Scandinavian Journal of Gastroenterology. 2014; 49: 595–603
ORIGINAL ARTICLE
Scand J Gastroenterol Downloaded from informahealthcare.com by Universitaet Zuerich on 07/09/14 For personal use only.
Prediction of spontaneous bacterial peritonitis in cirrhotic ascites by a simple scoring system
MALTE H. WEHMEYER, SARAH KROHM, FRIEDERIKE KASTEIN, ANSGAR W. LOHSE & STEFAN LÜTH I. Department of Medicine, University Medical Center Hamburg-Eppendorf, Hamburg, Germany
Abstract Background and aims. Spontaneous bacterial peritonitis (SBP) is a life-threatening complication in patients with liver cirrhosis. The aim of this prospective study was to identify predictors of SBP in order to develop a noninvasive method to identify or exclude an episode of SBP. Patients and methods. Three hundred and ninety-two consecutive patients, who underwent paracentesis from March 2008 through January 2012 in our department due to cirrhotic ascites, were screened. Ninety-six patients were excluded, mostly due to prior application of antibiotics. SBP was defined by an absolute neutrophil count ‡250 cells/mL ascites. We evaluated various clinical and laboratory parameters as potential predictors of SBP. A scoring system was developed in a training set of 220 and validated in a second set of 76 patients. Results. Fifty-eight patients (26%) in the training set and 17 patients in the validation set (22%) suffered from SBP. Thrombocytopenia £100,000 cells/mL, age >60 years and CRP >60 mg/L were identified as independent predictors of SBP. A scoring system combining these three parameters (weighting thrombocytopenia and age with 1 point each, but CRP with 2 points) reaches a positive predictive value for the diagnosis of SBP of 81.8% with a specificity of 98.8% (score ‡3). The negative predictive value at a threshold of 1 point is 93.5% with a sensitivity of 87.9%. Notably, a high MELD score is not associated with SBP (p = 0.3344). Conclusions. Combination of age, CRP and platelet count in a simple scoring system helps in the rapid diagnosis or exclusion of SBP.
Key Words: platelets, predictors, SBP, thrombocytes
Introduction Spontaneous bacterial peritonitis (SBP) is a frequent and life-threatening infection in patients with ascites due to liver cirrhosis, first described in the 1960s [1]. The prevalence of SBP ranges from 10% to 30% in hospitalized patients [2–5] and is ~3.5% in asymptomatic outpatients [6] suffering from cirrhotic ascites. The diagnosis of SBP has immediate therapeutic consequences, such as the start of an efficient antibiotic treatment and should not be missed or delayed, therefore. Translocation of bacteria from the intestinal lumen into mesenteric lymph nodes with colonization of blood, lymph, and ascites is the key event in the pathogenesis of SBP [7–10]. There are three
predisposing factors for SBP: First, patients with an advanced cirrhosis display a decreased mucosal barrier function against intestinal bacteria [7,8]. Second, bacterial overgrowth [11,12] is a common result of a disturbed intestinal motility in cirrhotic patients [13,14]. Third, reduced immune function in patients with liver cirrhosis permits an uncontrolled bacterial growth in ascites [15–17]. As a consequence of bacteremia and bacterascites vasodilatation may lead to renal failure [18–20]. Clinical risk factors for developing SBP, such as history of prior SBP [21] or variceal hemorrhage [22] and low ascitic fluid protein levels [2,6,23] are well known. However, available data concerning the prediction of SBP by routine laboratory tests are contradictory. An association between hyperbilirubinemia
Correspondence: Stefan Lüth, MD, I. Department of Medicine, University Medical Center Hamburg-Eppendorf, Martinistr. 52, 20246 Hamburg, Germany. Tel: +49 40 7410 58941. Fax: +49 40 7410 58531. E-mail: [email protected]
(Received 9 August 2013; revised 5 September 2013; accepted 20 September 2013) ISSN 0036-5521 print/ISSN 1502-7708 online 2014 Informa Healthcare DOI: 10.3109/00365521.2013.848471
596
M. H. Wehmeyer et al. 392 patients Exclusion criteria: Receiving antibiotics (n = 82), hemorrhagic or malignant ascites (n = 14).
Scand J Gastroenterol Downloaded from informahealthcare.com by Universitaet Zuerich on 07/09/14 For personal use only.
296 patients
96 patients excluded
Training set: 03/2008– 05/2010 (n = 220)
Validation set: 03/2011– 01/2012 (n = 76)
58 patients with SBP (≥ 3 billion neutrocytes/L)
17 patients with SBP (≥ 3 billion neutrocytes/L)
Culture positive 12%
CNNA 88%
Culture positive 25%
CNNA 75%
Figure 1. Study design and results. Period of time in which the two different sets of patients were collected is given, too. [n = number; SBP = spontaneous bacterial peritonitis; CNNA = culture-negative neutrocytic ascites].
and SBP has been demonstrated repeatedly [21,23– 26], but also reduced prothrombin activity [21,23,26], aspartate transaminase (AST) [23], C-reactive protein (CRP) [27], leukocytosis [26,27], serum creatinine [26], Child-Pugh score [24], MELD score [25] and platelet count [24,27] have been shown to be associated with an episode of SBP. However, either these studies contained only a relatively small collective [21,23] of patients, were designed to identify long-term risk factors for SBP [21,23,24,26] or were designed as a retrospective study [25]. A rather complex scoring system consisting of creatinine, bilirubin, thrombin time, and leucocyte count for risk stratefication for SBP was recently developed by Shi et al. [26] The aim of this prospective study was to identify clinical and laboratory parameters that predict SBP in patients with cirrhotic ascites and to develop a simple scoring system which allows a rapid diagnosis of SBP or to exclude SBP without the need for instant paracentesis. Patients and methods Study population Diagnostic with or without therapeutic paracentesis was performed in 392 patients admitted to the I. Department of Medicine of the University Medical Center Hamburg-Eppendorf with cirrhotic ascites from March 2008 through January 2012.
Our training set consisted of 256 patients. Of these, 36 patients were excluded from the study due to antibiotic treatment at the time of paracentesis (n = 32) or hemorrhagic ascites (n = 4). Two other exclusion-criteria (malignant ascites and evidence for secondary peritonitis) were not met by any of the remaining patients in the training set. A total of 220 consecutive patients were included in the training set. To verify our findings, we further screened 136 patients, of which a total of 60 patients were excluded (50 patients receiving antibiotics at the time of paracentesis and 10 showed evidence for hemorrhagic or malignant ascites). Thus 76 patients could be recruited for the validation set. The study design is summarized in Figure 1. Cases and controls A neutrophil cell count in the ascitic fluid >250 cells/m L defined the diagnosis of SBP regardless of the results of ascitic culture. Therefore, patients with culture-positive and culture-negative neutrocytic ascites (CNNA) were considered as SBP positive. At the time of paracentesis (+/- 1 day) laboratory and clinical data were assessed and used for analysis. Patients with a neutrophil cell count 60 years History of SBP Short-time history of variceal bleeding Etiology of Cirrhosis Alcohol Hepatitis C ± alcohol NASH/cryptogenic Autoimmune ± alcohol Hepatitis B ± alcohol Other Creatinine [mg/dL] Bilirubin [mg/dL] Bilirubin >5 mg/dL INR MELD score Child-Pugh stadium Child B Child C Albumin [g/L] CRP [mg/L] CRP >60 mg/L Platelet Count [103/mL] Platelets £100,000/mL
SBP present (n = 58) Mean ± SD, median (range), or N (%) 40 (69.0%) 60.2 ± 8.9 33 (56,9%) 5 (8.6%) 6 (10.3%)
SBP absent (n = 162) Mean ± SD, median (range), or N (%) 105 (64.8%) 56.5 ± 10.2 63 (38.9%) 13 (8%) 17 (10.5%)
0.5657 0.0164* 0.0177* 1 1 0.5839 0.7913 0.5925 0.6070 0.5916 0.3294 0.7265 0.2180 0.0092* 0.1729 0.3344 0.8415
35 13 4 3 2 1 1.12 2.10 20 1.38 16
(60.3%) (22.4%) (6.9%) (5.2%) (3.5%) (1.7%) (0.5–5.8) (0.2–43.0) (34.5%) (0.94–2.18) (7–37)
91 39 11 9 5 7 1.20 2.05 29 1.34 15
(56.2%) (24.1%) (6.8%) (5.6%) (3.1%) (4.3%) (0.4–7.6) (0.2–36.4) (17.9%) (0.90–5.00) (6–40)
37 21 28 25.5 14 110 25
(63.8%) (36.2%) (18–38) (5–131) (24.1%) (24–481) (43.1%)
101 61 27 17.5 7 130 46
(62.4%) (37.7%) (12–47) (5–202) (4.3%) (32–456) (28.4%)
Statistical analysis Continuous variables with an assumed Gaussian distribution (age) were compared between cases and controls via t-test. Continuous variables considered nonnormally distributed (MELD-Score, creatinine, bilirubin, INR, platelet count, albumin, CRP) were compared using rank-sum test. Receiver–operator characteristic (ROC) curves were calculated to define the most discriminating “cutoffs” for measured data, thus identifying dichotomous variables associated with SBP. p-Values for dichotomous variables (etiology, sex, Child-Pugh stadium, laboratory values and age at specific cutoff) were calculated using Fisher’s exact probability test. A p-Value < 0.05 was considered statistically significant. To identify independent predictors of SBP, variables being significantly correlated with the risk for SBP in univariate analysis were entered in a multivariate model using stepwise logistic regression. Scoring points based on estimated coefficients derived from the regression model were assigned to independent predictors of SBP and added up for each patient. The diagnostic value of this scoring system was then assessed by the area under the ROC curve. Findings were validated in our validation set.
