Abstract
Ascitic fluid samples are frequently sent to the laboratory for analysis. Although the underlying cause of the ascites is often thought to be clinically obvious, it is important to establish a definitive diagnosis. The value of a cell count and bacterial culture of the ascitic fluid is not disputed, but the role of biochemical testing is less clear. The use of ascitic fluid total protein to try to classify ascitic fluids as either an exudate or a transudate has contributed to this. The use of the physiologically based serum ascites albumin gradient to differentiate ascites caused by portal hypertension from other causes provides a better diagnostic approach. We recommend that the serum ascites albumin gradient is performed by laboratories as the first-line test and that interpretative reports are provided. Additional testing should be restricted to specific diagnostic queries and requires close collaboration between the laboratory and the clinician.
Introduction
The accumulation of fluid in the peritoneal cavity constitutes a peritoneal effusion. This is also termed ascites, which is derived from the Greek askos meaning bladder, belly or bag. 1 In this article the two are used interchangeably.
Along with pleural fluid, peritoneal fluid is frequently sent to the laboratory for biochemical analysis. The clinical utility of pleural fluid examination has been reviewed and this article takes a similar approach to assess critically the usefulness of the analytes often requested in the investigation of peritoneal fluid. 2
The American Association for the Study of Liver Disease (AASLD) and the British Society for Gastroenterology have both recently published guidelines on the management of ascites in cirrhosis and also make recommendations about ascitic fluid investigation (Table 1). 3,4 Much of the literature in this area is quite old and there have been major advances in imaging techniques in recent years. Nonetheless, the AASLD guidelines state that ‘abdominal paracentesis with appropriate ascitic fluid analysis is probably the most rapid and cost-effective method of diagnosing the cause of ascites’.
Summary of the investigations recommended in the AASLD and BSG guidelines
AASLD, American Association for the Study of Liver Disease; BSG, British Society of Gastroenterology; SAAG, serum ascites albumin gradient; LD, lactate dehydrogenase; AFB, acid fast bacilli
*Inoculated into blood culture bottles at the bedside
This review includes those analytes reported to be useful in the differential diagnosis of ascites, discusses the recently published guidelines and makes recommendations for a rational and pragmatic approach to the investigation of peritoneal fluid samples sent to the laboratory for evaluation.
Peritoneal fluid formation
A thin layer of peritoneal fluid, which is an ultrafiltrate of plasma, separates the two layers of peritoneum. Clinically ascites is detected by the presence of flank dullness to percussion, but is not usually apparent until more than 500 mL of fluid has accumulated. 5 Radiological techniques, such as rectal or transvaginal ultrasonography, may however be able to detect volumes of less than 50 mL, and may also suggest the cause. 6
In accordance with Starling's hypothesis, there are two important factors in the formation of ascites: the plasma colloid osmotic pressure and the portal venous pressure. Interchange of fluids between the blood and the tissue spaces is controlled by the balance between these.
7
As for pleural fluid, in simple terms, the accumulation of peritoneal fluid occurs when either fluid formation is increased due to:
elevation of the hydrostatic pressure gradient, in this instance the portal venous pressure, a decrease in colloid osmotic pressure, increased permeability of the peritoneal capillaries
or if fluid removal is decreased.
The pathophysiology of ascites associated with cirrhosis is more complicated. A number of factors are involved including:
portal hypertension,* splanchnic arterial vasodilation,* salt and water retention*, secondary hyperaldosteronism and alteration in hepatic and intestinal lymph formation.
(*denotes those thought to be of greatest importance
8,9
).
Causes of ascites and differential diagnosis
The most common cause of ascites in patients from Western Europe and North America is cirrhosis, which accounts for approximately 80% of cases. 3,4 It is reported that approximately 50% of patients with cirrhosis develop ascites within 10 y of observation and that half of these will die within 2 y. 10 The other common causes of ascites are malignancy (approximately 10%) and cardiac failure (approximately 5%), with tuberculosis (TB) assuming increasing importance in some populations. The main causes of peritoneal effusions are listed in Table 2.
Causes of peritoneal effusion
CCF, congestive cardiac failure; TB, tuberculosis; SLE, systemic lupus erythematosis
The major causes of malignant peritoneal effusions are shown separately in Table 3. The development of a malignant peritoneal effusion is of grave prognostic significance with survival in these patients averaging less than five months, although for those with ovarian cancer the average survival time was 10 months. 11
Main causes of malignant ascites – note that peritoneal carcinomatosis accounts for two-thirds of patients with malignancy-related ascites
Data adapted from Sears and Hajdu 11
Transudate or exudate?
