Research Article

C-Reactive protein predicts short-term mortality in patientswith cirrhosis

Jean-Paul Cervoni1, Thierry Thévenot1, Delphine Weil1, Emilie Muel1, Olivier Barbot1,Frances Sheppard2, Elisabeth Monnet1, Vincent Di Martino1,⇑

1Université de Franche Comté et Service d’hépatologie et de soins intensifs digestifs, Besançon, France; 2Centre d’investigation clinique(Inserm CBT 506), CHRU Jean Minjoz, Besançon, France

Background & Aims: We aimed at improving prediction of short- and CRP level (group A) (HR = 2.73; 95% CI: 1.41–5.26;

term mortality in cirrhotic inpatients by evaluating C-reactiveprotein (CRP) as a surrogate marker of systemic inflammatoryresponse syndrome (SIRS).Methods: One-hundred and forty-eight consecutive cirrhoticpatients with Child-Pugh score PB8 and without hepatocellularcarcinoma were prospectively included and followed for182 days. The primary end point was 6-month survival.Results: Main baseline characteristics were as follows: alcoholicliver disease in 88.5%; bacterial infection in 37%; hepatorenalsyndrome in 7% of cases. CRP range was 1–240 mg/L (median26 mg/L); 42 patients (28.4%) had SIRS as defined by ACCP/SCCM-criteria. CRP levels were higher in patients with SIRS (50vs. 21 mg/L; p <0.0001), infection (46 vs. 27 mg/L; p <0.0001),and alcoholic hepatitis (44 vs. 32 mg/L, p = 0.049). Forty-twopatients died within the first 6 months of follow-up. Short-termmortality was associated with extrahepatic co-morbidities(p = 0.002), high MELD score (p <0.001; AUROC = 0.67), renal fail-ure (p = 0.008), elevated blood lactates (p <0.001), and high base-line CRP levels (p = 0.003; AUROC = 0.63; best cut-off value at29 mg/L). Among patients with baseline CRP P29 mg/L, 32 stillhad CRP P29 mg/L at day 15 (group A). Group A was associatedwith 6-month mortality in the overall population (p <0.001) andalso through sentitivity analyses restricted to patients withoutinfection or alcoholic hepatitis. Multivariate analysis (Cox)adjusted for age identified three predictors of mortality: highMELD score (HR = 1.08; 95% CI: 1.03–1.12; p <0.001), extrahe-patic co-morbidities (HR = 2.51; 95% CI: 1.31–4.84; p = 0.006),

Journal of Hepatology 20

Received 21 June 2011; received in revised form 13 December 2011; accepted 29December 2011; available online 5 February 2012⇑ Corresponding author. Address: Service d’hépatologie, CHRU Besançon, 25000Besançon, France. Tel.: +33 381668421; fax: +33 381668417.E-mail addresses: vdimartino@chu-besancon.fr, vdimartino1@gmail.com (V. DiMartino).Abbreviations: CRP, C-reactive protein; SIRS, systemic inflammatory response s-yndrome; ACCP/SCCM, American College of Chest Physicians/Society of CriticalCare Medicine; MELD, Model of End Stage Liver Disease; HR, Hazard ratio; 95% CI,95% confidence interval; AUROC, area under the receiver operating characteristiccurve; APACHE, acute physiology and chronic health evaluation; SOFA, sequentialorgan failure assessment; US, ultrasonography; CT, computed tomography; SE,standard error; ANOVA, analysis of variance; HCC, hepatocellular carcinoma; PPV,positive predictive value; NPV, negative predictive value; IL6, interleukin 6; Akt,proteine kinase B; INR, International Normalized ratio; Na, natremia.

p = 0.003). The performance of the three variables taken togetherfor predicting death was 0.80 (AUROC).Conclusions: In Child-Pugh score PB8 cirrhotic patients, persis-tent CRP levels P29 mg/L predicted short-term mortality inde-pendently of age, MELD, and co-morbidities, and better thaninfection or clinically-assessed SIRS.� 2012 European Association for the Study of the Liver. Publishedby Elsevier B.V. All rights reserved.