p-Value
0.4898 0.0055* < 0.0001* 0.2109 0.0397*
MELD-score and Child-Pugh stadium were calculated using the established formula. The study was approved by the local ethics committee. All patients gave informed consent. Statistical analyses were performed using SPSS Version 19; the figures were created with GraphPadPrism 4. Results Patients characteristics Fifty-eight out of 220 patients (26%) in the training set and 17 out of 76 patients (22%) in the validation set were diagnosed with SBP. Patient characteristics of our training set are displayed in Table I. In the validation set significantly more patients suffered from cirrhosis due to alcohol abuse and significantly less patients suffered from HCV-related cirrhosis (p = 0.0062 and 0.0040). Otherwise, there were no significant differences between the two sets (see Table II for details). Twenty-one patients with SBP (36.2%) and 61 individuals in control-group (37.7%) had a Child-C-cirrhosis (p = 0.8415). All other patients were classified as Child-B-cirrhosis, while no patients with a ChildPugh stadium A were found in the study population (Table I).
598
M. H. Wehmeyer et al.
Table II. Comparison of clinical data between the training and the validation set. [SD = standard deviation; n = number; SBP = spontaneous bacterial peritonitis; NASH = nonalcoholic steatohepatitis; * p < 0.05]. Variable
Scand J Gastroenterol Downloaded from informahealthcare.com by Universitaet Zuerich on 07/09/14 For personal use only.
Male Gender Age [years] Prevalence of SBP Etiology of Cirrhosis Alcohol Hepatitis C ± alcohol NASH/cryptogenic Autoimmune ± alcohol Hepatitis B ± alcohol Other
Training set: March 2008 to May 2010 (n = 220) mean ± SD, or N (%)
Validation set: March 2011 to January 2012 (n = 76) mean ± SD, or N (%)
p-Value
145 (65.9%) 57.5 ± 10.0 58 (26.4%)
49 (64.5%) 59.1 ± 9.1 17 (22.4%)
0.8888 0.3064 0.5430
126 52 15 12 7 8
(57.3%) (23.6%) (6.8%) (5.5%) (3.2%) (3.6%)
Thirteen patients in the control group (8%) and five patients in the SBP group had a history of SBP (8.6%). Seventeen patients in the control group (10.5%) and six additional patients in the SBP group (10.3%) presented with a short-time history of variceal hemorrhage. Thirty-three ascitic fluid cultures were obtained in the SBP group (56.9%) with four positive results (12.1%) for two gram-negative and two Grampositive bacteria respectively. 61 cultures were performed in the control group (37.7%), diagnosing monomicrobial nonneutrocytic ascites in two individuals (3.3%; one Gram-positive and one Gramnegative bacteria, respectively). There was no patient with polymicrobial ascites infection. Ascitic fluid culture was conducted relatively more often in patients with SBP (p = 0.0110). Age, elevated CRP and thrombocytopenia are associated with the presence of SBP Univariate analysis identified significant differences between cases and controls only for the continuous variables age and CRP (p = 0.0164 and 0.0032, respectively). A platelet count considered as a dichotomous variable at a “cutoff” of £100,000/mL predicted SBP with a sensitivity of 43.1% and a specificity of 71.6% (p = 0.0397). The ROC curves for serum CRP indicated a sensitivity of 25.5% and a specificity of 96.5% for SBP at >60 mg/L (p < 0.0001). The area under the curve (AUC) was 0.6358. An age >60 years predicted the presence of SBP in our population with a sensitivity of 56.9% and a specificity of 61.1% (p = 0.0177). The AUC was 0.6054. The most discriminating “cutoff” identified by ROC-curve for bilirubin was a level >5 mg/dL (p = 0.0092). Details are depicted in Figure 2.
57 6 4 3 3 3
(75.0%) (7.9%) (5.3%) (4.0%) (4.0%) (4.0%)
0.0062* 0.0040* 0.7894 0.7671 1 1
Potential thresholds for creatinine, INR and albumin that were generated the same way by ROC curves did not discriminate between SBP patients and controls (data not shown). The best “cutoff” for MELD-score identified via ROC at >23 was not able to differentiate between cases and controls, too (p = 0.0680; Figure 2). A simple scoring system identifies patients with SBP Platelets, age, CRP and bilirubin were entered in a stepwise logistic regression model. Only platelets £100,000/mL (odds ratio (OR) 2.55, 95% CI 1.28– 5.07, p = 0.008), age >60 years (OR 2.72, 95% CI 1.39–5.34, p = 0.004) and CRP >60 mg/L (OR 9.79, 95% CI 3.46–27.66, p < 0.001) were identified as being independently affiliated with SBP. Diagnostic fidelity was improved when combining these parameters in a scoring system with age >60 years and platelets £100,000/mL each scoring one point (since the estimated coefficient for age was 1.0 and the coefficient for platelets was 0.9 in logistic regression) and CRP >60 mg/L scoring two points (the estimated coefficient in logistic regression was 2.3; “model A”). An overview of the scoring system is displayed in table III. At a “cutoff” of 1 scoring-point the positive predictive value (PPV) was 45.5% and the negative predictive value (NPV) was 93.5%. Thus in 93.5% of the patients with a scoring of 0 points a SBP was excluded correctly. Using a “cutoff” of 1 point led to false negative testresults in 7 out of 58 patients with SBP in the training set (sensitivity 87.9%). At a “cutoff” of 3 points PPV was 81.8% with a NPV of 76.6% and all patients with a score of 4 were suffering from SBP (PPV 100% at a “cutoff” of 4). Two patients in the control group displayed a scoring of 3 points (specificity 98.8%). Further statistical values from both the training and validation set are detailed in table IV.
A scoring system for rapid diagnosis of SBP B
A
90
220 200
80
180
70
160
60
140
Age [years]
CRP [mg/L]
599
120 100 80
50 40 30 20
40
10
20 0
0 SBP present
SBP absent
SBP present
SBP absent
SBP present
SBP absent
D
C
50
550 500 450
40
400 350
MELD score
Platelets [103/mL]
Scand J Gastroenterol Downloaded from informahealthcare.com by Universitaet Zuerich on 07/09/14 For personal use only.
60
120 250 200
30
20
150 10
100 50 0
0 SBP present
SBP absent
Figure 2. CRP [A], age [B], platelet count [C] and MELD score [D] in cases and controls (training set). The most discriminating “cutoff” for the prediction of SBP is marked for each variable (A to C: p < 0.0001, p = 0.0177 and p = 0.0092), platelet count was not affiliated with SBP analyzed as continuous variable. No significant “cutoff” was identified by ROC for MELD“score. [SBP = spontaneous bacterial peritonitis; CRP = C-reactive protein; MELD = model for end-stage liver disease; ROC = receiver operating characteristics].
The AUC was 0.71 in training set (95%CI 0.629– 0.786, p < 0.0001) and 0.68 in the validation set (95% CI 0.511–0.848, p = 0.0245). In order to improve the sensitivity of our diagnostic score, we evaluated a second scoring system named
Table III. Scoring system for diagnosis of SBP. The score is calculated by adding the points. Thus the total score for each patient is 0 to 4. Parameter Age Platelet count C-reactive protein
“Cutoff”
Scoring-Points
>60 years £100.000/mL >60 mg/L
1 1 2
“model B”: however, addition of a lower scoring step for a CRP between 30 and 60 mg/L gaining one point, does not significantly improve sensitivity and NPV, while there is a relevant loss in specificity and PPV, as depicted in Figure 3. Discussion The diagnosis of SBP has instant therapeutic consequences and may not be delayed. The combination of age, CRP and platelet count in a scoring system, as displayed in table III, accurately identifies a subset of patients with SBP and excludes the presence of SBP in another subset of patients. Our scoring system is
600
M. H. Wehmeyer et al.
A
B 100
50 25
50 25
0
0 1+
Scand J Gastroenterol Downloaded from informahealthcare.com by Universitaet Zuerich on 07/09/14 For personal use only.