Traditionally, as with pleural fluid, peritoneal fluid has been classified on the basis of the total protein (TP) content into either an exudate or transudate, attempting to reflect the different processes of fluid formation. 2 The cut-off values used have varied but mostly lie between 25 and 30 g/L. 12–14
The transudate/exudate concept in ascites has, however, resulted in many problems and exceptions. The pathophysiology operating in the formation of ascites does not always follow the expected pattern. For example, in malignant ascites due to hepatic metastases, the underlying cause of the ascites is the resulting portal hypertension which is a transudative process. Protein concentrations in the transudate range have been reported in many patients where the disease process would be expected to be exudative, for example, malignancy and infection, and the converse has also been commonly reported. For example Gupta et al, 1995, reported that 24% of patients with uncomplicated cirrhosis had an ascitic TP concentration greater than 25 g/L, confirming previous reports. 15,16 This is discussed further in the section on TP.
In addition to these misclassifications, the peritoneal fluid of normal healthy women has been reported to have a TP content of greater than 40 g/L. 17,18 This would define it as an exudate which invalidates the whole transudate/exudate concept. It was therefore proposed more than 15 y ago that the transudate/exudate concept for the classification of ascites should be discontinued. 19
Collection of peritoneal fluid and initial evaluation, including non-biochemical testing
Collection
Diagnostic abdominal paracentesis is viewed as an essential investigation in all patients with clinically apparent new onset ascites. 3,4,20 The removal of fluid from the peritoneal cavity (paracentesis) is a safe procedure and can usually be carried out at the bedside by aspirating through a needle inserted in the flank or midline in the area of maximal dullness to percussion. 21 Ultrasound-guided paracentesis may be required when the volume of fluid is very small.
Paracentesis is usually carried out for diagnostic purposes, particularly to diagnose spontaneous bacterial peritonitis (SBP), but also in patients with suspected malignancy. Even where the cause is thought to be obvious from the clinical history and examination, a diagnostic paracentesis is considered mandatory. 3,4 In addition, therapeutic removal of larger volumes of fluid may be used to alleviate discomfort in massive ascites. Biochemical analysis of fluid obtained in this situation is not required and there is also doubt about the value of routine cell counts to diagnose SBP. 3,22
Peritoneal lavage for the diagnosis of blunt abdominal trauma is rarely performed now with the ready availability of CT scanning and such samples usually do not require biochemical analysis except, occasionally, for amylase.
Gross appearance
The gross appearance of the fluid can provide useful diagnostic information. The clear straw coloured appearance of most ascitic fluids reflects the most common aetiology, namely cirrhosis. 23 A blood-stained sample is usually due to a traumatic tap and in these cases the fluid tends to clot on standing. Samples which remain homogeneously blood stained throughout the tap could indicate malignancy, pancreatitis, TB, intestinal infarction or recent abdominal trauma. Blood-stained ascites is often thought to be a characteristic of peritoneal carcinomatosis, but an ascitic fluid RBC count of >10,000/mm3 was only found in 8.3% of peritoneal carcinomatosis patients and 22% of malignancy-related ascites overall. 24
It may be difficult to distinguish true turbidity, usually due to neutrophils and associated with bacterial peritonitis, pancreatitis or malignancy from the opalescence due to triglyceride seen in the fluid of 20–30% of cirrhotic patients. 25 In turn this pseudochylous ascites needs to be differentiated from true chylous ascites in which there is presence of chylomicrons and a high concentration of triglycerides. Measurement of triglyceride (see later) is therefore required to make this distinction, which is important since, in adults, the frequency of malignancy in true chylous ascites may be as high as 80%, making a search for occult malignancy imperative. 26
Occasionally ascitic fluid samples with striking appearances may be encountered, which may indicate the aetiology (tea coloured: pancreatic ascites; black: haemorrhagic pancreatitis or melanoma; dark molasses: perforated gut; green/brown: biliary tract disease).
