with ascitic fluid detects growth in approximately 80%.¹⁰ Multiple prospective studies have demonstrated the superiority of the blood culture bottle method.¹⁰ Also, bedside inoculation is superior to delayed laboratory inoculation of blood culture bottles in the laboratory.²⁷ Gene probes have become commercially available for the detection of bacteremia; hopefully, they will also lead to rapid (30-minute) and accurate detection of organisms in ascitic fluid. Culture will continue to be required, however, for assessment of the susceptibility of the organism to antibiotics.

Total Protein

The protein concentration in ascitic fluid in the setting of cirrhosis is determined almost entirely by the serum protein concentration and portal pressure. A patient with cirrhosis and a relatively high serum protein concentration will have a relatively high ascitic fluid protein concentration. Because of this relationship, almost 20% of ascitic fluid samples in patients with cirrhosis will have a protein concentration greater than 2.5 g/dL (25 g/L). The ascitic fluid total protein concentration does not increase during SBP; it remains stable before, during, and after infection.²⁸ In fact, patients with the lowest ascitic protein concentrations are the most susceptible to spontaneous peritonitis.²⁹ During a 10-kg diuresis, however, the ascitic fluid total protein concentration doubles, and 67% of such patients with cirrhosis and ascites have a protein concentration greater than 2.5 g/dL (25 g/L) by the end of diuresis.¹⁹ In almost one third of patients with malignancy-related ascites, the ascites is caused by massive liver metastases or hepatocellular carcinoma, and the ascitic fluid in these patients has a low protein concentration.² In cardiac ascites, the ascitic fluid protein concentration is usually greater than 2.5 g/dL (25 g/L).³⁰

The SAAG classifies fluid by the presence or absence of portal hypertension and is much more physiologic and intuitive than the old exudate/transudate classification. The exudate/transudate method of classification of ascites places many patients with cirrhosis and ascites and all patients with cardiac ascites in the exudate category, and many patients with malignant ascites and essentially all patients with spontaneously infected ascites in the transudate category. Clearly, this method of classification is not useful.²⁴ The albumin gradient classifies cardiac ascites in the high-SAAG category, similar to ascites in cirrhosis. The high SAAG of cardiac ascites is presumably the result of high right-sided cardiac pressures. In patients with cardiac ascites, the SAAG may narrow with diuresis; such narrowing does not happen in patients with cirrhosis.

The combination of ascitic fluid total protein, glucose, and LDH is of value in distinguishing SBP from intestinal perforation with leakage of intestinal contents into ascites³¹ (Fig. 93-3). Patients who have neutrocytic ascitic fluid, in whom the clinical picture suggests bacterial peritonitis (rather than peritoneal carcinomatosis or tuberculous peritonitis) and who meet 2 of the following 3 criteria, are likely to have surgical peritonitis and merit immediate radiologic evaluation to determine if intestinal perforation with leakage of intestinal contents into ascites has occurred: total protein greater than 1 g/dL (10 g/L), glucose less than 50 mg/dL (2.8 mmol/L), and LDH greater than the upper limit of normal for serum.³¹

Glucose

The glucose molecule is small enough to diffuse readily into body fluid cavities. Therefore, the concentration of glucose in ascitic fluid is similar to that in serum, unless glucose is being consumed by ascitic fluid WBCs or bacteria.³¹ In early SBP, the ascitic fluid glucose concentration is similar to that of sterile fluid.²⁸ By contrast, in SBP detected late in its course (as well as in the setting of intestinal perforation into ascitic fluid), the ascitic fluid glucose concentration usually drops to 0 mg/dL (0 mmol/L) because of large numbers of stimulated neutrophils and bacteria.³¹

Lactate Dehydrogenase

The LDH molecule is too large to enter ascitic fluid readily from blood,³¹ and the ascitic fluid concentration of LDH usually is less than half of the serum level in uncomplicated ascites in the setting of cirrhosis. In SBP, the ascitic fluid LDH level rises because of the release of LDH from neutrophils, and the ascitic fluid concentration is greater than that of serum. In secondary peritonitis, the LDH level is even higher than that seen in SBP and may be several-fold higher than the serum LDH level.³¹

Amylase

In uncomplicated ascites in the setting of cirrhosis, the ascitic fluid amylase concentration usually is half that of the serum value, approximately 50 U/L.³² In patients with acute pancreatitis or intestinal perforation (with release of luminal amylase into the ascitic fluid), the fluid amylase concentration is elevated markedly, usually greater than 2000 U/L, and approximately 5-fold greater than simultaneous serum values.^31-33

Gram Stain

Gram stains of body fluids demonstrate bacteria only when more than 10,000 bacteria/mL are present. The median concentration of bacteria in ascitic fluid in SBP is only 1 organism/mL, similar to the colony count in bacteremia.²⁶ Requesting an ascitic fluid Gram stain to detect bacteria in SBP is analogous to requesting a Gram stain of blood to detect bacteremia. Bacteria are detected on Gram stain only with overwhelming infection, as in advanced SBP or asplenic pneumococcal sepsis. Gram stain of ascitic fluid is most helpful in the diagnosis of free perforation of the intestine into ascitic fluid. In this setting, sheets of multiple different bacteria are found. Gram stain of the centrifuged sediment of 50 mL of ascites has a sensitivity of only 10% for visualizing bacteria in SBP.²⁶

Smear and Culture for TB

A direct smear of ascitic fluid to detect mycobacteria is almost never positive because of the rarity of tuberculous peritonitis and the low concentration of mycobacteria in ascitic fluid in tuberculous peritonitis.³⁴ The older literature suggests that 1 L of fluid should be cultured. The largest centrifuge tube found in most laboratories, however, has a capacity of 50 mL. In general, only 1 50-mL aliquot of fluid is centrifuged, and the pellet is cultured. In contrast to a sensitivity of approximately 50% for ascitic fluid mycobacterial culture with optimal processing, laparoscopy with histology and culture of peritoneal biopsies has a sensitivity approaching 100% for detecting tuberculous peritonitis.³⁴ Tuberculous peritonitis can easily be confused with SBP, because both conditions are associated with abdominal pain and fever and half of the patients with tuberculous peritonitis have cirrhosis. A negative bacterial culture and predominance of mononuclear cells in the differential count, however, provide clues to the diagnosis of tuberculous peritonitis. DNA probes have become available to detect mycobacteria and probably will replace older methods of detection.³⁵ Nevertheless, cultures still will be required to determine susceptibility to antimicrobial agents.