Model A Model B
75 NPV
75 PPV
100
Model A Model B
2+
3+
4
1+
2+
3+
4
Figure 3. The positive [A] and negative [B] predictive value for predicting SBP of the two different scoring systems for the labeled “cutoffs” (for “model A” and “model B,” respectively). The 95%-CI is indicated by error bars. In both models 1 point is gained each for age >60 years and platelets £100,000/mL and 2 points for CRP >60 mg/L. Model B is defined by an additional scoring level for patients with a CRP from 30 to 60 mg/L who receive one 1 point. [SBP = spontaneous bacterial peritonitis; CRP = C-reactive protein; PPV = positive predictive value; NPV = negative predictive value].
solidly based on the prospective design of our study and the relatively large number of patients included. We found the scoring model A superior to the scoring model B because of the higher certainty in the diagnosis of SBP (see Figure 3). SBP is very unlikely in patients with a score of 0 (negative predictive value 93.5% at a threshold of one scoring point in the training set and 83.3% in the validation set). The risk for suffering from SBP increases with rising score. The positive predictive value at a “cutoff” of 3 or more points was 81.8% in training set and 100% in validation set. The simple scoring system consists only of parameters as collected in daily clinical routine at admission of patients with ascites to the hospital and shows clear and easy to remember “cutoffs.” Our scoring system helps to decide which patients need immediate antibiotic treatment if a prompt and secure paracentesis is not available or there is a lack of experience in this technique. Further, patients scoring at least three points should be considered as suffering from SBP given the high positive and negative predictive values in both sets of patients at this “cutoff.” In fact, only
2/162 patients of the control group scored 3 or 4 points. One limitation of our proposed scoring system is that it fails to establish or exclude the diagnosis of SBP in patients scoring one or two points. In addition, it is important to have in mind that our scoring system fails to diagnose SBP in 7 patients in the training set at a “cutoff” of 1 scoring point (sensitivity 87.9%). However, the parameters integrated in our score are easily and routinely assessed, and thereby allow to rapidly diagnose or exclude SBP in many patients without additional effort and with sufficient accuracy. An association between age and SBP has not been reported so far. One possible explanation for this observation is an increased comorbidity or advanced cirrhosis in older patients, leading to a greater susceptibility for infectious complications in general. Further, age was shown to alter functions of the innate immune system, which results in an elevated risk for bacterial infections in the elderly [28]. We found a platelet count £100,000/mL to be an independent predictor of SBP. A correlation of platelets as a continuous variable with SBP as observed in
Table IV. Diagnostic validity of the scoring system at different “cutoffs.” One scoring point assigned for each age >60 years and platelets £100,000/mL and two points for CRP >60 mg/L [SBP = spontaneous bacterial peritonitis; PPV = positive predictive value; NPV = negative predictive value; CRP = C-reactive protein; 95%CI = 95% confidence interval; n = number]. Cutoff
Sensitivity (%)
Training set (n = 220) 1 87.9 2 41.4 3 15.5 4 3.5 Validation set (n = 76) 1 76.5 2 47.1 3 29.4 4 11.5
95%CI (%)
PPV (%)
95% CI (%)
Specificity (%)
95%CI (%)
NPV (%)
95%CI (%)
76.7–95.0 28.6–55.1 7.4–27.4 0.4–11.9
45.5 53.3 81.8 100
36.2–55.2 38.0–68.1 47.8–96.8 19.8–100
37.7 87.0 98.8 100
30.2–45.6 80.9–91.8 95.6–99.9 97.8–100
93.5 80.6 76.6 74.3
86.6–97.1 73.8–86.0 70.1–82.0 67.9–79.9
50.1–93.2 23.0–72.2 10.3–56.0 1.5–36.4
25.0 53.3 100 100
14.5–39.2 27.4–77.7 46.3–100 19.8–100
33.9 88.1 100 100
22.1–47.4 77.1–95.1 93.9–100 93.9–100
83.3 85.3 83.1 79.7
61.8–94.5 73.3–92.6 71.9–90.6 68.5–87.9
Scand J Gastroenterol Downloaded from informahealthcare.com by Universitaet Zuerich on 07/09/14 For personal use only.
A scoring system for rapid diagnosis of SBP other studies [27] could not be demonstrated in our patients. However, a platelet count £98,000/mL has been described as an independent predictor for developing SBP in patients with low ascitic fluid protein before [24], and other authors found evidence that the degree of portal hypertension correlates with the risk for SBP development [29]. Recent findings suggest that platelets have numerous immunologic functions. In particular, platelets have been shown to play an important role by activating neutrophil granulocytes in bacterial infections [30–33]. Thus, a low platelet count might result in insufficient activation of neutrophils in cirrhotic patients and might be a risk factor for the development of infections in cirrhotic patients in general. Not surprisingly, our data confirmed that high CRP levels are also a strong and independent predictor of SBP [27]. The MELD score, originally developed for the prediction of survival after TIPS implantation, consists of creatinine, bilirubin and INR and is now widely used for the assessment of prognosis of cirrhotic patients [34,35]. A high MELD score has been shown to be associated with complications of liver cirrhosis, such as SBP [25], ascites, or hepatic encephalopathy [36]. But neither the MELD-score nor its components – including bilirubin as a widely accepted indicator of liver damage [21,25] – are independent predictors of SBP in our patients. A Child-Pugh stadium of B or C being another indicator for the prognosis of cirrhotic patients [36–38] was not associated with the presence of SBP. While these well established assessment tools for the grading of cirrhosis may be useful in identifying patients at risk for a future episode of SBP [21,23–25], they are not helpful in diagnosing or excluding the actual presence of SBP. Our data indicate that SBP takes place in all decompensated cirrhotic patients regardless of the underlying liver disease. However, many of our patients with an etiology other than solely ethanol showed a confounding abuse of alcohol and were not classified as alcoholic cirrhosis. Therefore, we cannot verify the observation that patients with alcoholic cirrhosis are at an increased risk for developing SBP [27]. We defined SBP solely via neutrophil cell count in ascitic fluid. Thus monomicrobial nonneutrocytic ascites (MBA) was not classified as SBP and consequently our scoring system fails to detect MBA. However, given the low prevalence of MBA in general [6,39] and in our study (3.3% in the control group of the training set) it is very unlikely that the results of our study were inflicted significantly. In addition, only 12% of the obtained ascites cultures were positive in training set, which is significantly less then reported before [40,41]. Further, significantly more cultures
601
were obtained in the SBP group. Treating MBA as SBP would have been an additional source of potential bias. Another proposed screening tool for rapid diagnosis of SBP is the use of leukocyte esterase reagent strips in ascitic fluid [42]. Therefore, future studies should compare diagnostic fidelity of our scoring system with the diagnostic fidelity of reagent strips for diagnosis of SBP. In conclusion, SBP must be ruled out in all patients with cirrhotic ascites regardless of the stage of cirrhosis as assessed by MELD- or Child-Pugh Score. Our simple scoring system is a valid screening test for rapid exclusion or diagnosis of SBP. However, despite high specify in patients with 3 or 4 scoring points, additional ascites and blood cultures are essential to identify drug-resistant bacteria strains. We suggest to treat all individuals with cirrhotic ascites as suffering from SBP who show elevated CRP above 60 mg/L and either are >60 years or have a platelet count of £100,000 cells/mL, especially if prompt paracentesis is not available or cannot be performed safely, for example, at private practice or due to lack of experience in this technique.
Acknowledgments Financial support: supported by the Deutsche Forschungsgemeinschaft (Grant LU B62/2-1).
Authors’ involvement with the manuscript Malte H. Wehmeyer: Study concept and design; acquisition of data; analysis and interpretation of data; statistical analysis; drafting of the manuscript. Sarah Krohm: Acquisition of data. Friederike Kastein: Acquisition of data. Ansgar W. Lohse: Critical revision of the manuscript. Stefan Lüth: Study concept and design; critical revision of the manuscript. Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the article.
References [1] Kerr DN, Pearson DT, Read AE. Infection of ascitic fluid in patients with hepatic cirrhosis. Gut 1963;4:394–8. [2] Runyon BA. Low-protein-concentration ascitic fluid is predisposed to spontaneous bacterial peritonitis. Gastroenterology 1986;91:1343–6. [3] Pinzello G, Simonetti RG, Craxi A, Di Piazza S, Spano C, Pagliaro L. Spontaneous bacterial peritonitis: a prospective
602
[4] [5]
[6]
[7]
Scand J Gastroenterol Downloaded from informahealthcare.com by Universitaet Zuerich on 07/09/14 For personal use only.