Non-biochemical testing
Three non-biochemical tests, namely cell count, bacterial culture and cytology have an important role in the evaluation of ascitic fluid and it has been stated that ‘the ascitic fluid cell count is the single most helpful test to perform on ascites’. 1
Where SBP (spontaneous infection in the absence of an intra-abdominal source of infection which has a prevalence of between 10% and 30% in patients with ascites) 27,28 or secondary peritonitis is suspected, the white blood cell count and bacterial culture are of prime importance. The presence of >250 polymorphonuclear cells (PMN) per mm3 is diagnostic of this condition and indeed it is recommended that, in a clinical setting compatible with ascitic fluid infection, patients with a PMN count >250 mm3 should receive antibiotic therapy without waiting for culture results. 27 Automated counting of nucleated cells appears to be as reliable as manual counting and can be performed more rapidly. 28,29
Leukocyte esterase urine reagent strips to detect neutrophils in ascitic fluid could provide even more rapid bedside diagnosis of peritonitis in peritoneal dialysis patients and in cirrhotic patients with ascites. 30,31 Confirmation of this approach is required since there are contradictory reports of diagnostic sensitivity (50–100% depending both on which strips were used and the grading of positivity) and the strips are not licensed for this clinical indication. 32
The sensitivity of bacterial culture may be low, either due to low bacterial counts in the patient or due to inoculation of the culture medium in the laboratory rather than at the bedside. Conventional culture has been shown to detect bacterial growth in only approximately 40–60% of cases compared with a yield of 70–90% if inoculation is performed at the bedside. 33,34 The recommendations are therefore that ascitic fluid should be inoculated into blood culture bottles immediately the sample is obtained. It has been suggested that differentiation of SBP from secondary peritonitis may be helped by use of an algorithm including protein, lactate dehydrogenase (LD) and glucose with secondary bacterial peritonitis or perforation peritonitis being more likely when two out of three of the following were present: TP >10 g/L, glucose <2.8 mmol/L and LD > upper reference limit for serum. 35 Routine testing for TB is not recommended; the sensitivity of smear for mycobacteria is very low and that of fluid culture only 50%. 36
Peritoneal carcinomatosis accounts for two-thirds of patients with malignancy-related ascites and nearly all of these have positive cytology, due to the shedding of viable malignant cells from the peritoneal seedlings into the ascitic fluid. 11,24,37 By contrast, ascitic fluid cytology is positive in less than 10% of patients with hepatocellular carcinoma and is also usually negative in those with liver metastases. 24 With the exception of suspected ovarian malignancy, the emphasis placed on ascitic fluid cytology in the early assessment of ascites is felt to be misplaced since this has a sensitivity of only 60%, although 100% specificity. 38
Tests used in the biochemical analysis of peritoneal fluids
Methodological considerations
The use of methods optimized for measurement of analytes in serum, plasma or urine has already been discussed for pleural fluid. 2,39 Briefly, the major considerations include use of assays not validated for fluids other than serum or plasma (with the possible need for ‘CE’ marking), suitability of the sample, measurement at analyte concentrations outside the optimized range and lack of reference ranges. A more detailed discussion of the methodological aspects of albumin measurement is included in the section on the use of the albumin gradient.
Apart from the recommendation that ascitic fluid should be inoculated into blood culture bottles at the bedside, the guidelines do not comment on the specimen type required. Some authors have specified the use of a plain universal container for all tests including glucose, whereas others have recommended anticoagulated samples for cell counts and cytology if the sample is heavily bloodstained and likely to clot. 23,40
Original markers
Total protein
TP has for many years been the major criterion used in the differential diagnosis of ascites, although as already discussed, the premise upon which its use is based is probably flawed. 12–14 Not surprisingly therefore the literature abounds with papers where the use of TP has resulted in misclassification, but it should be appreciated that the aetiology of the ascites in many of these studies was not definitely proven.
For example, Sampliner and Iber 15 showed that 12% of unselected patients with chronic liver disease (CLD) had an ascitic fluid protein >30 g/L. A similar proportion (17%) was reported by Boyer et al., and moreover, in 12% the ascitic fluid protein concentration was ≥40 g/L. Conversely, two of 14 patients in this study with proven malignant ascites had a TP <30 g/L. Use of an ascitic fluid protein/serum protein ratio was of minimal extra value. 14
Gupta et al. 16 reported that an ascitic fluid TP <25 g/L had a 100% diagnostic specificity and positive predictive value for a cirrhotic effusion, but sensitivity was less good at 76% and the overall diagnostic accuracy was calculated at 88% (compared with 92% for serum-ascitic fluid albumin gradient of 11 g/L).
In the study of Rector and Reynolds, 41 patients with ‘transudative’ (chronic heart failure and cirrhosis) ascites had lower fluid TP concentrations, but there was considerable overlap with ‘exudative’ (largely malignancy) ascites, particularly in those with congestive cardiac failure. This may relate to diuretic use. 42
The majority of the variation in fluid TP concentration in patients with ascites due to CLD has been shown to be due to differences in the serum protein concentration and in portal pressure, and to be relatively independent of peritoneal permeability. 43 Thus, a transudative process will generate a relatively high ascitic protein concentration provided the blood oncotic pressure, determined chiefly by the albumin, is preserved. The converse is also true and low protein ascites may be found in ‘exudative’ ascites if the serum albumin is low. 41
Contrary to expectation, ascitic fluid protein does not increase during episodes of SBP and indeed those with the lowest protein concentration were found to be the most likely to develop SBP. 44
Lactate dehydrogenase
Early studies reported uniformly high LD activity in malignant effusions and low activity in non-malignant effusions. 45–47
In the studies of Boyer et al., the mean ascitic fluid LD activity (expressed in Sigma units [SU]) was much lower in patients with liver disease than in those with malignant disease (167 ± 9 versus 913 ± 228 SU). There was however considerable overlap, which was not reduced by use of the fluid/serum LD ratio. Nonetheless, once the LD activity exceeded 500 SU (serum reference range 220–580), malignancy, tuberculous or pancreatic ascites was always present (100% diagnostic specificity), but the sensitivity for distinguishing these from liver disease was low and malignancy was not excluded by a low LD. 14
A wide range of diagnostic sensitivities and specificities have however been reported for LD with one study suggesting that LD is one of the poorest discriminants. 48–50 This probably relates to patient groupings, possibly to methodology and to selection of cut-offs. It is not clear where the cut-off should be set since the absolute values given are method dependent and most reports do not give the normal serum reference range. The recommendation that a value of greater than 70% of the upper limit of the serum reference range may be discriminatory should be viewed with caution since this relates to one small study using old methodology. Overall, it does appear that a raised LD indicates a cause other than liver disease, but a low LD does not exclude malignancy.