Cytologic Examination

Cytologic studies can be expected to detect malignancy only when tumor cells line the peritoneal cavity and exfoliate into the ascitic fluid (i.e., in peritoneal carcinomatosis). In the past, massive liver metastases and hepatocellular carcinoma superimposed on cirrhosis were not recognized as causes of “malignant” ascites.³⁶ Cytologic studies should not be expected to detect tumor when the peritoneum is uninvolved, as in ascites resulting from portal hypertension in patients with hepatocellular carcinoma or massive liver metastases or from lymph node obstruction in patients with malignant lymphoma.² In one study in which the location and type of tumor that caused ascites were confirmed by a standard test, only approximately two thirds of patients with malignancy-related ascites were found to have peritoneal carcinomatosis, but nearly 100% of patients with peritoneal carcinomatosis were reported to have positive findings on cytologic examination of ascitic fluid; the remaining one third of patients with massive liver metastases, chylous ascites caused by lymphoma, or hepatocellular carcinoma had negative cytologic findings.² Therefore, the sensitivity of cytology is approximately 100% for detecting peritoneal carcinomatosis but much lower for detecting malignancy-related ascites caused by conditions other than peritoneal carcinomatosis. Cytologic studies should not be falsely positive if performed carefully; I have never encountered a false-positive result.

Because hepatocellular carcinoma rarely metastasizes to the peritoneum, a positive ascitic fluid cytology in a patient with hepatocellular carcinoma is unusual enough to be the subject of a case report.³⁷ Measurement of the serum AFP concentration (which is always higher in serum than in ascitic fluid) may be of value in detecting hepatocellular carcinoma; serum AFP is much more sensitive than ascitic fluid cytology for this purpose (see Chapter 96).² In malignancy-related ascites, the fluid may have an elevated PMN count, presumably because dying tumor cells attract neutrophils.² The elevated PMN count may cause confusion with SBP; however, a predominance of lymphocytes in malignancy-related ascites is usual. Flow cytometry and magnetic enrichment of ascitic fluid as an adjunct to cytology may further increase diagnostic accuracy.³⁸

TG

A TG level should be measured in opalescent or frankly milky ascitic fluid (see Fig. 93-2). By definition, chylous ascites has a TG concentration greater than 200 mg/dL (2.26 mmol/L) and greater than the serum level; usually, the level is greater than 1000 mg/dL (11.30 mmol/L).³⁹ In sterile but slightly cloudy ascitic fluid specimens in the setting of cirrhosis, in the absence of an elevated cell count, the TG concentration is elevated: 64 ± 40 mg/dL (0.72 ± 0.45 mmol/L), compared with 18 ± 9 mg/dL (0.20 ± 0.10 mmol/L) in clear ascites in the setting of cirrhosis.¹⁶

Bilirubin

The bilirubin concentration should be measured in ascitic fluid that is dark brown. An ascitic fluid bilirubin level greater than 6 mg/dL (102 µmol/L) and greater than the serum level of bilirubin suggests biliary or proximal small intestinal perforation into ascitic fluid.^18,31

Tests That Are Seldom Helpful

Tests that have been proposed to be helpful in the analysis of ascitic fluid but shown subsequently to be of no benefit include determination of pH, lactate, fibronectin, and cholesterol. The studies that attempted to validate the value of pH and lactate included small numbers of patients and used suboptimal culture techniques. In the 2 largest studies, which did not have some of the deficiencies of smaller, earlier studies, the ascitic fluid pH and lactate were found not to be helpful.^40,41 The pH was found to have no impact on decision making regarding the use of empirical antibiotic therapy.⁴⁰

Fibronectin and cholesterol were proposed to be useful in detecting malignant ascites. The basic premise in studies of these markers was that ascitic fluid cytologic examination is insensitive. Unfortunately, the design of the studies was problematic, several subgroups of malignancy-related ascites (e.g., massive liver metastases, hepatocellular carcinoma with cirrhosis) were not considered, and appropriate control groups (e.g., patients with ascites caused by conditions other than cirrhosis or peritoneal carcinomatosis) were not included. Other studies have demonstrated that in patients with massive liver metastases, ascitic fluid fibronectin and cholesterol concentrations are not abnormally elevated.^42,43 Therefore, in patients with malignancy-related ascites and negative cytologic findings, these “humoral tests of malignancy” are usually negative. Additionally, patients with high-protein ascites in the absence of cirrhosis nearly always have ascitic fibronectin and cholesterol elevations despite the absence of malignancy.^2,42,43

CEA in ascitic fluid was proposed as a helpful marker for detecting malignant ascites.⁴⁴ The study that attempted to validate this proposal, however, was flawed, and more studies, with various subgroups of patients, are required before testing for ascitic fluid CEA can be considered validated.