Introduction

In the United States and Europe, liver transplant allocation pro-grams give the highest priority to the sickest patients. This avoidsfutile transplants which have recently been identified as deleteri-ous in a randomized trial [1]. Such a system emphasizes the needfor powerful prognostic markers, able to identify severe cirrhoticpatients with the highest short-term mortality risk. The MELD(Model of End Stage Liver Disease) score is an indicator of theseverity of cirrhosis used to sort patients on the waiting list.However, it was recently shown that short-term prognosisdepends largely on events that temporarily worsen or are super-imposed on liver failure. These events are better taken intoaccount by general prognostic scores reflecting organ failure(such as APACHE II and SOFA) than scores specifically dedicatedto liver failure such as Child-Pugh or MELD scores [2]. In thiscontext, the occurrence of a systemic inflammatory response syn-drome (SIRS) appears to have an important prognostic relevance:SIRS increases the risk of encephalopathy, renal failure, infectionand death during acute [3–5] or chronic [6–8] liver failure. Unfor-tunately, it can be difficult to assess SIRS in cirrhotic patients.Hypersplenism, hyperventilation associated with encephalopa-thy, hyperkinetic circulatory syndrome, or the use of beta-blockers may modify the clinical and biochemical parameters ofSIRS [9]. New markers of SIRS would thus be particularly usefulin this population. The C-reactive protein is synthesized in theacute phase of inflammation mainly in response to interleukin-6. Serum CRP level increases in situations responsible for a SIRS(with or without documented bacterial infection) and is able topredict mortality in different populations of intensive care unitpatients without cirrhosis [10–13]. Although synthesized by theliver, high CRP levels can be observed in the event of advanced

12 vol. 56 j 1299–1304

Table 1. Characteristics according to CRP level outcome between baseline and day 15.

Overall°°(n = 148)

Groups according to CRP assessment at baseline and day 15

Group A°(n = 32)

Group B°(n = 29)

Group C°(n = 84)

p‡

Age (yrs) 58 (56-59) 57 (50-62) 58 (56-60) 58 (54-59) n.s.

Sex (% M) 70.3 65.6 82.8 67.9 n.s.

Alcoholic cirrhosis (%) 88.5 90.6 89.7 88.1 n.s.

Histologically-proven alcoholic hepatitis (%)

22.1 37.5 27.6 14.3 0.019

Extrahepatic co-morbidities (%) 40.5 53.1 34.5 38.1 n.s.

Beta-blockers (%)* 28.9 21.9 27.6 32.1 n.s.

Ascites (%)* 71.7 78.1 69.0 70.2 n.s.

MELD* 22.85 (21.17-23.83) 25.43 (20.25-30.34) 24.27 (20.07-28.94) 21.94 (20.87-23.34) n.s.

ΔMELD** –1.19 (–2.05- –0.43) –0.09 (–1.95- +1.26) –2.68 (–3.41- –1.97) –1.18 (–2.26- –0.37) n.s.

Bilirubin (mg/dl)* 5.3 (4.3-6.1) 11.5 (6.5-15.9) 7.3 (4.2-16.4) 4.0 (3.5-4.7) 0.0003

Creatinine (mg/L)* 9.27 (8.59-10.06) 8.81 (7.57-10.17) 8.59 (7.01-13.67) 9.38 (8.59-10.51) n.s.

INR* 2.20 (2.04-2.36) 2.26 (1.70-2.70) 2.07 (1.76-2.70) 2.27 (2.02-2.40) n.s.

Albumin (g/L)* 25 (23-26) 24 (20-28) 22 (19-26) 25 (24-26) n.s.

Na (mEq/L)* 136 (134-137) 135 (132-138) 134 (131-137) 136 (135-138) n.s.

Lactates (mM)* 2.0 (2.0-2.0) 2.2 (2.0-3.0) 2.0 (1.17-2.30) 2.0 (1.70-2.0) n.s.

Infection (%)*† 37.2 56.3 37.9 31.0 0.043

Infection at day 15 (%)† 14.5 37.5 10.3 7.1 0.0001

SIRS (%)* 28.4 46.9 44.8 14.3 0.0001

Procalcitonin (ng/ml)* 0.27 (0.21-0.32) 0.56 (0.30-0.74) 0.36 (0.26-0.80) 0.17 (0.13-0.22) n.s.