Based on the approach of Light et al. 51 for pleural fluid classification, the value of combining LD with TP analysis has also been studied in ascitic fluid.The finding of two out of the three of the characteristics of an exudate (LD >400 SU, fluid/serum LD ratio >0.6 and fluid/serum TP ratio >0.5) tended to indicate a non-hepatic cause of the ascites. The absence of all three strongly suggested uncomplicated liver disease as the cause, combining the parameters made the diagnosis more difficult in some patients. 14
The LD isoenzyme pattern has been suggested to be helpful with higher LD-4 and -5 and lower LD-1 activity in malignant ascites than ascites from other causes. Although there were statistical differences in activity of the isoenzymes between the groups, there was a wide overlap so this seems unlikely to be of routine use. 52
Additional markers
Albumin gradient
Introduction
The failure of the traditional transudate/exudate concept in differentiating the aetiology of ascitic fluid has raised the question of adopting a more physiological approach, based on the presence or absence of portal hypertension. The ascitic fluid albumin gradient has been proposed as a way of assessing this. 41,53,54
Definition
The serum ascites albumin gradient (SAAG) is defined as the serum albumin concentration minus the ascitic fluid albumin concentration (not the ratio). It is thought that this parameter reflects the oncotic pressure gradient between the vascular bed and the ascitic fluid; an elevated gradient (greater than or equal to 11 g/L) usually being associated with increased portal pressure, whereas a low gradient (<11 g/L) is associated with conditions where the ascites is not related to portal hypertension. 41,53 This was demonstrated directly by Hoefs, 43 who showed that the SAAG correlated with the measured portal pressure in 56 patients with CLD (see Table 4).
Use of serum ascites albumin aradient (SAAG) in the differential diagnosis of ascites
*The classification of nephrotic syndrome is unclear and it has also been included in the category of high SAAG ascites 4
The literature on SAAG in the clinical setting is complex. As described above, initial reports determined the value of SAAG as an indicator of the pathophysiology and as an alternative to ascitic fluid TP and the traditional transudate–exudate concept.
The ability of this marker to classify patients with ascites based on the presence or absence of portal hypertension has subsequently been extended to encompass differential diagnosis in selected patient groups.
The first report by Paré 53 of SAAG in the differential diagnosis of ascites used the cut-off established by Hoefs of 11 g/L. 43 It should perhaps be noted that in an earlier paper by Hoefs, 42 the cut-off value was 10 g/L. Both groups proposed that SAAG was more effective than TP, LD or ratios of these tests in correctly identifying the aetiology (intrinsic liver disease from malignancy), but acknowledged that it probably could not establish the precise diagnosis since the same pathophysiological processes operate in malignant, tuberculous and inflammatory causes of ascites.
This theme was supported by Rector and Reynolds, 41 who suggested that SAAG should replace TP in the traditional classification of ‘transudative’ versus ‘exudative’ ascites, but also noted that it did not provide perfect discrimination.
In 1988 a study undertaken by Mauer and Mazione, 55 in a larger number of patients with more varied diagnoses confirmed and extended these observations. Their view was that SAAG best reflected elevated hydrostatic pressure, contributing significantly to the development of portal hypertensive ascites, even in patients with a low serum albumin concentration. They reported SAAG to be particularly helpful in distinguishing congestive heart failure with high TP from malignant ascites without liver metastases. However, in contrast to Paré, 53 SAAG was not superior to other markers in differentiating patients with CLD from those with malignant ascites.
Prieto et al. 48 also reported that SAAG was not the best discriminant of ascites due to liver disease and that caused by malignancy. Although they achieved a similar diagnostic accuracy (94%) with SAAG compared with that of Paré (95%), their view was that a sensitivity of only 80% made it an inappropriate single tool for the diagnosis of malignant ascites.
A large study of 285 patients in 1990 showed that a SAAG gradient less than 11 g/L was the single most accurate parameter for diagnosing malignant ascites (all patients had disseminated peritoneal malignancy). 56 The diagnostic efficiency of SAAG was quoted as 93%, but with a diagnostic sensitivity of only 62%. Nevertheless, it was concluded that SAAG should be the test of choice with the addition of an ascitic fluid PMN count to diagnose/exclude bacterial peritonitis.