Measurement of adenosine deaminase has been proposed as a useful test for detecting peritoneal TB. In the United States, however, where greater than 50% of patients with tuberculous peritonitis have underlying cirrhosis, the adenosine deaminase level has been found to be too insensitive to be helpful.³⁴

Classification

Of the 3 subtypes of spontaneous ascitic fluid infection, the prototype is SBP. The diagnosis of SBP is made when there is a positive ascitic fluid culture and an elevated ascitic fluid absolute PMN count (i.e., at least 250/mm³ [0.25 × 10⁹/L]) without evidence of an intra-abdominal surgically treatable source of infection.¹⁰ When Correia and Conn coined the term spontaneous bacterial peritonitis in 1975, their goal was to distinguish this form of infection from surgical peritonitis.⁶¹ Therefore, although many patients with SBP have a focus of infection (e.g., urinary tract infection or pneumonia), the diagnosis of SBP is still appropriate unless the focus requires surgical intervention (e.g., a ruptured viscus). I have not encountered a convincing case of polymicrobial SBP; all of the patients presumed to have SBP in whom ascitic fluid cultures initially grew more than 1 organism eventually were found to have surgical peritonitis or an erroneous culture result (e.g., a pathogen plus a contaminant or 2 colony morphologies of 1 species of bacteria).

The criteria for a diagnosis of monomicrobial non-neutrocytic bacterascites (MNB) include (1) a positive ascitic fluid culture for a single organism, (2) an ascitic fluid PMN count lower than 250/mm³ (0.25 × 10⁹/L), and (3) no evidence of an intra-abdominal surgically treatable source of infection.⁶² In the older literature, MNB was either grouped with SBP or labeled “asymptomatic bacterascites.” Because many patients with bacterascites have symptoms, the modifier asymptomatic seems inappropriate.

Culture-negative neutrocytic ascites (CNNA) is diagnosed when (1) the ascitic fluid culture grows no bacteria, (2) the ascitic fluid PMN count is 250/mm³ (0.25 × 10⁹/L) or greater, (3) no antibiotics have been given (not even a single dose), and (4) no other explanation for an elevated ascitic PMN count (e.g., hemorrhage into ascites, peritoneal carcinomatosis, TB, or pancreatitis) can be identified.⁶³ This variant of ascitic fluid infection is seldom diagnosed when sensitive culture methods are used.²⁶

Secondary bacterial peritonitis is diagnosed when (1) the ascitic fluid culture is positive (usually for multiple organisms), (2) the PMN count is 250/mm³ (0.25 × 10⁹/L) or greater, and (3) an intra-abdominal surgically treatable primary source of infection (e.g., perforated intestine, perinephric abscess) has been identified.^31,45 The importance of distinguishing this variant from SBP is that secondary peritonitis usually requires emergency surgical intervention (see Chapter 38).

Polymicrobial bacterascites is diagnosed when (1) multiple organisms are seen on Gram stain or cultured from the ascitic fluid and (2) the PMN count is lower than 250/mm³ (0.25 × 10⁹/L).⁶⁴ This diagnosis should be suspected when the paracentesis is traumatic or unusually difficult because of ileus or when stool or air is aspirated into the paracentesis syringe. Polymicrobial bacterascites is essentially diagnostic of intestinal perforation by the paracentesis needle.

Clinical Setting

The spontaneous variants of ascitic fluid infection—SBP, CNNA, and MNB—occur almost exclusively in the setting of severe liver disease. The liver disease usually is chronic (cirrhosis), but may be acute (fulminant hepatic failure) or subacute (alcoholic hepatitis). Cirrhosis of all causes can be complicated by spontaneous ascitic fluid infection. Spontaneous infection of noncirrhotic ascites is rare.

Essentially all patients with SBP have an elevated serum bilirubin level and prolonged prothrombin time because of advanced cirrhosis.¹⁰ Ascites appears to be a prerequisite for the development of SBP. The peritonitis is unlikely to precede the development of ascites. Usually, the infection develops when the volume of ascites is at its maximum.

Secondary bacterial peritonitis and polymicrobial bacterascites can develop with ascites of any type. The only prerequisite, in addition to the presence of ascites, is an intra-abdominal surgical source of infection.³¹ Such an infection can result from penetration of a needle into the bowel during attempted paracentesis.⁶⁴

Pathogenesis

Currently available evidence suggests that the spontaneous forms of ascitic fluid infection are the result of overgrowth of a specific organism in the intestine, translocation of that microbe from the intestine to mesenteric lymph nodes, and resulting spontaneous bacteremia and subsequent colonization of susceptible ascitic fluid^65,66 (Fig. 93-4). When bacteria enter the fluid in the abdomen, by whatever route, a battle ensues between the virulence factors of the organism and the immune defenses of the host.⁶⁷ The ascitic fluid protein concentration does not change with development of spontaneous infection.²⁸ Low-protein ascitic fluid (e.g., protein content < 1 g/dL [10 g/L]) is particularly susceptible to SBP.²⁹ The endogenous antimicrobial (opsonic) activity of human ascitic fluid correlates directly with the protein concentration of the fluid.⁶⁶ Patients with deficient ascitic fluid opsonic activity are predisposed to SBP.⁶⁸ Patients with detectable ascitic fluid opsonic activity appear to be protected from SBP unless they are exposed to a particularly virulent organism (e.g., Salmonella).^67,68

Studies in both patients and animals with cirrhosis demonstrate that MNB is common.^62,69 Pieces of bacterial DNA are commonly present in serum and ascitic fluid of patients with cirrhosis.⁷⁰ In both humans and rats, most episodes of bacterascites resolve without antibiotic treatment.^62,69 The fluid frequently becomes sterile without an increase in ascitic PMNs. Apparently, the host's defense mechanisms are able to eradicate the invading bacteria on most occasions. Uncontrolled infection probably develops only when the defenses are weak or the organism is virulent (see Fig. 93-4). Bacterascites probably is more common than SBP. Conceivably, ascitic fluid in the setting of cirrhosis is colonized regularly by bacteria, and almost just as regularly, the colonization resolves. The entry of PMNs into the fluid probably signals failure of the peritoneal macrophages to control the infection.⁷¹ A majority of episodes of MNB appear to resolve in rats and humans with cirrhosis, whereas untreated SBP is frequently fatal. In summary, MNB probably represents an early stage of ascitic fluid infection, which can resolve or progress to CNNA or to SBP.

