Potential conflict of interest: Nothing to report.
Recent studies have shown that the diagnosis of spontaneous bacterial peritonitis (SBP) can be rapidly obtained using leukocyte esterase reagent strips. However, published studies were restricted to one or two centers, and the number of patients with SBP was thus limited. The aims of the current prospective multicenter study were: (1) to assess the diagnostic accuracy of the Multistix 8SG urine test for the diagnosis of SBP; and (2) to assess the prevalence of SBP. From January to May 2004, 2 reactive strips were tested independently in inpatients with cirrhosis and in outpatients undergoing paracentesis. Cultures of ascitic fluid were performed at the bedside using aerobic and anaerobic blood culture bottles. Two thousand one hundred twenty-three paracenteses were performed in 1,041 patients from 70 centers. One hundred seventeen samples, obtained from 91 patients, had ascites polymorphonuclear cell (PMN) counts ≥250/μl (range, 250–34,000), among which 56 were associated with positive ascitic fluid cultures. The prevalence of SBP was 5.5% in the whole population, 9% in inpatients, and 1.3% in outpatients (P < 0.0001). The prevalence of SBP was 0.57% in asymptomatic outpatients versus 2.4% in symptomatic outpatients (P = 0.04). Using a threshold of 2+ for positivity of the reagent strip, sensitivity was 45.3% for the diagnosis of SBP, specificity was 99.2%, positive predictive value was 77.9%, and negative predictive value was 96.9%. Conclusion: This study confirms the low prevalence of SBP in asymptomatic outpatients according to a priori defined criteria, and indicates an absence of diagnostic efficacy for this specific strip test. (HEPATOLOGY 2007;45:1275–1281.)
Spontaneous bacterial peritonitis (SBP) is a frequent and severe complication of cirrhosis.1–4 The prognosis of SBP has been improved by the use of antibiotics with high peritoneal diffusion and low nephrotoxicity and by the prophylactic use of antibiotics in high-risk groups of patients.5 This improvement in survival also might be explained by more rapid diagnosis and treatment, thereby avoiding the occurrence of severe sepsis and septic shock, a condition well known for its frequently fatal outcome.6 The diagnosis of SBP is based on detection of a polymorphonuclear neutrophil count equal to or greater than 250/μl with or without positive culture. However, obtaining an ascitic cell count within a few hours is sometimes difficult, and the clinician may decide to start an empiric antibiotic based on clinical or biological signs suggestive of infection.
Recently, the use of urinary reagent strips has been proposed for rapid diagnosis of SBP. The urinary strips identify leukocytes by detecting their esterase activity via a colorimetric reaction.7, 8
The use of Multistix strips (Bayer Health Care) has been tested for the diagnosis of bacterial meningitis,9 pleural infection,10 synovial infection,11 and peritoneal infection in dialyzed patients.12–14 For diagnosis of SBP, use of a reagent strip may be promising, because it has high sensitivity, between 85% and 100%, and high specificity, between 98% and 100%.15–19
However, published studies have been limited by the number of SBP episodes, and have been restricted to one or two centers.15–19 The authors of those studies suggested validating this technique in a larger group of patients with cirrhosis in a multicenter study. Moreover, the prevalence of SBP depends on the studied population. Indeed, the prevalence of SBP is low in ambulatory asymptomatic patients in whom therapeutic paracenteses are performed for refractory ascites.20–22 Thus, the accuracy of the test should be evaluated according to the different subgroups of patients.
The aims of this large prospective multicenter study were: (1) to assess the diagnostic accuracy of the Multistix 8 SG urinary screening test for the diagnosis of SBP in patients admitted for complications of cirrhosis or for therapeutic paracentesis as outpatients; and (2) to assess the prevalence of SBP in a multicenter prospective study in inpatients with cirrhosis with ascites and in outpatients with cirrhosis treated by therapeutic paracentesis for refractory ascites.