CRP level (mg/L)* 26 (19-29) 65 (54-78) 50 (39-54) 12 (9-14) <0.0001

Quantitative variables are expressed as median and 95% confidence interval.n.s., not significant.�Group A: CRP P29 mg/L at baseline and day 15; Group B: CRP P29 mg/L at baseline and <29 mg/L at day 15; Group C: CRP <29 mg/L at baseline.��Including three patients who died within 15 days and are not included in group A, B, or C.⁄At baseline.⁄⁄Variation of MELD score between baseline and day 15.�Overt bacterial infection.�Assessed by Chi-square test or ANOVA.

Research Article

liver failure [14,15]. The aim of our study was to assess the valueof CRP for the diagnosis of SIRS and for prediction of short-termmortality in patients with severe cirrhosis. We focused on thevariation of CRP level within the first 15 days to better identifypatients with sustained systemic inflammation.

Patients and methods

Patients

The protocol for this study was approved by the Besançon University HospitalInstitutional Review Board. All consecutive cirrhotic patients admitted betweenOctober 2008 and August 2010 in our Hepatology Department were prospectivelyincluded if they fulfilled the following criteria: (1) cirrhosis either histologically-proven or considered obvious by clinical, biochemical and morphological criteria;(2) Child-Pugh stage above B7; (3) and no hepatocellular carcinoma. There wereno exclusion criteria regarding age or extrahepatic co-morbidities.

Data collection

One-hundred and fifty-eight variables were recorded at admission (n = 148patients) and during follow-up on day 15 (n = 135), month 3 (n = 99) and month6 (n = 85). These concerned demographics, cause and history of liver disease,

1300 Journal of Hepatology 2012

relevant co-morbidities as assessed by the Charlson co-morbidity index [16](including respiratory, cardiovascular, renal, neuropsychiatric disorders, extrahe-patic cancers, and diabetes), reason for admission, clinical status on admissionincluding the SIRS criteria, liver biochemistry, assessment of renal function,abdominal US or CT scan, chest X-ray, systematic culture of blood, urine, and asci-tes (if any), and measurement of serum routine CRP and procalcitonin.

Statistical analyses

Quantitative variables were expressed by their mean ± SE or median (with the 95%CI) in the event of abnormal distribution. Comparisons of subgroups of patientsregarding presence of SIRS or CRP level were performed using the Chi-square, Stu-dent t test and ANOVA. Correlations between CRP level and MELD score were eval-uated by the Spearman rank test. Diagnostic performance of CRP for predicting SIRSand mortality was assessed using the area under the receiver operating curve(AUROC). We chose to use the Youden index, calculated as the maximum (sensitiv-ity + specificity-1) value, to estimate the best discriminate threshold.

The primary outcome measure for prognosis was the 6-month mortality rate.Factors affecting prognosis were evaluated using the Kaplan–Meier model andlog-rank test for univariate analyses, and the proportional Hazard Regression(Cox) model for multivariate analysis. To include in the model the variation ofCRP level within the first 15 days as a predictive factor of mortality, day 15 wasconsidered as the starting point and patients who died within the first 15 dayswere not included in the survival analysis. Patients transplanted during follow-up were included and considered as alive with the liver transplant date markingthe end of follow-up. A logistic regression model was built to evaluate the AUROCof the independent predictors of 6-month mortality taken together. The c index

vol. 56 j 1299–1304

ivity

0.75

1.00

MELD

CRP(AUROC = 0.63)

JOURNAL OF HEPATOLOGY

given by the logistic regression model could be used to calculate for each patientthe probability p of dying within 6 months by using the following formula:p = 1/(1 + exp(�c)). The performances of MELD, baseline CRP, and multivariatemodel for predicting short-term mortality were compared using the comparisonof the AUROCs. All statistical analyses were performed using the NCSS package forWindows.

1-Specificity

Sen

sit

0.000.00

0.25

0.25

0.50

0.50 0.75 1.00

(AUROC = 0.67)

3 variable model(AUROC = 0.80)

ig. 1. ROC curves of MELD, baseline C-reactive protein (CRP) levels, and 3-ariable model for the prediction of the 6-month mortality. The 3-variableodel included MELD, extrahepatic co-morbidities, and variation of CRP level

etween baseline and day 15 (Group A was defined by CRP P29 mg/L at baselinend day 15) as follows: estimate = 1.2518 � (co-morbidity (1 if present; 0 ifbsent)) + 1.449 � (group A (1 or 0)) + 8.028E-02 � (MELD) � 3.828. The 3-variableodel was significantly better than baseline CRP level (p <0.001) and baselineELD score (p = 0.014) in predicting the 6-month mortality, whereas AUROCs of

RP and MELD were not significantly different.