Most episodes of CNNA are diagnosed by insensitive culture methods for which the numbers of bacteria are insufficient to reach the threshold of detectability.²⁶ Inoculation of ascitic fluid into blood culture bottles can lead to detection of a single organism in the cultured aliquot of fluid, whereas the older method of culture by inoculation of agar plates and broth probably requires at least 100 organisms/mL (see earlier).²⁶ Even when optimal culture methods are used, however, a variable percentage of specimens of neutrocytic ascitic fluid grow no bacteria. A study of rapid sequential paracenteses (before the initiation of antibiotic treatment) in patients with CNNA demonstrated that, in most cases, the PMN count dropped spontaneously and the culture results remained negative in the second specimen.⁷² When sensitive culture techniques are used, CNNA probably results from (1) previous antibiotic treatment (even 1 dose), (2) an inadequate volume of fluid inoculated, or (3) spontaneously resolving SBP in which the paracentesis is performed after all bacteria have been killed by host defenses but before the PMN count has normalized.

The pathogenesis of secondary bacterial peritonitis is more straightforward than that of SBP. When the intestine perforates, billions of bacteria flood into the ascitic fluid. In the absence of a frank perforation, bacteria may cross inflamed tissue planes and enter the fluid. The pathogenesis of polymicrobial bacterascites is also obvious.⁶⁴ A paracentesis needle enters the bowel, and the bowel contents are released into the ascites.

Symptoms and Signs

Although 87% of patients with SBP are symptomatic at the time the infection is diagnosed, the symptoms and signs of infection are often subtle, such as a slight change in mental status.⁶² Without prompt paracentesis, the diagnosis and treatment of infected ascites may be delayed, often resulting in the death of the patient. The symptoms and signs manifested in all 5 variants of ascitic fluid infection are listed in Table 93-3.

Frequency

Since the 1980s, routine paracenteses at the time of hospitalization in patients with ascites have provided data regarding the frequency of ascitic fluid infection. In the 1980s, approximately 10% of patients with ascites were infected at the time of hospital admission; of the subgroup of patients with cirrhosis, about 27% were infected.¹⁰ At present, because of measures to prevent SBP, the frequency has dropped significantly (see later). Of patients with culture-positive ascitic fluid, about two thirds have neutrocytic ascitic fluid (SBP), and one third have MNB.⁶² The frequency of CNNA depends largely on the culture technique (see earlier). Polymicrobial bacterascites occurs in only 1 in 1000 paracenteses. Secondary bacterial peritonitis is found in only 0% to 2% of patients with ascites at the time of hospital admission.^10,31

Bacteriology

Escherichia coli, streptococci (mostly pneumococci), and Klebsiella caused most episodes of SBP and MNB in the past in patients who were not receiving selective intestinal decontamination. Now more Gram-positive and resistant Gram-negative organisms are being isolated (Table 93-4).⁷³ Selective intestinal decontamination and recent exposure to antibiotics are causing a change in the bacteria isolated from patients in whom an ascitic fluid infection develops. We may encounter organisms for which we have no effective antibiotics.⁷⁴ The antibiotic susceptibility of bacteria that cause infections in patients with cirrhosis should be tested periodically in each hospital, and the preferred empirical antibiotic regimens should be adapted accordingly.⁷³

CNNA is, by definition, culture-negative, and polymicrobial bacterascites is, by definition, polymicrobial. The most apparent difference between the spontaneous forms of ascitic fluid infection and the secondary forms (secondary peritonitis and polymicrobial bacterascites) is that the former always are monomicrobial and the latter usually are polymicrobial. Although older papers reported that anaerobic bacteria were present in approximately 6% of cases of SBP, the detection of anaerobes probably reflected unrecognized cases of secondary bacterial peritonitis. In more recent series, anaerobes have been found in approximately 1% of cases of SBP and MNB.^26,62

Risk Factors

Patients with cirrhosis are unusually predisposed to bacterial infection because of multiple defects in immune defense. The concept that cirrhosis is a form of acquired immunodeficiency (in the generic sense) is relatively new. In a prospective study, a bacterial infection occurred in 34% of 405 patients with cirrhosis at the time of admission to the hospital or during the hospitalization.⁷⁵ Low ascitic fluid total protein concentrations, as well as the phagocytic (both motile and stationary) dysfunction associated with cirrhosis, are risk factors for bacterial infection.

Paracentesis itself has been proposed as a risk factor for ascitic fluid infection, but this theoretical risk has not been substantiated in prospective studies of paracentesis-related complications.¹² SBP is statistically more likely to be diagnosed on the first paracentesis than on subsequent taps.¹² Needle-induced ascitic fluid infections do not occur unless the bowel is penetrated by the paracentesis needle.^12,64 Fortunately, bowel penetration occurs in only 1 in 1000 taps. One would expect bacteria of the skin flora to be isolated more frequently if poor paracentesis technique were the cause of many cases of SBP; yet skin flora microorganisms are seldom isolated from ascitic fluid when sterile technique is used.²⁶ Iatrogenic peritonitis is most likely to occur when the paracentesis needle enters the bowel during a difficult paracentesis.