The study prospectively involved all consecutive patients with cirrhosis admitted between January and May 2004 to the 70 participating centers under the authority of the “Club Francophone pour l'Etude de l'Hypertension Portale” and the “Association Nationale des Hépato-Gastroentérologues des Hôpitaux Généraux de France.” Centers affiliated with the Club Francophone pour l'Etude de l'Hypertension Portale were academic hospitals and centers affiliated with the Association Nationale des Hépato-Gastroentérologues des Hôpitaux Généraux de France were primary referral hospitals.
Diagnosis of cirrhosis relied on clinical, biological, and morphological criteria. Patients were admitted either for treatment of ascites or for complications of cirrhosis (i.e., infection, gastrointestinal bleeding, hepatic encephalopathy, alcoholic hepatitis, acute renal failure, hepatocellular carcinoma), or were treated in an ambulatory setting. These outpatients were also included consecutively whether they presented with or without signs of SBP. Patients were defined as “symptomatic” according to the following criteria20: temperature above 38°C or below 36.5°C, chills, abdominal tenderness suggestive of peritonitis, developing or worsening hepatic encephalopathy, gastrointestinal bleeding within the last 15 days, acute renal failure (defined by an increase in the serum creatinine level to above 133 μl), or arterial hypotension (systolic arterial pressure below 80 mm Hg). These signs suggestive of infection were carefully searched for at admission by a senior physician. Exclusion criteria were: chylous ascites and ascites not related to portal hypertension (i.e., pancreatic ascites, peritoneal tuberculosis, peritoneal carcinomatosis). Patients with hemoperitoneum complicating hepatocellular carcinoma were also excluded.
The following data were recorded: age, sex, cause of cirrhosis (alcoholic, HBV, HCV, genetic hemochromatosis, other), the existence or not of prophylactic treatment by antibiotics, temperature, chills, systolic arterial pressure, abdominal pain suggestive of SBP, complications of cirrhosis, serum creatinine, prothrombin time, serum total bilirubin, serum albumin and platelet count, Child-Pugh score and grade.
Ascitic fluid was examined for leukocyte counts and polymorphonuclear cell (PMN) counts/μl, total protein, and bacteriological status. Ascitic fluid was centrifuged at 2500g for 10 minutes. A smear was stained with Giemsa. Total and differential cell counts were made with an optical microscope. Bacterial cultures were obtained by bedside inoculation of 10 ml of ascitic fluid into aerobic and anaerobic bottles (BacT/ALERT, Biomérieux, France).23
Each sample of the ascitic fluid was tested using Multistix 8 SG according to the technique described for urine by the manufacturer (Bayer Pharma SAS, Puteaux, France). All reagent areas were immersed in ascitic fluid, and the strip was removed immediately. At 120 seconds, the color of the leukocyte reagent was compared with the color chart on the bottle. A correlation between leukocytes and the colorimetric 5-grade scale (from 0 to 4) was suggested by the manufacturer as follows: grade 0, 0 leukocytes/μl; grade 1 (traces), 15 leukocytes/μl; grade 2 (1+), 70 leukocytes/μl; grade 3 (2+), 125 leukocytes/μl; grade 4 (3+), 500 leukocytes/μl.
The strip was considered positive at grade 3 (125 leukocytes/ml), because grade 4 is above the cutoff defining SBP. Moreover, most patients with SBP caused by Gram-positive cocci have a PMN count below the threshold of 250/μl.24
The strips were tested independently by each investigator and by a nurse, both of whom were unaware of the results of ascitic cytological examination at the time of reading.
All strips were read manually; in some cases, strips were read using an automatic reader (Clinitek 50, Bayer Diagnostics) in centers in which this automatic reader was available. When there was discordance between the 2 readers, the highest value was registered. For each patient, a test was repeated during the 5-month period of the study for a maximum of eight paracenteses.
The study protocol was approved by the local ethics committee of the University Hospital of Brest, France.