322984

242874

192566

162261

152058

141952

141648

Days

Sur

vivi

ng (%

)

00 30

25

60

50

90

75

120

100

150 180

p <0.001

Group A

Group A

Group C

Group C

Group B

Group B

Patients at risk

Fig. 2. Short-term mortality according to CRP level at baseline and day 15.Group A: CRP P29 mg/L at baseline and day 15; Group B: CRP P29 mg/L atbaseline and <29 mg/L at day 15; Group C: CRP <29 mg/L at baseline. Amonggroup C patients, only nine patients had CRP 29 mg/L at day 15 and were notseparately analyzed due to the low number of events.

Results

Study population

The main characteristics of the studied population are summa-rized in Table 1. One hundred and forty-eight cirrhotic patientswith Child-Pugh score higher than B7 and without HCC were con-secutively admitted to our institution and enrolled in this cohort.There were 105 men with a median age of 58 (95% CI: 56–59 yrs).The median MELD was 22.85 (95% CI: 21.17–23.83). The mainbaseline characteristics were as follows: alcoholic liver diseasein 131 patients (88.5%); documented bacterial infection in 55(37.2%), including 15 spontaneous peritonitis and nine bactere-mia; hepatorenal syndrome in 11 (7.4%); and CRP value rangedfrom 1 to 240 mg/L (median 26 mg/L, 95% CI: 19–29). Forty-twopatients (28.4%) had SIRS as defined by the American College ofChest Physicians/Society of Critical Care Medicine (ACCP/SCCM)Consensus Conference [9]. Sixty patients (40.5%) had extrahepaticco-morbidities, including 23 diabetes mellitus, 19 cardiomyopa-thy, 16 atherosclerosis, 15 extrahepatic malignancies, eightdementia, seven respiratory failure, seven HIV infection, and foursystemic inflammatory diseases.

Parameters associated with baseline CRP level

Baseline CRP level was higher in patients with infection (46 vs.27 mg/L; p <0.001), SIRS (53 vs. 27 mg/L; p <0.001), and alcoholichepatitis (44 vs. 32 mg/L, p = 0.049). Eight (14.8%) patients withbacterial infection, five (11.9%) patients with SIRS, and three(9.7%) patients with alcoholic hepatitis had CRP 610 mg/L. Con-versely, 62.5% of patients with baseline CRP >10 mg/L had noinfection, SIRS, or alcoholic hepatitis. CRP value was not signifi-cantly higher in the event of extrahepatic co-morbidities (39 vs.31 mg/L; p = 0.12), or hepatorenal syndrome (46 vs. 34 mg/L;p = 0.29). The CRP level was positively correlated with the biliru-bin level (Spearman-R = 0.413; p <0.001), heart rate (R = 0.347;p <0.001), and leukocyte count (R = 0.475, p <0.001). There wereno correlations between CRP level and MELD score (R = 0.158;p = 0.081), INR (R = �0.069, p = 0.66), natremia (R = �0.144,p = 0.35), creatinine (R = �0.053, p = 0.35), and albumin(R = �0.182, p = 0.82).

Factors associated with SIRS

SIRS was associated with overt bacterial infection (43.0% vs.19.4%, p = 0.002), high procalcitonin (0.35 vs. 0.24 mg/L,p <0.001) and high CRP levels (53 vs. 27 mg/L, p <0.001). TheAUROC of CRP for the diagnosis of SIRS was 0.73 (95% CI: 0.62–0.81), and was significantly higher (p = 0.027) than the AUROCof procalcitonin (0.65 (95% CI: 0.50–0.75)) for the same diagnosis.The best predictive value of CRP for SIRS given by the Youdenindex was 26 mg/L, with sensitivity, specificity, positive predic-tive value (PPV), and negative predictive value (NPV) of 76.2%,63.8%, 45.7%, and 87.0%, respectively. Normal CRP level

Journal of Hepatology 2012

FvmbaamMC

(610 mg/L) had a NPV for SIRS of 86.5%. High thresholds of CRPlevels had PPV for SIRS that did not exceed 75%.