In 1992, based on a study of 901 specimens collected over a six-year period from 330 patients, Runyon et al. 19 concluded that the exudate–transudate concept should be discarded in the classification of ascites and replaced by SAAG as a marker for portal hypertension. TP was however retained as a second-line test firstly to discriminate spontaneous from secondary bacterial peritonitis, secondly to predict those at high risk of SBP on the basis of a low ascitic TP and finally to suggest heart failure as a diagnosis in patients with high gradient ascites.
Other studies have supported the sole use of SAAG rather than traditional markers, using an algorithmic approach with established markers or in conjunction with less commonly measured parameters such as interleukin-1-α. 50,57,58 However, Chen et al. 59 reflected that although SAAG offered the best diagnostic accuracy (90.2%) and specificity (98.9%) for the diagnosis of malignancy, its sensitivity of 62.1% was too low.
In 1988 Runyon et al. 24 attempted to determine the utility of SAAG in differentiating the subtypes of malignancy-related ascites. Not surprisingly, due to the heterogeneous nature of the malignant diseases causing ascites, SAAG did not discriminate between these conditions any better than the other parameters. Indeed only one test, the serum alkaline phosphatase (ALP), was found to discriminate completely between the various patient subgroups and to separate outpatients with and without massive liver metastases.
Overall, these studies support SAAG as a marker of portal hypertension and the use of this index to replace the exudate–transudate concept in ascitic fluid. This approach has been adopted in the British and the American guidelines, both of which include SAAG in the initial testing strategy. 3,4 The message is less clear however in terms of any further discriminatory ability. This is due to a variety of issues including the number and types of patient included in the various studies; many did not reflect the patients in whom this test might be used. Other factors include a lack of consistency in evaluating the performance of the various parameters and the choice of analytical method to measure albumin.
SAAG methodology
There are a number of concerns about the albumin methodology used in these studies and hence the validity of the conclusions.
Many of the papers express only the albumin gradient and do not give the serum and ascitic fluid concentrations for individual patients. This is particularly important since the concentration encountered may be below the recognized dynamic range of the assays used. Details of the methodology and particularly the intra- and interassay coefficients of variation at the relevant concentrations are rarely given. The concentration of albumin in ascitic fluid varies depending upon the aetiology, but in patients with cirrhosis mean values are generally <15 g/L. External Quality Assurance schemes do not routinely assess assay performance at such low albumin concentrations.
Dye binding methods for measurement of albumin in serum and plasma are recognized to overestimate albumin at low concentrations with immunochemical methods recommended. 60 At low albumin concentrations bromocresol green (BCG) methods are pH dependent and particularly prone to interference from transferrin and lipoproteins, all of which may be relevant in patients with ascites. 61 One author has clearly stated that the BCG method is unsuitable for albumin determination in ascitic fluid and that results must be obtained using immunochemical methods. 62 The majority of studies have used BCG rather than a more specific method for measuring albumin and several reports do not state the method used. 53,59,62 Practically, most laboratories use dye binding methods for albumin and do not have access to immunochemical methods. This is borne out by participation in the UK NEQAS scheme for serum albumin measurement, the majority of labs using either BCG or bromocresol purple (BCP).
Other points to be considered are preanalytical factors that affect the serum albumin concentration, such as posture, prolonged tourniquet application, fluctuations in sick patients and the effect of diuretic use on the gradient. 42 The importance of simultaneous sampling of serum and ascitic fluid is unclear. 63
Other tests and specific situations where these may be useful
Glucose
Since glucose diffuses readily across membranes, the concentration in ascitic fluid will reflect the prevailing plasma concentration unless it is being metabolized by bacteria or fluid white blood cells. Decreased ascitic fluid glucose concentration has been reported in TB peritonitis, carcinomatosis and SBP, 64,65 and has been regarded as helpful in differentiating infection and malignant ascites from other causes. However, in early SBP the ascitic fluid glucose was shown to be similar to that in sterile fluid and overall glucose appears to be of low diagnostic sensitivity and specificity, limiting its routine use. 66
The one area where fluid glucose may be of use is in the differentiation of patients with tuberculous ascites where significantly lower concentrations are typical. 67 In one small study a fluid:blood glucose ratio of <0.7 gave 100% sensitivity and specificity. 50 The use of ascitic fluid glucose may be of more relevance in laboratories serving populations where there is a high prevalence of TB.
If the measurement of glucose is required then collection of the sample into a container with an appropriate preservative (fluoride oxalate) is recommended and use of a ratio is probably appropriate given the high prevalence of diabetes.
Amylase (is pancreatitis the cause?)
Pancreatic ascites is the accumulation of high protein, amylase-rich intraperitoneal fluid, which occurs during the course of chronic pancreatitis or is associated with the rupture of a pseudocyst or the disruption of a pancreatic duct. Extremely elevated ascitic amylase activity is characteristic. 68 Elevations above the serum reference range may be found in up to 90% of patients with acute pancreatitis or pancreatic pseudocyst. 69
It should however be noted that raised ascitic fluid amylase may also be found in patients with small bowel perforation, ischaemia or in mesenteric thrombosis, and so it is not totally specific.