Continuous variables are expressed as mean ± standard deviation. For variables that were not normally distributed, median and ranges are given. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), accuracy and likelihood ratios of each reagent strip in the diagnosis of SBP were calculated. Likelihood ratios were calculated as follows: likelihood ratio for a positive test result (LR+) = sensitivity/(1 − specificity); likelihood ratio for a negative result (LR -) = (1 − sensitivity)/specificity. A 95% confidence interval for each statistic was calculated from binomial distribution.
Concordance between readings by the investigator and the nurse was evaluated using kappa statistics.
A total of 2,123 paracenteses were performed on 1,041 patients from 70 centers. Cirrhosis was mainly due to alcohol alone in 80.6% (n = 840) of the cases. Cirrhosis was related to alcoholism and chronic hepatitis B or C in 5% of cases, and to chronic viral hepatitis alone in 8.5% of cases. The main characteristics of patients are shown in Table 1. The median number of paracenteses per patient was 2.0 (range, 1–8). Seven hundred forty-eight patients (71.9%) were male and 293 (28.1%) were female. Mean age was 60 years ± 11.5 (range, 25–93).
Table 1. Characteristics of the 1041 Patients Who Underwent Diagnostic or Therapeutic Paracentesis
Male sex, n (%)
Age (years), mean (range)
Median number of paracenteses
Cause of cirrhosis, n (%)
Alcohol + viral hepatitis (B or C)
Child-Pugh score, n (%)
Prothrombin time (%), median (range)
Serum bilirubin (μmol/l), median (range)
Albumin (g/l), median (range)
Platelet count (109/l), median (range)
Serum creatinine (μmol/l), median (range)
A total of 238 samples were collected in inpatients and outpatients receiving primary or secondary prophylaxis with norfloxacine, and 351 samples in patients receiving other antibiotics for presumptive or documented bacterial infection.
A total of 1,147 samples were obtained from 686 inpatients and 976 samples from 355 outpatients; 581 patients (262 outpatients, 319 inpatients) were “asymptomatic” according to the following criteria (cf. supra): temperature above 38°C or below 36.5°C, chills, abdominal tenderness suggestive of peritonitis, developing or worsening hepatic encephalopathy, gastrointestinal bleeding within the last 15 days, acute renal failure, or arterial hypotension (systolic arterial pressure below 80 mm Hg). A total of 460 patients were defined as “symptomatic” when one or more criteria were noted. The prevalence of each specific symptom was as follows: fever > 38°C or hypothermia below 36.5°C (n = 317), chills (n = 54), abdominal tenderness (n = 130), hepatic encephalopathy (n = 177), gastrointestinal bleeding (n = 140), acute renal failure (n = 147), and arterial hypotension (n 41).
A total of 117 samples obtained from 91 patients had SBP. The PMN count ranged from 250/μl to 34,000/μl. The PMN count was below 1,000/μl in 50 samples.
These 117 SBP cases were associated with positive ascitic fluid cultures in 56 cases (47.9%).
Table 2 shows the sensitivity, specificity, PPV, NPV, likelihood ratio for a positive test, likelihood ratio for a negative test, and 95% CI when we considered a reagent strip positive with grade 3.
Table 2. Sensitivity, Specificity, Positive Predictive Value, Negative Predictive Value, and Likelihood Ratios of the Multistix 8 SG Reagent Strip in the Diagnosis of Spontaneous Bacterial Peritonitis
Briefly, in the overall population, sensitivity was 45.3%, specificity 99.2%, PPV 77.9%, and NPV 96.9%. When considering symptomatic patients, whether inpatients or outpatients, sensitivity was 48.1%, specificity was 98.6%, PPV was 80.6%, and NPV was 94.1%. For inpatients without a priori clinical or biological signs of infection,20 sensitivity fell to 16.7%, specificity was 99.2%, PPV only 33.3%, and NPV 98.0%. Considering asymptomatic outpatients, sensitivity was 25%, specificity 100%, PPV 100%, and NPV 99.6%.