Follow-up examination and factors associated with short-termmortality

During a median follow-up of 182 days (95% CI: 134–208; range:3–1580 days), 57 (38.5%) patients died and 16 were transplanted.Thirty-five patients (23.6%) died within 3 months of follow-up,and 42 (28.4%) died within 6 months of follow-up. Of these, threepatients died within the first 15 days. Death was liver-related in36 cases: variceal bleeding in three cases, portopulmonaryhypertension in two cases, and end stage cirrhosis in 31 cases

vol. 56 j 1299–1304 1301

Table 2. Multivariate analysis of factors associated with short-term mortality.

Variables Cox model on 6-month mortality42 events, R² = 0.2016

HR 95% CI HR p value % variance explained (pseudo-R2)Age° 1.03 0.99-1.06 0.0678 0.0857MELD° 1.08 1.03-1.12 0.0005 0.2529CRP (Group A*) 2.73 1.41-5.26 0.0028 0.2009Extrahepatic co-morbidities† 2.51 1.31-4.84 0.0059 0.1758

�Included as a continuous quantitative variable.⁄Group A was defined according to CRP outcome between baseline and day 15 as follows: CRP >29 mg/L at baseline and CRP >29 mg/L at day 15.�At least one extrahepatic co-morbidity.HR, hazard ratio; 95% CI, 95% confidence interval.

Research Article

(including 16 cases of sepsis and four cases of hepatorenal syn-drome). The remaining causes of death were extrahepatic cancerin two cases, extrahepatic co-morbidity in two cases, or unknownin two cases.

Univariate analyses

The 6-month mortality was associated with high MELD score(26.1 vs. 22.0, p <0.001), high INR (2.24 vs. 1.81, p <0.001), highbilirubin (199 vs. 133 lMol/L, p = 0.010), high serum creatinine(122 vs. 92 lMol/L, p = 0.008), elevated blood lactates (2.83 vs.1.75, p <0.001), and high baseline CRP levels (47 vs. 29 mg/L,p = 0.003). Death also occurred more frequently within 6-monthfollow-up in cases of extrahepatic co-morbidities (42.4% vs.19.3%, p = 0.002), particularly cardiac failure (53.8% vs. 26.1%,p = 0.035), respiratory failure (71.4% vs. 26.4%, p = 0.010), demen-tia (75% vs. 26.1%, p = 0.003) or diabetes (47.8% vs. 25.0%,p = 0.026) whereas alcoholic hepatitis (40.6% vs. 25.2%, p =0.088), natremia, albumin, procalcitonin level, bacterial infectionat baseline (29.1% vs. 28.3%, n.s.) or at day 15 (42.9% vs. 26.0%,n.s.), or SIRS as defined by the ACCP/SCCM criteria (29.5% vs.26.2%, n.s.) had no significant influence on the 6-month mortalityin this series. The performance of the MELD score for predicting6-month mortality was 0.67 (AUROC; 95% CI: 0.56–0.76). Theperformance of baseline CRP level for the same prediction wasnot significantly different to that of the MELD score: AUROCwas 0.63 (95% CI: 0.51–0.72; Fig. 1). The best cut-off value ofCRP for predicting 6-month mortality given by the Youden indexwas 29 mg/L, with a sensitivity, specificity, PPV, NPV of 54.8%,64.8%, 38.3%, and 78.2%, respectively. Baseline CRP levels werealso able to predict 3-month mortality with a similar perfor-mance to that of the MELD score (data not shown).