GI hemorrhage is an under-recognized risk factor for the development of spontaneous bacteremia and SBP (see Chapters 20 and 92). The cumulative probability of infection during a single hospitalization for bleeding is approximately 40%.⁷⁶ The risk appears to peak 48 hours after the onset of hemorrhage. The high risk of infection probably is mediated by a shock-induced increase in the translocation of bacteria from the intestine to extraintestinal sites. Urinary tract infections also constitute an under-recognized risk factor for SBP.⁷⁷

Diagnosis

Timely diagnosis of ascitic fluid infection requires a high index of suspicion and a low threshold for performing a paracentesis. Clinical deterioration, especially fever or abdominal pain, in a patient with ascites should raise the suspicion of infection and prompt a paracentesis. If the ascitic fluid PMN count is elevated, the working diagnosis is ascitic fluid infection until proved otherwise. Although peritoneal carcinomatosis, pancreatitis, hemorrhage into ascites, and TB can lead to an elevated ascitic fluid PMN count, most cases of neutrocytic ascites are caused by bacterial infection. A predominance of PMNs in the WBC differential count lends further support to the diagnosis of infection. An elevated absolute ascitic fluid PMN count with a predominance of neutrophils in a clinical setting compatible with infection should prompt empirical antibiotic therapy (see later).

Although SBP is approximately 20 times as common as surgical peritonitis in a patient with ascites, secondary peritonitis should be considered in any patient with neutrocytic ascites (see Chapter 38).⁴⁵ Clinical symptoms and signs do not distinguish patients with secondary peritonitis from those with SBP (see Fig. 93-3).³¹ Even with free perforation of the colon into ascitic fluid, a classic surgical abdomen does not develop. Peritoneal signs require contact of inflamed visceral and parietal peritoneal surfaces, and such contact does not occur when there is a large volume of fluid separating these surfaces. Intestinal perforation can be suspected and pursued if a specimen of ascites is neutrocytic and meets 2 of the following 3 criteria (see Fig. 93-3): (1) total protein greater than 1 g/dL (10 g/L), (2) glucose less than 50 mg/dL (2.8 mmol/L), and (3) LDH greater than the upper limit of normal for serum.³¹ In the setting of a perforated viscus, cultures of ascitic fluid nearly always disclose multiple organisms, except in gallbladder rupture, which is usually monomicrobial.¹⁸ Brown ascitic fluid with a bilirubin concentration that is greater than 6 mg/dL (102 µmol/L) and greater than the serum level is indicative of biliary or proximal small intestinal perforation into ascites.¹⁸ An ascitic fluid amylase level that is greater than 5-fold that of the serum level also may be indicative of intestinal rupture (but not gallbladder rupture) with the release of luminal amylase.^31,32

The initial ascitic fluid analysis is helpful in delineating which patients are likely to have a ruptured viscus (see Fig. 93-3). Within minutes of the detection of neutrocytic ascitic fluid, these patients should undergo abdominal CT to confirm and localize the site of rupture.⁴⁵ If perforation is documented, emergency surgical intervention is the next step. Timing is crucial; after septic shock occurs, death is nearly certain. Antibiotic therapy without surgical intervention in the treatment of a ruptured viscus is predictably unsuccessful.

In contrast to patients with peritonitis resulting from perforation of a viscus, patients with secondary peritonitis unrelated to perforation tend not to have a diagnostic initial ascitic fluid analysis.³¹ The need to make the diagnosis of secondary peritonitis in patients without free perforation is less urgent, and there may be time to evaluate the response of the ascitic PMN count and fluid culture to treatment with antibiotics. It is best to repeat the paracentesis to assess the response to treatment after 48 hours of therapy; by 48 hours, the ascitic PMN count will be lower than the pretreatment value and the ascitic culture will be negative in essentially every patient with SBP who has been treated with an appropriate antibiotic.³¹ Before 48 hours of treatment, the ascitic PMN count may rise to a value higher than baseline in either SBP or secondary peritonitis.³¹ The culture remains positive in secondary peritonitis and becomes rapidly negative in SBP (see Fig. 93-3).³¹ Whereas antibiotics alone cannot control secondary peritonitis, medical therapy cures SBP rapidly.³¹

Treatment

Patients with an ascitic fluid PMN count of 250/mm³ (0.25 × 10⁹/L) or greater and a clinical scenario compatible with ascitic fluid infection should receive empirical antibiotic treatment (Table 93-5).^10,78 Patients with hemorrhage into the ascitic fluid, peritoneal carcinomatosis, pancreatic ascites, or tuberculous peritonitis may have an elevated ascitic PMN count that is unrelated to SBP and usually do not require empirical antibiotic treatment. If they do receive antibiotics, the ascitic PMN count usually fluctuates randomly, in contrast to the dramatic reduction in PMN count typical of SBP. If the clinical picture is unclear initially, the physician should err on the side of antibiotic treatment (with a non-nephrotoxic antibiotic). If ascitic fluid cultures are negative, the antibiotic can be stopped after 48 hours. In patients with uninfected neutrocytic ascitic fluid (except those with hemorrhage), lymphocytes usually predominate in the ascitic fluid differential count, in contrast to those with SBP, in whom PMNs predominate. In patients with bloody ascitic fluid, a “corrected” PMN count should be calculated (as discussed earlier). Antibiotic therapy is not necessary for patients with bloody fluid unless the corrected ascitic fluid PMN count is 250/mm³ (0.25 × 10⁹/L) or greater.