Exact concordance between the 2 readers (a doctor and a nurse) was high (K = 0.805; P < 0.0001). A strong correlation was seen between ascitic fluid total leukocyte count and PMN (Fig. 1). Three hundred sixteen samples were read using an automatic reader. In 18 samples, the PMN count was above 250/μl. The correlation between ascitic leukocyte count and results obtained by automatic reading was good (Spearman ρ: 0.719; P < 0.0001). The diagnostic performances of the test according to the different subgroups of patients are shown in Table 2.
The prevalence of SBP using cell count (i.e., PMN ≥ 250/μl) was 9% in inpatients and 1.33% in outpatients, including patients with a previous diagnosis of SBP and patients on antibiotics. In the ambulatory setting, the prevalence of SBP was 0.57% in asymptomatic outpatients. The prevalence of SBP in these different settings was higher in patients with a previous diagnosis of SBP (data not shown). Among 4 patients who had an absolute neutrophil count of ≥250/μl, one was detected positive with 3,000 PMN/μl and a positive culture for Streptococcus. For the other three patients, the strip was negative with 1290, 308, and 504 neutrophils/μl, respectively. Culture of ascitic fluid was negative in all cases. One patient received empiric antibiotic therapy. The other 2 patients did not take the antibiotics prescribed, and they were seen 1 and 2 months later without any symptoms.
Our study confirms that the specificity of reagent strips is very high in the diagnosis of SBP. However, the sensitivity of the Multistix 8 SG test was low in this large national multicenter prospective study involving patients with cirrhosis followed both in academic hospitals and in primary referral hospitals.
Using the threshold of 125 leukocytes/μl, sensitivity was only 45.3%, which is much lower than the sensitivity reported in previous studies.15–19 Moreover, the cutoff of the strip that we defined for detection of SBP was low: indeed, the next grade of the Multistix strip corresponded to 500 leukocytes/μl.
Several explanations are possible for this poor sensitivity. First, published studies were limited to a small number of patients with SBP. According to the 95% CI, in a larger population of patients with SBP, there might be a high rate of false-negative tests.19 Moreover, those studies were limited to one or two centers, making extrapolation to real-life situations difficult.18, 19 In the study performed by Butani et al.,15 two samples for which the leukocyte esterase test were negative had the 2 lowest absolute PMN values, 1088 and 368 PMN/μl, respectively. The authors suggested lower sensitivity of the test in a weakly positive setting. In our study, the larger number of samples with low PMN values below 1000/μl may account for the lower sensitivity reported thus far.
Second, the strip was initially designed for detection of urinary tract infections in which the number of leukocytes is significantly higher than in SBP. In initial reports, the sensitivity of the reagent strips ranged from 85% to 100%. The sensitivity of the test may vary according to the setting. Meta-analysis showed that the sensitivity of leukocyte esterase tests in the diagnosis of urinary tract infections ranged from 56% in an emergency setting to 87% for the family physician.25 Even in this situation, when the strip is negative and there is a high probability of urinary infection, the clinician relies on clinical signs. In the case of SBP, infection is associated with high morbidity and a high mortality rate. Therefore, delaying the start of antibiotics based on a negative result of the test would be deleterious. The strip may be helpful for the clinician when the test is positive, as the specificity of the test is very high. Indeed, in our study, specificity was 99.2%. Conversely, a negative result cannot exclude the diagnosis of SBP, particularly in symptomatic patients, in whom sensitivity was only 48.1%.
We chose to use the Multistix 8 SG in this study for several reasons: previous studies had shown its excellent sensitivity,15, 16 and this test is widely available in public institutions in France. Nonetheless, other reagent strips can be used with greater accuracy. Castellote et al.17 used Aution sticks manufactured in Italy and observed 89% sensitivity. However, in that study, some patients had ascites unrelated to portal hypertension, that is, carcinomatous ascites with a high ascitic leukocyte count. To our knowledge, no study had yet compared Aution sticks and Multistix 8 SG strips. Sapey et al.19 compared the Multistix 10 SG and the Nephur test and showed that the Nephur test was more sensitive (88.2% vs 64.7%). The Combur test, which is a modified version of the Nephur test, has also been recently compared with the Multistix in 2 studies.18, 26 The sensitivity of the tests was identical in the first study,19 whereas the Combur test was more sensitive than the Multistix test in the second when using the threshold of grade 2 on a colorimetric scale (63.0 versus 45.7%).26 Our results are in accordance with the latter study. Indeed, we defined the threshold of positivity with grade 3 of the test (125 PMN/μl). The sensitivity of the test was much higher when using grade 2 or grade 1 (data not shown), but specificity decreased slightly. Results of diagnostic performances in the main published studies including ours are presented in Table 3. Thus, some tests probably outperform Multistix 8 SG. Other strips should be tested and developed.