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
M. H. Wehmeyer et al. investigation in predominantly nonalcoholic cirrhotic patients. Hepatology 1983;3:545–9. Caly WR, Strauss E. A prospective study of bacterial infections in patients with cirrhosis. J Hepatol 1993;18:353–8. Bac DJ, Siersma PD, Mulder PGH. Spontaneous bacterial peritonitis: outcome and predictive factors. Eur J Gastroenterol Hepatol 1993;5:635–40. Evans LT, Kim WR, Poterucha JJ, Kamath PS. Spontaneous bacterial peritonitis in asymptomatic outpatients with cirrhotic ascites. Hepatology 2003;37:897–901. Runyon BA, Squier SU, Borzio M. Translocation of gut bacteria in rats with cirrhosis to mesenteric lymph nodes partially explains the pathogenesis of spontaneous bacterial peritonitis. J Hepatol 1994;21:792–6. Cirera I, Bauer TM, Navasa M, Vila J, Grande L, Taura P, et al. Bacterial translocation of enteric organisms in patients with cirrhosis. J Hepatol 2001;34:32–7. Guarner C, Soriano G. Bacterial translocation and its consequences in patients with cirrhosis. Eur J Gastroenterol Hepatol 2005;17:27–31. Llovet JM, Bartoli R, Planas R, Cabre E, Jiminez M, Urban A, et al. Bacterial translocation in cirrhotic rats. Its role in the development of spontaneous bacterial peritonitis. Gut 1994;35:1648–52. Guarner C, Runyon BA, Young S, Heck M, Sheikh MY. Intestinal bacterial overgrowth and bacterial translocation in cirrhotic rats with ascites. J Hepatol 1997;26:1372–8. Casafont Morencos F, de las Heras Castano G, Martin Ramos L, Lopez Arias MJ, Ledesma F, Pons Romero F. Small bowel bacterial overgrowth in patients with alcoholic cirrhosis. Dig Dis Sci 1995;40:1252–6. Madrid AM, Cumsille F, Defilippi C. Altered small bowel motility in patients with liver cirrhosis depends on severity of liver disease. Dig Dis Sci 1997;42:738–42. Chang CS, Chen GH, Lien HC, Yeh HZ. Small intestine dysmotility and bacterial overgrowth in cirrhotic patients with spontaneous bacterial peritonitis. Hepatology 1998; 28:1187–90. Runyon BA. Patients with deficient ascitic fluid opsonic activity are predisposed to spontaneous bacterial peritonitis. Hepatology 1988;8:632–5. Runyon BA, Morrissey RL, Hoefs JC, Wyle FA. Opsonic activity of human ascitic fluid: a potentially important protective mechanism against spontaneous bacterial peritonitis. Hepatology 1985;5:634–7. Fiuza C, Salcedo M, Clemente G, Tellado JM. In vivo neutrophil dysfunction in cirrhotic patients with advanced liver disease. J Infect Dis 2000;182:526–33. Jimenez W, Ros J, Morales-Ruiz M, Navasa M, Sole M, Colmenero J, et al. Nitric oxide production and inducable nitric oxide synthase expression in peritoneal macrophages of cirrhotic patients. Hepatology 1999;30:670–6. Such J, Hillebrand DJ, Guarner C, Berk L, Zapater P, Westengard J, et al. Nitric oxide in ascitic fluid is an independent predictor of renal impairment in patients with cirrhosis and spontaneous bacterial peritonitis. Eur J Gastroenterol Hepatol 2004;16:571–7. Frances R, Zapater P, Gonzales-Navajas JM, Munoz C, Cano R, Moreu R, et al. Bacterial DNA in patients with cirrhosis and noninfected ascites mimics the soluble immune response established in patients with spontaneous bacterial peritonitis. Hepatology 2008;47:978–85. Tito L, Rimola A, Gines P, Llach J, Arroyo V, Rodes J. Recurrence of spontaneous bacterial peritonitis in cirrhosis: frequency and predictive factors. Hepatology 1988;8:27–31.
[22] Bernard B, Cadranel JF, Valla D, Escolano S, Jarlier V, Opolon P. Prognostic significance of bacterial infection in bleeding cirrhotic patients: a prospective study. Gastroenterology 1995;108:1828–34. [23] Llach J, Rimola A, Navasa M, Gines P, Salmeron JM, Gines A, et al. Incidence and predictive factors of first episode of spontaneous bacterial peritonitis in cirrhosis with ascites: relevance of ascitic fluid protein concentration. Hepatology 1992;16:724–7. [24] Guarner C, Sola R, Soriano G, Andreu M, Novella MT, Vila MC, et al. Risk of a first community-acquired spontaneous bacterial perionitis in cirrhotics with low ascitic fluid protein levels. Gastroenterology 1999;117:414–19. [25] Obstein KL, Campbell MS, Reddy KR, Yang YX. Association between model for end-stage liver disease and spontaneous bacterial peritonitis. Am J Gastroenterol 2007;102: 2732–6. [26] Shi KQ, Fan YC, Ying L, Lin XF, Song M, Li LF, et al. Risk stratification of spontaneous bacterial peritonitis in cirrhosis with ascites based on classification and regression tree analysis. Mol Biol Rep 2012;39:6161–9. [27] Lata J, Fejfar T, Krechler T, Musil T, Husova L, Senkyrik M, et al. Spontaneous bacterial peritonitis in the czech republic: prevalence and aetiology. Eur J Gastroenterol Hepatol 2003;15:739–43. [28] Solana R, Tarazona R, Gayoso I, Lesur O, Dupuis G, Fulop T. Innate immunosenescence: effect of aging on cells and receptors of the innate immune system in humans. Semin Immunol 2012;24:331–41. [29] Serste T, Bourgeois N, Lebrec D, Evrard S, Deviere J, Le Moine O. Relationship between the degree of portal hypertension and the onset of sponateous bacterial peritonitis in patients with cirrhosis. Acta Gastroenterologica Belgica 2006;69:355–60. [30] Elzey BD, Tian J, Jensen RJ, Swanson AK, Lees JR, Lentz SR, et al. Platelet-mediated modulation of adaptive immunity: a communication link between innate and adaptive immune compartments. Immunity 2003;19:9–19. [31] Aslam R, Speck ER, Kim M, Crow AR, Bang KW, Nestel FP, et al. Platelet Toll-like receptor expression modulates lipopolysaccharide-induced thrombocytopenia and tumor necrosis factor-a production in vivo. Blood 2006; 107:637–41. [32] Clark SR, Ma AC, Tavener SA, McDonald B, Goodarzi Z, Kelly MM, et al. Platelet TLR4 activates neutrophil extracellular traps to ensnare bacteria in septic blood. Nat Med 2007;13:463–9. [33] Kuckleburg CJ, Tiwari R, Czuprynski CJ. Endothelial cell apoptosis induced by bacteria-activated platelets requires caspase-8 and -9 and generation of reactive oxygen species. Thromb Haemost 2008;99:363–72. [34] Kamath PS, Wiesner RH, Malinchoc M, Kremers W, Therneau TM, Kosberg CL, et al. A model to predict survival in patients with end-stage liver disease. Hepatology 2001;33:464–70. [35] Botta F, Giannini E, Romagnoli P, Fasoli A, Malfatti F, Chiarbonello B, et al. MELD scoring system is useful for predicting prognosis in patients with liver cirrhosis and is correlated with residual liver function: a European study. Gut 2003;52:134–9. [36] Huo TI, Wu JC, Lin HC, Lee FY, Hou MC, Lee PC, et al. Evaluation of the increase in model for end-stage liver disease (deltaMELD) score over time as a prognostic predictor in patients with advanced cirrhosis: risk factor analysis and
A scoring system for rapid diagnosis of SBP
Scand J Gastroenterol Downloaded from informahealthcare.com by Universitaet Zuerich on 07/09/14 For personal use only.
comparison with initial MELD and Child-Turcotte-Pugh score. J Hepatol 2005;42:826–32. [37] Pugh RNH, Murray-Lyon IM, Dawson JL, Pietroni MC, Williams R. Transection of the oesophagus for bleeding oesophageal varices. Br J Surg 1973;60:646–9. [38] Albers I, Hartmann H, Bircher J, Creutzfeldt W. Superiority of the Child-Pugh classification to quantitative liver function tests for assessing prognosis of liver cirrhosis. Scand J Gastroenterol 1989;24:269–76. [39] Castellote J, Girbau A, Maisterra S, Charhi N, Ballester R, Xiol X. Spontaneous bacterial peritonitis and bacterascites prevalence in asymptomatic cirrhotic outpatients undergoing large-volume paracentesis. J Gastroenterol Hepatol 2008;23: 256–9.
603
[40] Llovet JM, Planas R, Morillas R, Quer JC, Cabre E, Biox J, et al. Short-term prognosis of cirrhotics with spontaneous bacterial peritonitis: multivariate study. Am J Gastroenterol 1993;88:388–92. [41] Navasa M, Follo A, Llovet JM, Clemente G, Vargas V, Rimola A, et al. Radomized, comparative study of oral ofloxacin versus intravenous cefotaxime in spontaneous bacterial peritonitis. Gastroenterology 1996; 111:1011–17. [42] Koulaouzides A, Leontiadis GI, Abdullah M, Moschos J, Gasem J, Tharakan J, et al. Leucocyte esterase reagent strips for the diagnosis of spontaneous bacterial peritonitis: a systematic review. Eur J Gastroenterol Hepatol 2008;20: 1055–60.