The decision to begin empirical antibiotic treatment in patients with bacterascites must be individualized. Many episodes resolve without treatment⁶²; however, the hospital mortality rate of 32% in patients when MNB is attributable, at least, in part, to infection.⁶² Therefore, treatment appears to be warranted in many patients. By definition, the ascitic PMN count is lower than 250/mm³ (0.25 × 10⁹/L) in this variant of ascitic fluid infection, and the PMN count cannot be the only parameter on which to base the decision about empirical therapy. Most patients with MNB in whom the colonization does not resolve progress to SBP and have symptoms or signs of infection at the time of the paracentesis that documents bacterascites.⁶² Therefore, patients with cirrhosis and ascites who have convincing symptoms or signs of infection should receive treatment regardless of the ascitic fluid PMN count. Empirical treatment can be discontinued after only 2 to 3 days if the culture demonstrates no growth. Asymptomatic patients may not need treatment.⁶² The paracentesis should be repeated for cell count and culture in patients without clinical evidence of infection, as soon as it is known that the initial culture result is positive. If the PMN count has risen to at least 250/mm³ (0.25 × 10⁹/L) or if symptoms or signs of infection have developed, antibiotics should be started. Culture results usually are negative in patients without a rise in the ascitic fluid PMN count on repeat paracentesis and without clinical evidence of infection, and these persons do not require antibiotics⁶² because colonization has been eradicated by host immune defenses.

The physician will not know initially that the ascitic fluid culture is destined to be negative in a patient with CNNA; therefore, empirical antibiotic treatment should be started. When the preliminary culture demonstrates no growth, it is helpful to repeat the paracentesis after 48 hours of therapy to assess the response of the PMN count to antibiotics. A dramatic decline in PMN count (always below the baseline pretreatment value and frequently a reduction of more than 80%) confirms a response to treatment.³¹ In such cases, a few more days of therapy is probably warranted. A stable ascitic fluid PMN count, especially with a predominance of lymphocytes and monocytes, suggests a nonbacterial (or mycobacterial) cause of ascitic fluid neutrocytosis, and the fluid should be sent for cytologic examination and mycobacterial culture. Because a negative culture result may be due to insensitive culture techniques, the prevalence of CNNA in a hospital that still uses older methods of culture (sending a tube or syringe of fluid to the laboratory) can be reduced by convincing the microbiology laboratory to accept and process ascitic fluid submitted in blood culture bottles.²⁶

Gram stain of the ascitic fluid is most helpful in detecting secondary peritonitis, in which multiple different bacterial forms are seen, but is of little value in guiding the choice of empirical antibiotic treatment for spontaneous ascitic infections (see earlier). I have found that use of the Gram stain did not help narrow the antibiotic coverage in even 1 patient of approximately 700 with SBP. Only approximately 10% of Gram stains demonstrate organisms in SBP.²⁶ If a Gram stain indicates secondary peritonitis, coverage for anaerobic flora, in addition to coverage for aerobic and facultative anaerobic flora, is required, as is an emergency search for the source of the infection (see Fig. 93-3 and Table 93-5).³¹ Therefore, a positive Gram stain may lead to broader antibiotic coverage, rather than narrower coverage. Narrow coverage based on a misinterpretation of the significance of the results of the Gram stain may lead to uncontrolled infection and death before resistance of the isolated organism to the chosen antibiotic becomes apparent.

Until the results of susceptibility testing are available, relatively broad-spectrum antibiotic therapy is warranted in patients with suspected ascitic fluid infection. After sensitivities are known, the spectrum of coverage can usually be narrowed. The antibiotics that have been recommended for empirical treatment have changed over the years. In the late 1970s, the combination of ampicillin and gentamicin was promoted, but this recommendation was not based on susceptibility testing or efficacy data. One study has documented an adjusted odds ratio of 4.0 for aminoglycosides as a risk factor for renal dysfunction in patients with cirrhosis.⁷⁹ Evidence that newer aminoglycosides are less nephrotoxic than gentamicin is lacking.

Several antibiotics are now available for the treatment of ascitic fluid infection. Cefotaxime, a third-generation cephalosporin, has been shown in a controlled trial to be superior to ampicillin plus tobramycin for the treatment of SBP.⁸⁰ Fully 98% of causative organisms were susceptible to cefotaxime, which did not result in superinfection or nephrotoxicity.⁸⁰ Cefotaxime or a similar third-generation cephalosporin appears to be the treatment of choice for suspected SBP.¹⁰ Anaerobic coverage is not needed.²⁶ Cefotaxime, 2 g IV every 8 hours, has been shown to result in excellent ascitic fluid levels (20-fold killing power after 1 dose).⁸¹ In patients with a serum creatinine level greater than 3 mg/dL, the dosing interval may be extended to 12 hours.⁸¹ Neither a loading dose nor an intraperitoneal dose appears to be necessary or appropriate. The clinician should, however, order the first dose “STAT,” to avoid a delay in administration of the life-saving agent. More broad-spectrum antibiotics are warranted if local susceptibility testing demonstrates increasing resistance to third-generation cephalosporins.

Prognosis

In the past, 48% to 95% of patients with a spontaneous ascitic fluid infection died during the hospitalization in which the diagnosis was made, despite antibiotic treatment.^10,23 By 1991, mortality rates were low (<5% if antibiotics are administered in a timely fashion), probably because of earlier detection and treatment of infection, as well as the avoidance of nephrotoxic antibiotics.⁹¹ The trial in which cefotaxime plus albumin was studied reported an inpatient mortality rate of 10%.⁸⁵ Even now, however, some patients are cured of their infection but die of liver failure or GI bleeding because of the severity of the underlying liver disease. In fact, spontaneous ascitic fluid infection is a marker of end-stage liver disease and has been proposed as an indication for liver transplantation in a patient who is otherwise a candidate.