Table 3. Sensitivity, Specificity, Positive Predictive Value, and Negative Predictive Value in Published Studies on the Diagnosis of SBP Using Different Reagent Strips
However, in terms of the severity of SBP, the rate of false-negative results remains high. These tests may help the clinician in some circumstances when a cell count is not available within a few hours. Reagent strips may be useful in developing countries without sufficient resources. The cost of the strip is only 0.15 euros. Our results do not support systematic replacement of standard ascitic fluid analyses by the use of reagent strips at this time.
Although the concordance between the investigator and the nurse was excellent, it was not total. We suggest performing 2 tests to increase sensitivity, because interpretation of the test may depend on the reader and on environmental factors such as room lighting. We expected that automatic reading would increase the performance of the test. Unfortunately, we did not observe greater accuracy with the automatic reader. A few false-negative results were observed with automatic reading in patients with SBP and a positive strip, as noted by the investigator.
Our study confirms the low prevalence of SBP in asymptomatic outpatients (0.57%) in a population chosen based on previously well-defined clinical criteria.20–22 In three previous studies with a number of paracenteses ranging from 117 to 427, the prevalence of SBP was low (0%–3.5%).20, 21 In addition, Runyon27 recently reported a 2% prevalence of SBP in a series of 400 paracenteses performed in 2 years in an outpatient setting. The results of the present study are comparable to those reported by Evans et al.21 and by Romney et al.22 In their retrospective study of 427 patients with cirrhosis seen in a single outpatient clinic, Evans et al.21 analyzed 427 exploratory paracenteses performed over a 6-year period. Their exclusion criteria were similar to those used by Jeffries et al.,20 but patients receiving primary or secondary prophylaxis with norfloxacin were excluded. For Evans et al.,27 the prevalence of SBP was 1.4% and the prevalence of neutrocytic ascites was 2.1% (giving a combined prevalence of 3.5%). In the prospective multicenter study of Romney et al.,22 the prevalence of SBP was nil. The results of these different studies in outpatients without suspected infection are thus similar, with a prevalence of SBP varying from 0 to 3.5% (Table 4). The higher prevalence of SBP was reported by Evans et al.21 and could be explained by less restrictive a posteriori exclusion criteria and possible inclusion of infected patients. Moreover, in the study by Evans et al., patients receiving norfloxacin were not included. In our multicenter prospective study, the prevalence of SBP in asymptomatic outpatients was 0.57%, confirming a very low prevalence of SBP in this setting.
Table 4. Prevalence of SBP in Asymptomatic Outpatients
In summary, our study confirms the excellent specificity of Multistix 8 SG strips in the diagnosis of SBP, but reveals the poor sensitivity and positive predictive value of the test.
Therefore, it should not systematically replace standard ascitic fluid analyses because of its weak sensitivity and because it cannot rule out SBP. Other reagent strips may be more accurate but still lack sensitivity. Routine cytological examination remains mandatory for the diagnosis of SBP.
We thank LFB Laboratories (Laboratoire Français du Fractionnement et des Biotechnologies, France) for their support, Dr. Christophe Bessaguet for statistical help, and Jerri Bram for revision of the manuscript.