ORIGINAL ARTICLE
Scand J Gastroenterol Downloaded from informahealthcare.com by Universitaet Zuerich on 07/09/14 For personal use only.
Prediction of spontaneous bacterial peritonitis in cirrhotic ascites by a simple scoring system
MALTE H. WEHMEYER, SARAH KROHM, FRIEDERIKE KASTEIN, ANSGAR W. LOHSE & STEFAN LÜTH I. Department of Medicine, University Medical Center Hamburg-Eppendorf, Hamburg, Germany
Abstract Background and aims. Spontaneous bacterial peritonitis (SBP) is a life-threatening complication in patients with liver cirrhosis. The aim of this prospective study was to identify predictors of SBP in order to develop a noninvasive method to identify or exclude an episode of SBP. Patients and methods. Three hundred and ninety-two consecutive patients, who underwent paracentesis from March 2008 through January 2012 in our department due to cirrhotic ascites, were screened. Ninety-six patients were excluded, mostly due to prior application of antibiotics. SBP was defined by an absolute neutrophil count ‡250 cells/mL ascites. We evaluated various clinical and laboratory parameters as potential predictors of SBP. A scoring system was developed in a training set of 220 and validated in a second set of 76 patients. Results. Fifty-eight patients (26%) in the training set and 17 patients in the validation set (22%) suffered from SBP. Thrombocytopenia £100,000 cells/mL, age >60 years and CRP >60 mg/L were identified as independent predictors of SBP. A scoring system combining these three parameters (weighting thrombocytopenia and age with 1 point each, but CRP with 2 points) reaches a positive predictive value for the diagnosis of SBP of 81.8% with a specificity of 98.8% (score ‡3). The negative predictive value at a threshold of 1 point is 93.5% with a sensitivity of 87.9%. Notably, a high MELD score is not associated with SBP (p = 0.3344). Conclusions. Combination of age, CRP and platelet count in a simple scoring system helps in the rapid diagnosis or exclusion of SBP.
Key Words: platelets, predictors, SBP, thrombocytes
Introduction Spontaneous bacterial peritonitis (SBP) is a frequent and life-threatening infection in patients with ascites due to liver cirrhosis, first described in the 1960s [1]. The prevalence of SBP ranges from 10% to 30% in hospitalized patients [2–5] and is ~3.5% in asymptomatic outpatients [6] suffering from cirrhotic ascites. The diagnosis of SBP has immediate therapeutic consequences, such as the start of an efficient antibiotic treatment and should not be missed or delayed, therefore. Translocation of bacteria from the intestinal lumen into mesenteric lymph nodes with colonization of blood, lymph, and ascites is the key event in the pathogenesis of SBP [7–10]. There are three
predisposing factors for SBP: First, patients with an advanced cirrhosis display a decreased mucosal barrier function against intestinal bacteria [7,8]. Second, bacterial overgrowth [11,12] is a common result of a disturbed intestinal motility in cirrhotic patients [13,14]. Third, reduced immune function in patients with liver cirrhosis permits an uncontrolled bacterial growth in ascites [15–17]. As a consequence of bacteremia and bacterascites vasodilatation may lead to renal failure [18–20]. Clinical risk factors for developing SBP, such as history of prior SBP [21] or variceal hemorrhage [22] and low ascitic fluid protein levels [2,6,23] are well known. However, available data concerning the prediction of SBP by routine laboratory tests are contradictory. An association between hyperbilirubinemia
Correspondence: Stefan Lüth, MD, I. Department of Medicine, University Medical Center Hamburg-Eppendorf, Martinistr. 52, 20246 Hamburg, Germany. Tel: +49 40 7410 58941. Fax: +49 40 7410 58531. E-mail: [email protected]
(Received 9 August 2013; revised 5 September 2013; accepted 20 September 2013) ISSN 0036-5521 print/ISSN 1502-7708 online 2014 Informa Healthcare DOI: 10.3109/00365521.2013.848471
596
M. H. Wehmeyer et al. 392 patients Exclusion criteria: Receiving antibiotics (n = 82), hemorrhagic or malignant ascites (n = 14).
Scand J Gastroenterol Downloaded from informahealthcare.com by Universitaet Zuerich on 07/09/14 For personal use only.
296 patients
96 patients excluded
Training set: 03/2008– 05/2010 (n = 220)
Validation set: 03/2011– 01/2012 (n = 76)
58 patients with SBP (≥ 3 billion neutrocytes/L)
17 patients with SBP (≥ 3 billion neutrocytes/L)
Culture positive 12%
CNNA 88%
Culture positive 25%
CNNA 75%
Figure 1. Study design and results. Period of time in which the two different sets of patients were collected is given, too. [n = number; SBP = spontaneous bacterial peritonitis; CNNA = culture-negative neutrocytic ascites].
and SBP has been demonstrated repeatedly [21,23– 26], but also reduced prothrombin activity [21,23,26], aspartate transaminase (AST) [23], C-reactive protein (CRP) [27], leukocytosis [26,27], serum creatinine [26], Child-Pugh score [24], MELD score [25] and platelet count [24,27] have been shown to be associated with an episode of SBP. However, either these studies contained only a relatively small collective [21,23] of patients, were designed to identify long-term risk factors for SBP [21,23,24,26] or were designed as a retrospective study [25]. A rather complex scoring system consisting of creatinine, bilirubin, thrombin time, and leucocyte count for risk stratefication for SBP was recently developed by Shi et al. [26] The aim of this prospective study was to identify clinical and laboratory parameters that predict SBP in patients with cirrhotic ascites and to develop a simple scoring system which allows a rapid diagnosis of SBP or to exclude SBP without the need for instant paracentesis. Patients and methods Study population Diagnostic with or without therapeutic paracentesis was performed in 392 patients admitted to the I. Department of Medicine of the University Medical Center Hamburg-Eppendorf with cirrhotic ascites from March 2008 through January 2012.
Our training set consisted of 256 patients. Of these, 36 patients were excluded from the study due to antibiotic treatment at the time of paracentesis (n = 32) or hemorrhagic ascites (n = 4). Two other exclusion-criteria (malignant ascites and evidence for secondary peritonitis) were not met by any of the remaining patients in the training set. A total of 220 consecutive patients were included in the training set. To verify our findings, we further screened 136 patients, of which a total of 60 patients were excluded (50 patients receiving antibiotics at the time of paracentesis and 10 showed evidence for hemorrhagic or malignant ascites). Thus 76 patients could be recruited for the validation set. The study design is summarized in Figure 1. Cases and controls A neutrophil cell count in the ascitic fluid >250 cells/m L defined the diagnosis of SBP regardless of the results of ascitic culture. Therefore, patients with culture-positive and culture-negative neutrocytic ascites (CNNA) were considered as SBP positive. At the time of paracentesis (+/- 1 day) laboratory and clinical data were assessed and used for analysis. Patients with a neutrophil cell count 60 years History of SBP Short-time history of variceal bleeding Etiology of Cirrhosis Alcohol Hepatitis C ± alcohol NASH/cryptogenic Autoimmune ± alcohol Hepatitis B ± alcohol Other Creatinine [mg/dL] Bilirubin [mg/dL] Bilirubin >5 mg/dL INR MELD score Child-Pugh stadium Child B Child C Albumin [g/L] CRP [mg/L] CRP >60 mg/L Platelet Count [103/mL] Platelets £100,000/mL
SBP present (n = 58) Mean ± SD, median (range), or N (%) 40 (69.0%) 60.2 ± 8.9 33 (56,9%) 5 (8.6%) 6 (10.3%)
SBP absent (n = 162) Mean ± SD, median (range), or N (%) 105 (64.8%) 56.5 ± 10.2 63 (38.9%) 13 (8%) 17 (10.5%)
0.5657 0.0164* 0.0177* 1 1 0.5839 0.7913 0.5925 0.6070 0.5916 0.3294 0.7265 0.2180 0.0092* 0.1729 0.3344 0.8415
35 13 4 3 2 1 1.12 2.10 20 1.38 16
(60.3%) (22.4%) (6.9%) (5.2%) (3.5%) (1.7%) (0.5–5.8) (0.2–43.0) (34.5%) (0.94–2.18) (7–37)
91 39 11 9 5 7 1.20 2.05 29 1.34 15
(56.2%) (24.1%) (6.8%) (5.6%) (3.1%) (4.3%) (0.4–7.6) (0.2–36.4) (17.9%) (0.90–5.00) (6–40)
37 21 28 25.5 14 110 25
(63.8%) (36.2%) (18–38) (5–131) (24.1%) (24–481) (43.1%)
101 61 27 17.5 7 130 46
(62.4%) (37.7%) (12–47) (5–202) (4.3%) (32–456) (28.4%)
Statistical analysis Continuous variables with an assumed Gaussian distribution (age) were compared between cases and controls via t-test. Continuous variables considered nonnormally distributed (MELD-Score, creatinine, bilirubin, INR, platelet count, albumin, CRP) were compared using rank-sum test. Receiver–operator characteristic (ROC) curves were calculated to define the most discriminating “cutoffs” for measured data, thus identifying dichotomous variables associated with SBP. p-Values for dichotomous variables (etiology, sex, Child-Pugh stadium, laboratory values and age at specific cutoff) were calculated using Fisher’s exact probability test. A p-Value < 0.05 was considered statistically significant. To identify independent predictors of SBP, variables being significantly correlated with the risk for SBP in univariate analysis were entered in a multivariate model using stepwise logistic regression. Scoring points based on estimated coefficients derived from the regression model were assigned to independent predictors of SBP and added up for each patient. The diagnostic value of this scoring system was then assessed by the area under the ROC curve. Findings were validated in our validation set.