In a more systematic study, Runyon 70 established that the amylase activity in ascites of non-pancreatic origin was generally about half the plasma value giving a mean ratio of 0.44 ± 0.33 (standard deviation) (range 0.10–1.55). Two patients in the study had pancreatitis and in both the ascites/serum ratio was >5.
The measurement of peritoneal lavage fluid amylase did not aid the diagnosis of pancreatic injury. 71 There have been reports of increased ascitic fluid amylase in a variety of malignancies, some, for example lung and ovarian associated with the S isoenzyme and pancreatic malignancies with the P isoenzyme. An unusual anodic migrating isoamylase has been reported in the ascites of a small number of patients with ovarian and gastric carcinoma. 72 Ascitic fluid lipase has also been suggested to be useful in the diagnosis of pancreatic ascites. 73
Tumour markers (is it malignant?)
A number of different proteins have been evaluated for their utility in diagnosing malignancy-related ascites. These include well-established ‘tumour markers’ such as AFP, CEA, CA19-9 and CA-125, as well as inflammatory markers such as ferritin, caeruloplasmin, α-2 macroglobulin, α-1-anti-trypsin, interleukins and other proteins such as fibronectin and complement. 74,75 These have often been looked at in combination with other parameters such as SAAG, LD and cholesterol and multivariate statistical methods used to improve diagnostic accuracy. 24,58,59,74 There are case reports of other tumour markers such as prostate-specific antigen. 76
Overall there appears to be little added value from measurement of the traditional tumour markers in ascitic fluid compared with measurement in serum and the same limitations with respect to using these for diagnosis apply. 59 Indeed in one group of patients with proven hepatocellular carcinoma the serum AFP was raised more frequently than the ascitic fluid AFP (72.7% versus 63.6%) and similar findings were reported by another group. 77,78 Also it should be noted that the most accurate (90%) and sensitive (87%) test for the diagnosis of ovarian cancer, and superior to serum CA-125 although both had similar specificity, was peritoneal fluid LD. 79
High concentrations of CA-125, CA19-9 and CA15-3 have been reported in the peritoneal fluid of women with endometriosis and other causes of pelvic pain and infection with only CA-125 being higher in the endometriosis group. 80 Perhaps not surprisingly, ascitic fluid CA-125 has been found to correlate closely with serum concentrations and so probably in this context provides no further information. 81 There has also been some interest in the use of fibronectin with higher concentrations in ascitic fluid from patients with peritoneal malignancy, but despite earlier encouraging reports this has not entered routine use. 49,58,82
There is current interest in the prognostic value of finding circulating tumour specific p53 DNA mutations in women with epithelial ovarian cancer. 83
Urea/creatinine (is it urine?)
Urea and creatinine measurements are used most widely to indicate the presence of urine in the peritoneal cavity. Creatinine is preferred since the urine plasma ratio is greater.
Adenosine deaminase (is it TB?)
Several reports have shown that ascitic fluid adenosine deaminase activity (ADA) is significantly greater in patients with tuberculous ascites compared with activity in those with cirrhosis, malignant disease and SBP, although high concentrations have been reported in fluids from patients with lymphoma and rheumatoid arthritis. 84
A recent meta-analysis of the four (out of 12) studies meeting the prespecified criteria of the meta-analysis (which included 264 patients) confirmed the high sensitivity (100%) and specificity (97%) using cut-offs from 36 to 40 IU/L. The authors strongly supported ADA determination as a fast and discriminatory test for peritoneal TB with an optimal cut-off value of 39 IU/L. 85 Indeed one small study of tuberculous effusions in children showed better sensitivity of ADA (81% versus 74%) albeit lower specificity (75% versus 88%) than polymerase chain reaction for mycobacterium TB. 86
Although ADA may be of value in areas with high prevalence of TB, it may not be helpful in low prevalence areas. A much lower sensitivity of 30% has been reported in a study from the US where more than 50% of the patients with TB had underlying cirrhosis. 87
Lipids (is this chylous ascites?)
Chylous ascites is a rare occurrence and is generally secondary to a malignant process. As discussed, it must be appreciated that a milky appearance does not necessarily indicate a chylous effusion. 25 Triglyceride should therefore be measured in all such effusions or chylomicrons should be demonstrated. Chylous ascites is defined on the basis of an ascitic fluid triglyceride concentration greater than 2.25 mmol/L and which is also greater than the corresponding serum concentration.