Participating Clinical Centers
Abbeville: J. Butel, Aix en Provence: C. Wartelle, M. Picon, J. Lafon; Amiens: D.Capron, E. Nguyen-Khac; Angers: F. Oberti, P.Cales; Arles: G. Boulay; Aulnais sous Bois: G. Bellaïche, JL. Slama; Auxerre: B. Champigneulle; Avignon: JP. Arpurt, Dr Bellon, Dr Bramli, Dr Coulibaly; Besançon: S. Bresson-Hadni, C. Vanlemmens, V. Di Martino, JP. Miguet; Bondy: P. Nahon, N. Ganne-Carrié, JC. Trinchet, M. Beaugrand; Bordeaux Pessac: X. Adhoute, V. De Lédinghen, P. Couzigou; Bordeaux Saint André: PH. Bernard, C. Balabaud; Bourg en Bresse: D. Pillon; Brest: JB. Nousbaum, F Cholet, H. Gouérou; Cambrai: T. Thévenot; Chalons en Champagne: N. Abdelli; Chalons sur Saône: P. Bernard, A. Soupison; Chambéry: F. Bourrhis, M. Jego; Châteauroux: T. Sapey, E. Fort; Cherbourg: H. Bertrand; Clermont-Ferrand: A. Abergel, C. Bonny; Clichy: R. Moreau, D. Lebrec, F. Durand, F. Degos, D. Valla; Clichy (Gouin): S. Levy, C. Halimi; Creil: JF. Cadranel, V. Jouannaud, P. Lemaitre, D. Belloula, NE. Farouj; Dunkerque: T. Paupard, Dr Belhous; Eaubonne: M. Howaizi; Evry-Corbeil: B. Wisniewski, D. Labayle, D. Fischer, J Denis; Foix: JM. Dramard; Fourmies: P. Bohon; Gonesse: A. Pauwels, A. Medini; Jolimont: J. Henrion; Le Kremlin Bicêtre: M. Gelu-Siméon, G. Pelletier, C. Buffet; Le Mans: C. Pilette, B. Bour, A. Blanchi; Lens: T. Davion, C. Becker, Dr Touze, Dr Djédir; Lille: P. Mathurin, V. Canva-Delcambre, JC. Paris; Lomme: D. Lucidarme, B. Filoche; Lourdes: JJ. Meurisse; Marseille Conception: D. Botta-Fridlund; Marseille Saint-Joseph: P. Castellani, M. Bourlière; Metz: JJ. Raabe, JM Perarnau; Montargis: Dr Blondon; Montauban: JL. Payen, L. Escudié; Montélimar: B. Nalet, F Belouineau; Montfermeil: B. Lesgourgues, S. Nahon; Morlaix: A. Gourlaouen, P. Loudu; Nancy: D. Ancel, JP. Bronowicki; Narbonne: E. Vaucher, L. Billès; Nice: G. Vanbiervliet, A. Tran; Paris (Pitié Salpétrière): D. Thabut, T. Poynard; Paris Saint Antoine: N. Carbonnel, L. Serfaty, T. Andréani, O. Chazouillères, R. Poupon, A. Poujol-Robert; Paris Tenon: X. Amiot, JD. Grangé; Pau: A. Pariente; Ploermel: M. Besseau, JM. Rotty; Poissy: RL. Vitte, C. Eugène, S. Beaulieu; Poitiers: C. Silvain, C. Chagneau, M. Beauchant; Pontivy: J.A. Seyrig, J. Abdulahad; Pontoise: O. Danne, P. Hervio; Saint Brieuc: O. Nouel; Saint Denis: H. Labadie, P. Periac; Saint Etienne: JC. Audigier; Sarrebourg: H. Jouin; Senlis: A. Menzel; Strasbourg: F. Habersetzer, D. Vetter, M. Doffoel, Dr Ananna; Tarbes: T. Morin, I. Kamran; Toulouse Purpan 1: C Bureau, JM. Peron, K. Barange, JP. Vinel; Toulouse Purpan 2: L. Alric, M. Duffaut, S. Thebault; Valenciennes: A. Boruchowicz; Vannes: D. Grasset; Villejuif: B. Roche, JC. Duclos-Vallée, D. Samuel; Villeneuve Saint Georges: L. Bettan.