To maximize survival, it is important that paracentesis is performed in all patients with ascites at the time of hospitalization, so that infection can be detected and treated promptly. The ascitic fluid cell count should be reviewed as soon as the results are available (≈60 minutes), and appropriate treatment should be instituted if indicated. The first dose of antibiotic should be given immediately. Because dipstick test results (see earlier) are available in 90 to 120 seconds, this tool may speed treatment of SBP and improve survival.^20,22

Paracentesis should be repeated during the hospitaliza­tion if any manifestation of clinical deterioration develops, including abdominal pain, fever, change in mental status, renal failure, acidosis, peripheral leukocytosis, or GI bleeding. If the physician waits to perform a paracentesis until convincing symptoms and signs of peritoneal infection have developed, the infection is likely to be advanced by the time the diagnosis is made. No survivors of SBP have been reported when the diagnosis was made after the serum creatinine level had risen above 4 mg/dL (350 µmol/L).

Without surgical intervention, the mortality rate for secondary peritonitis in hospitalized patients with ascites approaches 100%. When secondary peritonitis is diagnosed early and treated with emergency laparotomy, the mortality rate is 50% to 67%.^31,45

Prevention

The identification of risk factors for SBP (including an ascitic fluid protein concentration < 1.0 g/dL, variceal hemorrhage, and previous episode of SBP) has led to controlled trials of prophylactic antibiotics.^29,92-94 Norfloxacin, 400 mg/day orally, has been reported to reduce the risk of SBP in inpatients with low-protein ascites and those with previous SBP.^92,93 Norfloxacin, 400 mg orally twice daily for 7 days, helps prevent infection in patients with variceal hemorrhage⁹⁴ and is cost-effective in preventing recurrent SBP.⁹⁵ IV ceftriaxone 1 g daily for 7 days has been found to be even more effective than norfloxacin in the setting of GI bleeding; this regimen allows administration of antibiotics to patients who are vomiting blood.⁹⁶ Oral antibiotics select for resistant organisms in the intestinal flora in patients, and in animals these organisms can then cause spontaneous ascitic fluid infection.^90,97 Despite this concern, 2 randomized trials of primary prevention of ascitic fluid infection with prophylactic norfloxacin or ciprofloxacin have demonstrated a survival advantage for the antibiotic-treated patients (Table 93-6).^98-100

Trimethoprim/sulfamethoxazole has also been shown to prevent SBP in an animal model and in patients; in animals survival was increased.^101,102 The recommended dose for patients is 1 double-strength tablet daily.¹⁰²

Use of parenteral antibiotics to prevent endoscopic sclerotherapy–related or band ligation–related infections in nonbleeding patients does not appear to be warranted, as indicated by a controlled trial.¹⁰³ Active bleeding, not endoscopic treatment, appears to be the risk factor for ascitic fluid infection. On the other hand, bacterial infection is associated with failure to control variceal hemorrhage.¹⁰⁴ This observation provides additional incentive to try to prevent, detect, and treat infections aggressively in this setting to minimize mortality related not only to infection, but also to hemorrhage.

Cellulitis

Cellulitis of the lower extremities or abdominal wall is a common cause of soft tissue infection in obese patients with edema. One study has documented a 19% cumulative probability of cellulitis during hospitalization of patients with cirrhosis and ascites, compared with only a 4% likelihood of SBP.¹⁰⁵ Risk factors for cellulitis included obesity (which is increasing in frequency in patients with cirrhosis), homelessness, and greater degree of edema.¹⁰⁵ Another study has demonstrated a 22% risk of renal failure in patients with cellulitis and cirrhosis.¹⁰⁶ A high index of suspicion and low threshold for treatment with a first-generation cephalosporin or other antibiotic may help decrease morbidity and mortality from uncontrolled cellulitis.

Tense Ascites

Some patients with ascites do not seek medical attention until they can no longer breathe or eat comfortably because of the pressure of the intra-abdominal fluid on the diaphragm. Tense ascites requires urgent therapeutic paracentesis (see later). Contrary to folklore, tense ascites can be drained without untoward hemodynamic effects.¹⁰⁷ “Total paracentesis,” even more than 22 L, has been demonstrated to be safe.¹⁰⁷ In the setting of tense ascites, therapeutic paracentesis improves venous return and hemodynamics; the myth of paracentesis-related hemodynamic catastrophes was based on anecdotal observations in small numbers of patients.

Pleural Effusions

Pleural effusions can occur in patients with cirrhosis and ascites. They usually are unilateral and right-sided but occasionally may be bilateral and larger on the right side than on the left. A unilateral left-sided effusion suggests TB. A large effusion in a patient with cirrhotic ascites is designated hepatic hydrothorax.¹⁰⁸ Most carefully studied patients with hepatic hydrothorax have been shown to have a small defect in the right hemidiaphragm. With large diaphragmatic defects, ascites may be undetectable on clinical examination despite a large pleural effusion.