In adults, while malignancy has been reported to be implicated in up to 80% of true chylous ascites (mean triglyceride concentration 18.0 mmol/L: range 2.5–91.5), this definition also encompasses patients with cirrhosis and lesser elevations of ascitic fluid triglyceride (2.3–4.5 mmol/L) thought to be due to peritoneal lymphatic tears. 1,26
Cholesterol
A number of studies have reported that cholesterol measurement aids discrimination of malignant ascites from other causes with higher concentrations being found in patients with malignant peritoneal seedlings; suggested causal factors include increased vascular permeability, increased cholesterol synthesis and release from neoplastic cells. 16,88– 90
For example, Prieto et al. 48 showed that ascitic fluid cholesterol concentrations were significantly higher in patients with peritoneal metastases and was superior to ascitic fluid TP, LD and SAAG for discriminating ascites due to peritoneal metastases from that due to liver disease. There was however no difference in the ascitic fluid cholesterol between those with liver disease and those with superimposed hepatocellular carcinoma. Using a higher cut-off of 1.8 mmol/L compared with 1.2 mmol/L in the above study, another group suggested that cholesterol is the optimal test to distinguish ascites due to malignant disease from that due to liver disease with a 100% specificity and diagnostic efficiency of 94% compared with SAAG (86%), cytology (82%) and ascitic protein (70%). 90
Cholesterol and LD were found to be the two (of 11) most useful analytes for distinguishing malignant ascites from cirrhosis and hepatocellular-associated ascites. Additionally, by combining the log-transformed results of these two analytes and applying discriminant analysis, 100% sensitivity and specificity was achieved. 49
Bansal et al. 50 did not find such good discrimination, but still concluded that cholesterol was a useful tool for diagnosing malignant ascites. Contrastingly, a study of a similar patient group reported that cholesterol was less useful and it has also been demonstrated that cholesterol is not elevated in patients with massive liver metastases and conversely the patients with high protein non-cirrhotic ascites have high values. 91
Although it appears that ascitic fluid cholesterol is higher in those with peritoneal malignancy, all the studies can be criticized on the basis of patient selection and all have used different cut-offs, which lie at the lower end of the usual measuring range for cholesterol assay. Interestingly, most of the studies showed that ascitic fluid:serum cholesterol ratio decreased rather than improved diagnostic sensitivity and specificity. 48,50
Another consideration is whether the now widespread use of lipid-lowering agents would affect the previous findings.
Bilirubin
Although it has been reported that an ascitic fluid:serum bilirubin ratio of greater than 0.6 had a statistically significant association with fluids classified as an exudate, the addition of this to established criteria did not improve diagnostic accuracy and, given questions over validity of such a classification, seems to be of limited use 92 Although ascitic fluid bilirubin may not be useful in a general strategy, if a biliary leak is suspected then bilirubin measurement is indicated. An ascitic fluid value >103 μmol/L and which is also greater than the serum value is consistent with an intrahepatic or gallbladder fistula or upper gut perforation. 93
pH
An initial study showed a highly significant lowering of the ascitic fluid pH in six patients with SBP and complete discrimination from ascites of other causes when using a cut off between 7.32 and 7.38. 94 In other studies, low values were also found in patients with malignant ascites, TB peritonitis and pancreatic ascites, and discrimination between the groups was poor. A pH cut-off of <7.32 only identified 42% of the patients with SBP. 95 pH measurement is not useful in decisions about antibiotic treatment since it does not drop until the PMN count is in excess of 2000; cell counts being inexpensive and routinely available. 1
In a much larger study, although there were statistically significant differences in the pH of the ascitic fluid from patients with SBP compared with controls, no specimens had a pH of <7.35 and the sensitivity was therefore 0%. 96 The methodology used for pH measurement is, however, not stated and some studies have used pH indicator paper or pH meter, both of which would be considered inadequate. It has been concluded that measurement of pH added nothing to ascitic fluid analysis, except cost and confusion, and has no place in clinical decision-making. 1
Other analytes
Other peritoneal fluid analytes have been investigated, usually in very small numbers of patients and with inconsistent results. 50,58
Ascitic fluid ALP is increased in metastatic liver disease and peritoneal carcinomatosis, but does not provide any additional information compared with serum ALP. 24 ALP activity has been measured in peritoneal lavage samples and was found to be significantly increased in those with injury to the small intestine. 97 It was also increased in patients with perforation-related secondary peritonitis, as was CEA. 98 Increased ALP activity that was heat stable and confirmed as the placental isoform is reported in patients with intestinal endometriosis. 99
Although gamma glutamyl transferase has been reported to be increased in the ascitic fluid of patients with hepatoma, this has not been confirmed. 100,101
As for pH, measurement of ascitic fluid lactate has not been proven to be useful for the diagnosis of SBP, since although the concentration is increased, it is also high in malignant ascites, tuberculous ascites and pancreatic ascites. 95,96
Conclusions
As for pleural fluid, a wide range of tests and testing strategies have been proposed for the investigation of peritoneal fluid collections. Although at first sight the issues around the investigation of these fluids are similar, ascitic fluid often presents a more difficult diagnostic problem.