Considering the 29 mg/L cut-off value of CRP at baseline andat day-15, three groups of patients were defined: patients withCRP level persistently higher than 29 mg/L (group A, n = 32);patients with baseline CRP level P29 mg/L with subsequentdecrease below 29 mg/L (group B, n = 29); and patients withCRP level <29 mg/L at baseline (group C, n = 84, including onlynine patients with CRP P29 mg/L at day 15 who were not sepa-rately analyzed). The 3-month and 6-month survival rates were48.1% and 45.1%, 86.1% and 77.6%, and 84.0% and 78.4%, in groupA, group B, and group C patients, respectively (p <0.001, Fig. 2). Ingroup A patients, SIRS, infection at baseline or at day 15, histolog-ically proven alcholic hepatitis, and high bilirubin levels weremore common (Table 1). No significant relationship was observedbetween group A and sex, age, natremia, creatinine, blood

1302 Journal of Hepatology 2012

lactates, MELD score and its variation within baseline and day15, and Pugh score. (Table 1). Identifying group A patientsprovided a prediction of 6-month mortality with a sensitivity,specificity, PPV and NPV of 41.0%, 84.9%, 50.0%, and 79.6%, respec-tively. This therefore increased specificity and PPV by 20.1% and11.7%, respectively, as compared with the single measurementof baseline CRP level with the 29 mg/L cut-off value.

Multivariate analyses

Multivariate survival analysis using Cox model adjusted for ageidentified three predictors of short-term mortality (Table 2):MELD score (HR:1.08; 95% CIHR: 1.03–1.12; p <0.001), presenceof at least one extrahepatic co-morbidity (HR = 2.51; 95% CI:1.31–4.84; p = 0.006), and belonging to group A (HR = 2.73; 95%CI: 1.41–5.26; p = 0.003). The performance of these three vari-ables taken together in a multiple logistic regression analysisfor predicting 6-month mortality was 0.80 (AUROC; 95% CI:0.73–0.87). It was significantly better than the performance ofthe baseline CRP level (p <0.001) and baseline MELD score(p = 0.014). The 3-variable model estimate was: c = 1.2518 �(co-morbidity (1 if present; 0 if absent)) + 1.449 � (group A (1or 0)) + 8.028E-02 � (MELD) � 3.828 (Fig. 1). For example, apatient with a MELD score at 20, with no co-morbidity and a withCRP level <29 mg/L had an estimate at �2.22 and a probability of6-month mortality at 0.10. For the same MELD score, theprobability of 6-month mortality increased to 0.27 in cases ofco-morbidity with CRP <29 mg/L, 0.32 in group A patientswithout co-morbidity, and 0.62 in group A patients with co-morbidity.

Sensitivity analyses

Sensitivity analyses were performed to investigate whether beingin group A predicted 6-month mortality when excluding patientswith histologically-proven alcoholic hepatitis, bacterial infection,or both. When restricting survival analysis to patients withoutalcoholic hepatitis (n = 112), group A patients had lower 6-monthsurvival (43.5%), than group B (84.8%) and group C (80.6%)patients, p <0.001. When restricting survival analysis to patientswithout bacterial infection (n = 92), 6-month survival was47.1%, 74.9%, and 76.3% in group A, group B and group C patients,respectively (p = 0.014). When restricting the analysis to the 78patients without bacterial infection or alcoholic hepatitis, the 6-month survival was 46.2%, 82.1%, and 79.8% in group A, group Band group C patients, respectively (p = 0.011). Through

vol. 56 j 1299–1304

JOURNAL OF HEPATOLOGY

multivariate analyses, group A and MELD still had independentimpact on short-term mortality.

Discussion

In this prospective study, we observed that CRP was able to pre-dict short-term mortality of patients with severe cirrhosis asdefined by a Child-Pugh score B8 or higher. The prognostic valueof CRP was independent of the usual prognostic criteria such asMELD scores. High CRP levels were associated with SIRS, infec-tion, and alcoholic hepatitis. However, in our series, the majorityof patients with CRP >10 mg/L did not present with any of thesethree events. More importantly, these three events failed to pre-dict 6-month mortality, unlike CRP. We hypothesize that elevatedCRP value could be a surrogate marker of systemic inflammationthat was underestimated by the common criteria of SIRS in ourpopulation. The involvement of interleukin-6 in the activationof a proinflammatory response during cirrhosis is well estab-lished [17] and may be favored by sustained stimulation causedby the lipopolysaccharide-binding protein [18]. Hence, weassume that measuring CRP is a simple and accurate way of diag-nosing systemic inflammation and has a relevant impact on prog-nosis in cirrhotic patients, since CRP actually reflects IL6synthesis [19]. Moreover, repeating the measurement of CRP atday 15 helped improve the prediction of short-term mortalityby eliminating cases of quickly resolved bacterial infection andby identifying persistent systemic inflammation that stronglyimpacts on prognosis.