The most common symptom associated with hepatic hydrothorax is shortness of breath. Infection of the fluid can occur, usually as a result of SBP and transmission of bacteria across the diaphragm.¹⁰⁹ The analysis of uncomplicated hepatic hydrothorax fluid is similar but not identical to that of ascitic fluid, because the pleural fluid is subject to hydrostatic pressures different from those that affect the portal bed. The total protein concentration is higher (by ≈ 1.0 g/dL [10 g/L]) in the pleural fluid than in ascitic fluid.¹⁰⁸

The treatment of hepatic hydrothorax was difficult until the transjugular intrahepatic portosystemic shunt (TIPS) became available (see later).¹⁰⁸ The effusions tend to occur in patients who are the least adherent to treatment regimens or in whom ascites is most refractory to therapy. Some authors have recommended chest tube insertion and sclerosing of the pleurae with tetracycline; however, chest tubes inserted to treat hepatic hydrothorax are usually difficult to remove,¹¹⁰ Moreover, shortness of breath may recur when the tube is clamped, and fluid may leak around the insertion site of the tube. A peritoneovenous shunt (see later) can be considered when the patient with hepatic hydrothorax has large-volume ascites, but the shunt usually clots after a short time. Direct surgical repair of the diaphragmatic defect can be considered, but the patients typically are poor operative candidates. Video thoracoscopic suture of the hole in the diaphragm followed by pleurodesis has been reported to be successful.¹¹¹ Sodium restriction plus use of diuretics with intermittent thoracentesis is the safest and most effective first-line therapy of hepatic hydrothorax. TIPS placement has been reported to be successful and constitutes reasonable second-line treatment.¹⁰⁸ If the patient is a candidate for liver transplantation, proceeding with a transplant evaluation may be the best approach.

Abdominal Wall Hernias

Abdominal wall hernias are common in patients with ascites. They usually are umbilical or incisional and occasionally inguinal. Up to 20% of patients with cirrhosis and ascites have umbilical hernias at the time of hospitalization.¹¹² Some of these hernias incarcerate or perforate. Because of these potential complications, elective surgical treatment should be considered in a patient with a hernia and ascites. Insertion of mesh should be avoided because of the potential for the mesh to become infected. The ascitic fluid should be medically removed preoperatively because the hernia recurs in 73% of patients who have ascites at the time of hernia repair but in only 14% of those who have no ascitic fluid at the time of repair.¹¹³ Nevertheless, hernia repair is not without hazard. Successful laparoscopic repair of a recurrent strangulated umbilical hernia has been described.¹¹⁴ TIPS has also been reported to lead to good control of symptoms and may obviate the need for surgical repair.¹¹⁵ Many transplant surgeons prefer to avoid repair of the hernia or postpone it until the time of liver transplantation. An elastic abdominal binder can be used as a temporizing measure to reduce pain and hernia enlargement.

Surgical repair of a hernia or TIPS should be performed urgently in patients with skin ulceration, crusting, or black discoloration and emergently for refractory incarceration or rupture. Rupture is the most feared complication of an umbilical hernia. If TIPS is used, it must be performed prior to bacteremia. Infection of the TIPS may be difficult to eradicate.

Low-Albumin-Gradient Ascites

Peritoneal carcinomatosis is the most common cause of low-albumin-gradient ascites.² Peripheral edema in affected patients can be managed with diuretics. By contrast, patients without peripheral edema who receive diuretics lose only intravascular volume, without loss of ascitic fluid. The mainstay of treatment of nonovarian peritoneal carcinomatosis is outpatient therapeutic paracentesis. Patients with peritoneal carcinomatosis usually live only a few months. Patients with ovarian malignancy are an exception to this rule and may exhibit a good response to surgical debulking and chemotherapy.

Ascites caused by tuberculous peritonitis (without cirrhosis) is cured by anti-TB therapy. Diuretics do not speed weight loss unless the patient has underlying portal hypertension from cirrhosis. Pancreatic ascites may resolve spontaneously, require endoscopic placement of a stent in the pancreatic duct or operative intervention, or respond to treatment with somatostatin.¹¹⁶ A postoperative lymphatic leak from a distal splenorenal shunt or radical lymphadenectomy also may resolve spontaneously but on occasion may require surgical intervention or placement of a peritoneovenous shunt. Chlamydia peritonitis is cured by doxycycline. Ascites caused by lupus serositis may respond to glucocorticoids.⁷ Dialysis-related ascites may respond to aggressive dialysis.⁵¹

High-Albumin-Gradient Ascites

Cirrhosis is the most common cause of liver disease that leads to high-albumin-gradient ascites (see Table 93-1). Many patients with cirrhosis experience multiple insults to the liver, including excessive alcohol use, NASH, and chronic hepatitis C.⁴ One of the most important steps in treating high-albumin-gradient ascites in a patient with alcoholic liver disease, with or without other causes of liver injury, is to convince the patient to stop drinking alcohol. In a period of months, abstinence from alcohol can result in healing of the reversible component of alcoholic liver disease, and the ascites may resolve or become more responsive to medical therapy.

Baclofen has been shown in a randomized controlled trial to reduce alcohol craving and alcohol consumption, specifically in patients with alcoholic liver disease.¹¹⁷ It can be started at 5 mg orally 3 times daily and increased to 10 mg 3 times daily or even higher until craving is suppressed.

Similarly, patients with other forms of treatable liver disease (e.g., autoimmune hepatitis, chronic hepatitis B with reactivation, hemochromatosis, Wilson disease) should receive specific therapy for these diseases. Occasionally, cirrhosis due to causes other than alcohol or hepatitis B is reversible⁵; however, these diseases are usually less reversible than alcoholic liver disease, and by the time ascites is present, these patients may be better candidates for liver transplantation than for protracted medical therapy.

Refractory Ascites

Refractory ascites is defined as ascites unresponsive to a sodium-restricted diet and high-dose diuretic treatment. Refractoriness may manifest as minimal or no weight loss despite diuretics or the development of complications of diuretics.¹³⁹ Studies have shown that ascites in the setting of cirrhosis is refractory to standard medical therapy in fewer than 10% of patients.^124,128

Viable options for patients refractory to routine medical therapy include careful attention to medications, liver transplantation, serial therapeutic paracenteses, TIPS, and peritoneovenous shunts (Fig. 93-5).¹⁰