p-Value
0.4898 0.0055* < 0.0001* 0.2109 0.0397*
MELD-score and Child-Pugh stadium were calculated using the established formula. The study was approved by the local ethics committee. All patients gave informed consent. Statistical analyses were performed using SPSS Version 19; the figures were created with GraphPadPrism 4. Results Patients characteristics Fifty-eight out of 220 patients (26%) in the training set and 17 out of 76 patients (22%) in the validation set were diagnosed with SBP. Patient characteristics of our training set are displayed in Table I. In the validation set significantly more patients suffered from cirrhosis due to alcohol abuse and significantly less patients suffered from HCV-related cirrhosis (p = 0.0062 and 0.0040). Otherwise, there were no significant differences between the two sets (see Table II for details). Twenty-one patients with SBP (36.2%) and 61 individuals in control-group (37.7%) had a Child-C-cirrhosis (p = 0.8415). All other patients were classified as Child-B-cirrhosis, while no patients with a ChildPugh stadium A were found in the study population (Table I).
598
M. H. Wehmeyer et al.
Table II. Comparison of clinical data between the training and the validation set. [SD = standard deviation; n = number; SBP = spontaneous bacterial peritonitis; NASH = nonalcoholic steatohepatitis; * p < 0.05]. Variable
Scand J Gastroenterol Downloaded from informahealthcare.com by Universitaet Zuerich on 07/09/14 For personal use only.
Male Gender Age [years] Prevalence of SBP Etiology of Cirrhosis Alcohol Hepatitis C ± alcohol NASH/cryptogenic Autoimmune ± alcohol Hepatitis B ± alcohol Other
Training set: March 2008 to May 2010 (n = 220) mean ± SD, or N (%)
Validation set: March 2011 to January 2012 (n = 76) mean ± SD, or N (%)
p-Value
145 (65.9%) 57.5 ± 10.0 58 (26.4%)
49 (64.5%) 59.1 ± 9.1 17 (22.4%)
0.8888 0.3064 0.5430
126 52 15 12 7 8
(57.3%) (23.6%) (6.8%) (5.5%) (3.2%) (3.6%)
Thirteen patients in the control group (8%) and five patients in the SBP group had a history of SBP (8.6%). Seventeen patients in the control group (10.5%) and six additional patients in the SBP group (10.3%) presented with a short-time history of variceal hemorrhage. Thirty-three ascitic fluid cultures were obtained in the SBP group (56.9%) with four positive results (12.1%) for two gram-negative and two Grampositive bacteria respectively. 61 cultures were performed in the control group (37.7%), diagnosing monomicrobial nonneutrocytic ascites in two individuals (3.3%; one Gram-positive and one Gramnegative bacteria, respectively). There was no patient with polymicrobial ascites infection. Ascitic fluid culture was conducted relatively more often in patients with SBP (p = 0.0110). Age, elevated CRP and thrombocytopenia are associated with the presence of SBP Univariate analysis identified significant differences between cases and controls only for the continuous variables age and CRP (p = 0.0164 and 0.0032, respectively). A platelet count considered as a dichotomous variable at a “cutoff” of £100,000/mL predicted SBP with a sensitivity of 43.1% and a specificity of 71.6% (p = 0.0397). The ROC curves for serum CRP indicated a sensitivity of 25.5% and a specificity of 96.5% for SBP at >60 mg/L (p < 0.0001). The area under the curve (AUC) was 0.6358. An age >60 years predicted the presence of SBP in our population with a sensitivity of 56.9% and a specificity of 61.1% (p = 0.0177). The AUC was 0.6054. The most discriminating “cutoff” identified by ROC-curve for bilirubin was a level >5 mg/dL (p = 0.0092). Details are depicted in Figure 2.
57 6 4 3 3 3
(75.0%) (7.9%) (5.3%) (4.0%) (4.0%) (4.0%)
0.0062* 0.0040* 0.7894 0.7671 1 1
Potential thresholds for creatinine, INR and albumin that were generated the same way by ROC curves did not discriminate between SBP patients and controls (data not shown). The best “cutoff” for MELD-score identified via ROC at >23 was not able to differentiate between cases and controls, too (p = 0.0680; Figure 2). A simple scoring system identifies patients with SBP Platelets, age, CRP and bilirubin were entered in a stepwise logistic regression model. Only platelets £100,000/mL (odds ratio (OR) 2.55, 95% CI 1.28– 5.07, p = 0.008), age >60 years (OR 2.72, 95% CI 1.39–5.34, p = 0.004) and CRP >60 mg/L (OR 9.79, 95% CI 3.46–27.66, p < 0.001) were identified as being independently affiliated with SBP. Diagnostic fidelity was improved when combining these parameters in a scoring system with age >60 years and platelets £100,000/mL each scoring one point (since the estimated coefficient for age was 1.0 and the coefficient for platelets was 0.9 in logistic regression) and CRP >60 mg/L scoring two points (the estimated coefficient in logistic regression was 2.3; “model A”). An overview of the scoring system is displayed in table III. At a “cutoff” of 1 scoring-point the positive predictive value (PPV) was 45.5% and the negative predictive value (NPV) was 93.5%. Thus in 93.5% of the patients with a scoring of 0 points a SBP was excluded correctly. Using a “cutoff” of 1 point led to false negative testresults in 7 out of 58 patients with SBP in the training set (sensitivity 87.9%). At a “cutoff” of 3 points PPV was 81.8% with a NPV of 76.6% and all patients with a score of 4 were suffering from SBP (PPV 100% at a “cutoff” of 4). Two patients in the control group displayed a scoring of 3 points (specificity 98.8%). Further statistical values from both the training and validation set are detailed in table IV.
A scoring system for rapid diagnosis of SBP B
A
90
220 200
80
180
70
160
60
140
Age [years]
CRP [mg/L]
599
120 100 80
50 40 30 20
40
10
20 0
0 SBP present
SBP absent
SBP present
SBP absent
SBP present
SBP absent
D
C
50
550 500 450
40
400 350
MELD score
Platelets [103/mL]
Scand J Gastroenterol Downloaded from informahealthcare.com by Universitaet Zuerich on 07/09/14 For personal use only.
60
120 250 200
30
20
150 10
100 50 0
0 SBP present
SBP absent
Figure 2. CRP [A], age [B], platelet count [C] and MELD score [D] in cases and controls (training set). The most discriminating “cutoff” for the prediction of SBP is marked for each variable (A to C: p < 0.0001, p = 0.0177 and p = 0.0092), platelet count was not affiliated with SBP analyzed as continuous variable. No significant “cutoff” was identified by ROC for MELD“score. [SBP = spontaneous bacterial peritonitis; CRP = C-reactive protein; MELD = model for end-stage liver disease; ROC = receiver operating characteristics].
The AUC was 0.71 in training set (95%CI 0.629– 0.786, p < 0.0001) and 0.68 in the validation set (95% CI 0.511–0.848, p = 0.0245). In order to improve the sensitivity of our diagnostic score, we evaluated a second scoring system named
Table III. Scoring system for diagnosis of SBP. The score is calculated by adding the points. Thus the total score for each patient is 0 to 4. Parameter Age Platelet count C-reactive protein
“Cutoff”
Scoring-Points
>60 years £100.000/mL >60 mg/L
1 1 2
“model B”: however, addition of a lower scoring step for a CRP between 30 and 60 mg/L gaining one point, does not significantly improve sensitivity and NPV, while there is a relevant loss in specificity and PPV, as depicted in Figure 3. Discussion The diagnosis of SBP has instant therapeutic consequences and may not be delayed. The combination of age, CRP and platelet count in a scoring system, as displayed in table III, accurately identifies a subset of patients with SBP and excludes the presence of SBP in another subset of patients. Our scoring system is
600
M. H. Wehmeyer et al.
A
B 100
50 25
50 25
0
0 1+
Scand J Gastroenterol Downloaded from informahealthcare.com by Universitaet Zuerich on 07/09/14 For personal use only.