We have found the area of ascitic fluid investigation challenging to evaluate. The evidence base is difficult to interpret because of the use of different definitions, different patient groups, non-defined and non-standardized methodology, and the small numbers of patients included in the final analyses of many studies that are often a combination of retrospective and prospective data.
Many of the early studies focused on the classification of fluids into either a transudate or an exudate using TP. TP is routinely measured in laboratories with robust methodology, and it can be measured precisely and accurately within the required measuring range. Interpretation of results using the traditional transudate/exudate concept is however probably invalid and this is demonstrated by the poor diagnostic accuracy of the various studies.
Several authors have clearly stated that the use of TP should be abandoned in favour of SAAG, although it is still one of the three tests considered routine in the AASLD guidelines along with the SAAG and cell count. The British guidelines prefer albumin and allow for protein only if the former is not available.
Although logically because it is based on physiological principles, the albumin gradient is probably the investigation of choice and could perhaps replace TP measurement, there are a number of issues to be considered. Perhaps the most important of these is the methodology.
The methods used routinely in most laboratories for measuring serum albumin are not necessarily appropriate for measuring albumin in ascitic fluid and could lead to misclassification. Simultaneous timing of samples of serum and ascites is also important to ensure that the calculated gradient is an accurate reflection of the physiological state. 40
Irrespective of the methodology, SAAG does seem to discriminate consistently between most groups, perhaps because of its sound physiological basis. It is however a different approach to classifying effusions. There are particular issues with the diagnosis of malignancy, which may be associated with either high or low gradient ascites depending on aetiology. Clinicians however tend to think about cancer as a single entity. Peritoneal seedlings cause a low gradient ascites thought to be related to abnormal capillary permeability, whereas hepatic metastases cause a high gradient ascites due to intrahepatic congestion leading to portal hypertension.
SAAG is also of use in decision-making about therapy since those with a high gradient usually have portal hypertension and will respond to salt restriction. 102
Since the classification of ascites according to the transudate/exudate concept has been superseded, it makes LD redundant in this context. Measurement may however still have some clinical utility in identifying patients with malignancy as very high activities are almost always associated with this. 45–47 In this context it has been shown to be more useful than specific tumour markers, which at the moment seem to have little evidence to support their use but remain an active area of research.
There are situations in which other biochemical tests are of value. Their use should be restricted to specific clinical questions. This includes measurement of ascitic fluid amylase where pancreatic disease is suspected to be the cause of the ascites. A ratio of ascitic fluid:serum amylase is preferable since it eliminates the use of arbitrary ‘cut-offs’, which may not apply to individual laboratory methods; simultaneous sampling is important since rapid changes in serum amylase can occur.
For areas with a high prevalence of TB, measurement of glucose and ADA in ascitic fluid may provide supporting evidence for this diagnosis. This may change as more laboratories have access to newer molecular techniques for TB identification.
Creatinine measurement should be offered to identify a urinary leak into the peritoneal cavity. Triglyceride measurement is required for the diagnosis of chylous ascites. The only other biochemical test that seems to provide any additional information is that of bilirubin specifically where a biliary leak is suspected.
In our view there is insufficient evidence to recommend measurement of cholesterol, pH, lactate, tumour markers or other enzymes. The ASSLD Practice Guideline also states that these tests are unhelpful. 3
In conclusion, ascitic fluid analyses need to be interpreted in the light of the pretest probability of the likely cause of the ascites. Most ascitic fluid biochemical measurements only provide supporting and supplementary diagnostic information. Many diagnoses may be reached more effectively by traditional biochemical testing in serum, non-biochemical testing and other investigations such as imaging. It has been stated by several authors that the practice of ordering a battery of tests on every ascitic fluid specimen should be abandoned in favour of directed selective testing with the aim of establishing the aetiology of the ascites. 1,50 This approach is likely to be more efficient and cost-effective and has been adopted by both the American and British guidelines, but mandates close interaction between the clinician and the laboratory. 3,4
Recommendations
Both the American and British guidelines concur in the recommendation for cell count which we endorse. Ascitic fluid culture (with inoculation at the bedside) is also essential to exclude the presence of SBP.
In patients without an established cause for their ascites, the preferred biochemical investigation is the SAAG. Ideally samples should be collected simultaneously. Laboratories should ensure albumin measurements are optimized for this purpose, ideally using immunochemical methods or BCP if the former is unavailable.
Reports must include a description of the fluid's appearance, the serum and ascitic fluid albumin concentrations, as well as the calculated SAAG and an interpretative comment.
We believe that ascitic fluid TP measurement should be abandoned in favour of the SAAG, but may be retained in situations where the SAAG is unavailable.
There is a role for measurement of other analytes in specific circumstances such as amylase in cases of suspected pancreatic disease. These additional tests should only be performed after discussion with the relevant clinical team.
It is also important for laboratories to undertake a local risk assessment to address any health and safety issues related to handling potentially infective samples.
DECLARATIONS