Model A Model B
75 NPV
75 PPV
100
Model A Model B
2+
3+
4
1+
2+
3+
4
Figure 3. The positive [A] and negative [B] predictive value for predicting SBP of the two different scoring systems for the labeled “cutoffs” (for “model A” and “model B,” respectively). The 95%-CI is indicated by error bars. In both models 1 point is gained each for age >60 years and platelets £100,000/mL and 2 points for CRP >60 mg/L. Model B is defined by an additional scoring level for patients with a CRP from 30 to 60 mg/L who receive one 1 point. [SBP = spontaneous bacterial peritonitis; CRP = C-reactive protein; PPV = positive predictive value; NPV = negative predictive value].
solidly based on the prospective design of our study and the relatively large number of patients included. We found the scoring model A superior to the scoring model B because of the higher certainty in the diagnosis of SBP (see Figure 3). SBP is very unlikely in patients with a score of 0 (negative predictive value 93.5% at a threshold of one scoring point in the training set and 83.3% in the validation set). The risk for suffering from SBP increases with rising score. The positive predictive value at a “cutoff” of 3 or more points was 81.8% in training set and 100% in validation set. The simple scoring system consists only of parameters as collected in daily clinical routine at admission of patients with ascites to the hospital and shows clear and easy to remember “cutoffs.” Our scoring system helps to decide which patients need immediate antibiotic treatment if a prompt and secure paracentesis is not available or there is a lack of experience in this technique. Further, patients scoring at least three points should be considered as suffering from SBP given the high positive and negative predictive values in both sets of patients at this “cutoff.” In fact, only
2/162 patients of the control group scored 3 or 4 points. One limitation of our proposed scoring system is that it fails to establish or exclude the diagnosis of SBP in patients scoring one or two points. In addition, it is important to have in mind that our scoring system fails to diagnose SBP in 7 patients in the training set at a “cutoff” of 1 scoring point (sensitivity 87.9%). However, the parameters integrated in our score are easily and routinely assessed, and thereby allow to rapidly diagnose or exclude SBP in many patients without additional effort and with sufficient accuracy. An association between age and SBP has not been reported so far. One possible explanation for this observation is an increased comorbidity or advanced cirrhosis in older patients, leading to a greater susceptibility for infectious complications in general. Further, age was shown to alter functions of the innate immune system, which results in an elevated risk for bacterial infections in the elderly [28]. We found a platelet count £100,000/mL to be an independent predictor of SBP. A correlation of platelets as a continuous variable with SBP as observed in
Table IV. Diagnostic validity of the scoring system at different “cutoffs.” One scoring point assigned for each age >60 years and platelets £100,000/mL and two points for CRP >60 mg/L [SBP = spontaneous bacterial peritonitis; PPV = positive predictive value; NPV = negative predictive value; CRP = C-reactive protein; 95%CI = 95% confidence interval; n = number]. Cutoff
Sensitivity (%)
Training set (n = 220) 1 87.9 2 41.4 3 15.5 4 3.5 Validation set (n = 76) 1 76.5 2 47.1 3 29.4 4 11.5
95%CI (%)
PPV (%)
95% CI (%)
Specificity (%)
95%CI (%)
NPV (%)
95%CI (%)
76.7–95.0 28.6–55.1 7.4–27.4 0.4–11.9
45.5 53.3 81.8 100
36.2–55.2 38.0–68.1 47.8–96.8 19.8–100
37.7 87.0 98.8 100
30.2–45.6 80.9–91.8 95.6–99.9 97.8–100
93.5 80.6 76.6 74.3
86.6–97.1 73.8–86.0 70.1–82.0 67.9–79.9
50.1–93.2 23.0–72.2 10.3–56.0 1.5–36.4
25.0 53.3 100 100
14.5–39.2 27.4–77.7 46.3–100 19.8–100
33.9 88.1 100 100
22.1–47.4 77.1–95.1 93.9–100 93.9–100
83.3 85.3 83.1 79.7
61.8–94.5 73.3–92.6 71.9–90.6 68.5–87.9
Scand J Gastroenterol Downloaded from informahealthcare.com by Universitaet Zuerich on 07/09/14 For personal use only.
A scoring system for rapid diagnosis of SBP other studies [27] could not be demonstrated in our patients. However, a platelet count £98,000/mL has been described as an independent predictor for developing SBP in patients with low ascitic fluid protein before [24], and other authors found evidence that the degree of portal hypertension correlates with the risk for SBP development [29]. Recent findings suggest that platelets have numerous immunologic functions. In particular, platelets have been shown to play an important role by activating neutrophil granulocytes in bacterial infections [30–33]. Thus, a low platelet count might result in insufficient activation of neutrophils in cirrhotic patients and might be a risk factor for the development of infections in cirrhotic patients in general. Not surprisingly, our data confirmed that high CRP levels are also a strong and independent predictor of SBP [27]. The MELD score, originally developed for the prediction of survival after TIPS implantation, consists of creatinine, bilirubin and INR and is now widely used for the assessment of prognosis of cirrhotic patients [34,35]. A high MELD score has been shown to be associated with complications of liver cirrhosis, such as SBP [25], ascites, or hepatic encephalopathy [36]. But neither the MELD-score nor its components – including bilirubin as a widely accepted indicator of liver damage [21,25] – are independent predictors of SBP in our patients. A Child-Pugh stadium of B or C being another indicator for the prognosis of cirrhotic patients [36–38] was not associated with the presence of SBP. While these well established assessment tools for the grading of cirrhosis may be useful in identifying patients at risk for a future episode of SBP [21,23–25], they are not helpful in diagnosing or excluding the actual presence of SBP. Our data indicate that SBP takes place in all decompensated cirrhotic patients regardless of the underlying liver disease. However, many of our patients with an etiology other than solely ethanol showed a confounding abuse of alcohol and were not classified as alcoholic cirrhosis. Therefore, we cannot verify the observation that patients with alcoholic cirrhosis are at an increased risk for developing SBP [27]. We defined SBP solely via neutrophil cell count in ascitic fluid. Thus monomicrobial nonneutrocytic ascites (MBA) was not classified as SBP and consequently our scoring system fails to detect MBA. However, given the low prevalence of MBA in general [6,39] and in our study (3.3% in the control group of the training set) it is very unlikely that the results of our study were inflicted significantly. In addition, only 12% of the obtained ascites cultures were positive in training set, which is significantly less then reported before [40,41]. Further, significantly more cultures
601
were obtained in the SBP group. Treating MBA as SBP would have been an additional source of potential bias. Another proposed screening tool for rapid diagnosis of SBP is the use of leukocyte esterase reagent strips in ascitic fluid [42]. Therefore, future studies should compare diagnostic fidelity of our scoring system with the diagnostic fidelity of reagent strips for diagnosis of SBP. In conclusion, SBP must be ruled out in all patients with cirrhotic ascites regardless of the stage of cirrhosis as assessed by MELD- or Child-Pugh Score. Our simple scoring system is a valid screening test for rapid exclusion or diagnosis of SBP. However, despite high specify in patients with 3 or 4 scoring points, additional ascites and blood cultures are essential to identify drug-resistant bacteria strains. We suggest to treat all individuals with cirrhotic ascites as suffering from SBP who show elevated CRP above 60 mg/L and either are >60 years or have a platelet count of £100,000 cells/mL, especially if prompt paracentesis is not available or cannot be performed safely, for example, at private practice or due to lack of experience in this technique.
Acknowledgments Financial support: supported by the Deutsche Forschungsgemeinschaft (Grant LU B62/2-1).
Authors’ involvement with the manuscript Malte H. Wehmeyer: Study concept and design; acquisition of data; analysis and interpretation of data; statistical analysis; drafting of the manuscript. Sarah Krohm: Acquisition of data. Friederike Kastein: Acquisition of data. Ansgar W. Lohse: Critical revision of the manuscript. Stefan Lüth: Study concept and design; critical revision of the manuscript. Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the article.
References [1] Kerr DN, Pearson DT, Read AE. Infection of ascitic fluid in patients with hepatic cirrhosis. Gut 1963;4:394–8. [2] Runyon BA. Low-protein-concentration ascitic fluid is predisposed to spontaneous bacterial peritonitis. Gastroenterology 1986;91:1343–6. [3] Pinzello G, Simonetti RG, Craxi A, Di Piazza S, Spano C, Pagliaro L. Spontaneous bacterial peritonitis: a prospective
602
[4] [5]
[6]
[7]
Scand J Gastroenterol Downloaded from informahealthcare.com by Universitaet Zuerich on 07/09/14 For personal use only.
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
M. H. Wehmeyer et al. investigation in predominantly nonalcoholic cirrhotic patients. Hepatology 1983;3:545–9. Caly WR, Strauss E. A prospective study of bacterial infections in patients with cirrhosis. J Hepatol 1993;18:353–8. Bac DJ, Siersma PD, Mulder PGH. Spontaneous bacterial peritonitis: outcome and predictive factors. Eur J Gastroenterol Hepatol 1993;5:635–40. Evans LT, Kim WR, Poterucha JJ, Kamath PS. Spontaneous bacterial peritonitis in asymptomatic outpatients with cirrhotic ascites. Hepatology 2003;37:897–901. Runyon BA, Squier SU, Borzio M. Translocation of gut bacteria in rats with cirrhosis to mesenteric lymph nodes partially explains the pathogenesis of spontaneous bacterial peritonitis. J Hepatol 1994;21:792–6. Cirera I, Bauer TM, Navasa M, Vila J, Grande L, Taura P, et al. Bacterial translocation of enteric organisms in patients with cirrhosis. J Hepatol 2001;34:32–7. Guarner C, Soriano G. Bacterial translocation and its consequences in patients with cirrhosis. Eur J Gastroenterol Hepatol 2005;17:27–31. Llovet JM, Bartoli R, Planas R, Cabre E, Jiminez M, Urban A, et al. Bacterial translocation in cirrhotic rats. Its role in the development of spontaneous bacterial peritonitis. Gut 1994;35:1648–52. Guarner C, Runyon BA, Young S, Heck M, Sheikh MY. Intestinal bacterial overgrowth and bacterial translocation in cirrhotic rats with ascites. J Hepatol 1997;26:1372–8. Casafont Morencos F, de las Heras Castano G, Martin Ramos L, Lopez Arias MJ, Ledesma F, Pons Romero F. Small bowel bacterial overgrowth in patients with alcoholic cirrhosis. Dig Dis Sci 1995;40:1252–6. Madrid AM, Cumsille F, Defilippi C. Altered small bowel motility in patients with liver cirrhosis depends on severity of liver disease. Dig Dis Sci 1997;42:738–42. Chang CS, Chen GH, Lien HC, Yeh HZ. Small intestine dysmotility and bacterial overgrowth in cirrhotic patients with spontaneous bacterial peritonitis. Hepatology 1998; 28:1187–90. Runyon BA. Patients with deficient ascitic fluid opsonic activity are predisposed to spontaneous bacterial peritonitis. Hepatology 1988;8:632–5. Runyon BA, Morrissey RL, Hoefs JC, Wyle FA. Opsonic activity of human ascitic fluid: a potentially important protective mechanism against spontaneous bacterial peritonitis. Hepatology 1985;5:634–7. Fiuza C, Salcedo M, Clemente G, Tellado JM. In vivo neutrophil dysfunction in cirrhotic patients with advanced liver disease. J Infect Dis 2000;182:526–33. Jimenez W, Ros J, Morales-Ruiz M, Navasa M, Sole M, Colmenero J, et al. Nitric oxide production and inducable nitric oxide synthase expression in peritoneal macrophages of cirrhotic patients. Hepatology 1999;30:670–6. Such J, Hillebrand DJ, Guarner C, Berk L, Zapater P, Westengard J, et al. Nitric oxide in ascitic fluid is an independent predictor of renal impairment in patients with cirrhosis and spontaneous bacterial peritonitis. Eur J Gastroenterol Hepatol 2004;16:571–7. Frances R, Zapater P, Gonzales-Navajas JM, Munoz C, Cano R, Moreu R, et al. Bacterial DNA in patients with cirrhosis and noninfected ascites mimics the soluble immune response established in patients with spontaneous bacterial peritonitis. Hepatology 2008;47:978–85. Tito L, Rimola A, Gines P, Llach J, Arroyo V, Rodes J. Recurrence of spontaneous bacterial peritonitis in cirrhosis: frequency and predictive factors. Hepatology 1988;8:27–31.
[22] Bernard B, Cadranel JF, Valla D, Escolano S, Jarlier V, Opolon P. Prognostic significance of bacterial infection in bleeding cirrhotic patients: a prospective study. Gastroenterology 1995;108:1828–34. [23] Llach J, Rimola A, Navasa M, Gines P, Salmeron JM, Gines A, et al. Incidence and predictive factors of first episode of spontaneous bacterial peritonitis in cirrhosis with ascites: relevance of ascitic fluid protein concentration. Hepatology 1992;16:724–7. [24] Guarner C, Sola R, Soriano G, Andreu M, Novella MT, Vila MC, et al. Risk of a first community-acquired spontaneous bacterial perionitis in cirrhotics with low ascitic fluid protein levels. Gastroenterology 1999;117:414–19. [25] Obstein KL, Campbell MS, Reddy KR, Yang YX. Association between model for end-stage liver disease and spontaneous bacterial peritonitis. Am J Gastroenterol 2007;102: 2732–6. [26] Shi KQ, Fan YC, Ying L, Lin XF, Song M, Li LF, et al. Risk stratification of spontaneous bacterial peritonitis in cirrhosis with ascites based on classification and regression tree analysis. Mol Biol Rep 2012;39:6161–9. [27] Lata J, Fejfar T, Krechler T, Musil T, Husova L, Senkyrik M, et al. Spontaneous bacterial peritonitis in the czech republic: prevalence and aetiology. Eur J Gastroenterol Hepatol 2003;15:739–43. [28] Solana R, Tarazona R, Gayoso I, Lesur O, Dupuis G, Fulop T. Innate immunosenescence: effect of aging on cells and receptors of the innate immune system in humans. Semin Immunol 2012;24:331–41. [29] Serste T, Bourgeois N, Lebrec D, Evrard S, Deviere J, Le Moine O. Relationship between the degree of portal hypertension and the onset of sponateous bacterial peritonitis in patients with cirrhosis. Acta Gastroenterologica Belgica 2006;69:355–60. [30] Elzey BD, Tian J, Jensen RJ, Swanson AK, Lees JR, Lentz SR, et al. Platelet-mediated modulation of adaptive immunity: a communication link between innate and adaptive immune compartments. Immunity 2003;19:9–19. [31] Aslam R, Speck ER, Kim M, Crow AR, Bang KW, Nestel FP, et al. Platelet Toll-like receptor expression modulates lipopolysaccharide-induced thrombocytopenia and tumor necrosis factor-a production in vivo. Blood 2006; 107:637–41. [32] Clark SR, Ma AC, Tavener SA, McDonald B, Goodarzi Z, Kelly MM, et al. Platelet TLR4 activates neutrophil extracellular traps to ensnare bacteria in septic blood. Nat Med 2007;13:463–9. [33] Kuckleburg CJ, Tiwari R, Czuprynski CJ. Endothelial cell apoptosis induced by bacteria-activated platelets requires caspase-8 and -9 and generation of reactive oxygen species. Thromb Haemost 2008;99:363–72. [34] Kamath PS, Wiesner RH, Malinchoc M, Kremers W, Therneau TM, Kosberg CL, et al. A model to predict survival in patients with end-stage liver disease. Hepatology 2001;33:464–70. [35] Botta F, Giannini E, Romagnoli P, Fasoli A, Malfatti F, Chiarbonello B, et al. MELD scoring system is useful for predicting prognosis in patients with liver cirrhosis and is correlated with residual liver function: a European study. Gut 2003;52:134–9. [36] Huo TI, Wu JC, Lin HC, Lee FY, Hou MC, Lee PC, et al. Evaluation of the increase in model for end-stage liver disease (deltaMELD) score over time as a prognostic predictor in patients with advanced cirrhosis: risk factor analysis and
A scoring system for rapid diagnosis of SBP
Scand J Gastroenterol Downloaded from informahealthcare.com by Universitaet Zuerich on 07/09/14 For personal use only.
comparison with initial MELD and Child-Turcotte-Pugh score. J Hepatol 2005;42:826–32. [37] Pugh RNH, Murray-Lyon IM, Dawson JL, Pietroni MC, Williams R. Transection of the oesophagus for bleeding oesophageal varices. Br J Surg 1973;60:646–9. [38] Albers I, Hartmann H, Bircher J, Creutzfeldt W. Superiority of the Child-Pugh classification to quantitative liver function tests for assessing prognosis of liver cirrhosis. Scand J Gastroenterol 1989;24:269–76. [39] Castellote J, Girbau A, Maisterra S, Charhi N, Ballester R, Xiol X. Spontaneous bacterial peritonitis and bacterascites prevalence in asymptomatic cirrhotic outpatients undergoing large-volume paracentesis. J Gastroenterol Hepatol 2008;23: 256–9.
603
[40] Llovet JM, Planas R, Morillas R, Quer JC, Cabre E, Biox J, et al. Short-term prognosis of cirrhotics with spontaneous bacterial peritonitis: multivariate study. Am J Gastroenterol 1993;88:388–92. [41] Navasa M, Follo A, Llovet JM, Clemente G, Vargas V, Rimola A, et al. Radomized, comparative study of oral ofloxacin versus intravenous cefotaxime in spontaneous bacterial peritonitis. Gastroenterology 1996; 111:1011–17. [42] Koulaouzides A, Leontiadis GI, Abdullah M, Moschos J, Gasem J, Tharakan J, et al. Leucocyte esterase reagent strips for the diagnosis of spontaneous bacterial peritonitis: a systematic review. Eur J Gastroenterol Hepatol 2008;20: 1055–60.