There is growing evidence that systemic inflammation iscommon in patients with advanced cirrhosis and portal hyper-tension. Inflammatory response activation may be caused byoccult infections associated with bacterial translocation thatcomplicates the increase of intestinal permeability in thesepatients [20]. The incidence of bacterial translocation may belowered by administering preventive antibiotherapy or beta-blockers, which shorten the intestinal transit time and thusdecrease bacterial overgrowth [21,22]. An excessive proinflam-matory response has also been recently described in cirrhoticpatients, due to a defect in Akt-mediated glycogen-synthasekinase 3 phosphorylation [23].

From a pragmatic point of view, the key point is that systemicinflammation, with or without documented bacterial infection, isassociated with a negative outcome in cirrhotic patients [6–8]. Arecent systematic review of 11,987 cirrhotic patients showed afourfold increase in the risk of mortality in the event of bacterialinfection, which was more marked in studies conducted before2000 [24]. Systemic inflammation has been shown to favor seri-ous complications such as variceal bleeding, encephalopathy andacute-on-chronic liver failure [6–8]. Accordingly, Thabut et al. [8]showed that inflammation increases the risk of complicationsand death in cirrhotic patients with acute renal damage. Cazzan-iga and colleagues confirmed and extended these conclusions toinpatients with decompensated cirrhosis even in the absence ofrenal failure [7].

Not all the mechanisms involved in the prognostic conse-quences of SIRS in cirrhotic patients are fully understood. SIRSmay induce hemodynamic changes, in particular an increase inportal hypertension and a reduction of hepatic blood flow [25],that are associated with an accumulation of a NO-synthase inhib-itor, the asymmetric dimethylarginine [26]. More recently, it has

Journal of Hepatology 2012

been suggested that SIRS was associated with a change in thefunctional capacities of albumin, caused by an accumulation ofoxidative albumin that may account for a higher risk of mortalityper se [27].

We have shown that a simple measurement of CRP at baselineand at day 15 provided a powerful tool for predicting short-termmortality in hospitalized cirrhotic patients. Interestingly, highbaseline CRP levels could remain elevated over time despite theresolution of bacterial infection in the majority (62.5%, seeTable 1) of patients. This suggests that systemic inflammationmay become persistent and acts as an autonomous disease [2],and the prognostic value of CRP levels persistently higher than29 mg/L was still significant in sensitivity analyses restricted topatients without infection or alcoholic hepatitis. Conversely, theprognosis was significantly better in patients who experiencedreduction in the CRP levels in our series (Fig. 2). Another interest-ing point was that the negative impact on prognosis that could beobserved for small increases in CRP levels, (over 29 mg/L), may beconsidered irrelevant in clinical practice, especially given thepoor condition of the studied population.

Our study had several limitations. First, we failed to demon-strate a significant relationship between infection or clinically-assessed SIRS and survival, unlike other reports. Nonetheless,our population received up-to-date curative and prophylacticantibiotic strategies, and our study was based on all the ACCP/SCCM criteria of SIRS [9] that can be misinterpreted in the eventof cirrhosis. Second, our results would benefit from external val-idation with an independent cohort. We believe such a validationshould be conducted by establishing a new ‘‘MELD-CRP’’ score,since the prognostic value of CRP was independent of MELDscores in our series. Moreover, CRP was much more predictivethan other variables taken separately, such as creatinine [28] orhyponatremia [29], despite previous descriptions of their impacton prognosis. The challenge will be to better allocate grafts to thesickest patients. In the context of a transplant indication, how-ever, the first concern should be to carefully rule out an overtbacterial infection.

In conclusion, we have shown that the CRP value may identifypatients who have a higher short-term risk of mortality. Wehypothesized that persistently elevated CRP levels highlight theexistence of systemic inflammation that may be underestimatedby the criteria of SIRS but have a great significance for prognosis.This prognostic effect may be used alongside other discriminatingparameters such as MELD, and a ‘‘MELD-CRP’’ score should beevaluated.

Conflict of interest

The authors who have taken part in this study declared that theydo not have anything to disclose regarding funding or conflict ofinterest with respect to this manuscript.

Authors’ contributions

Study concept and design: V.D.M., J.P.C.; acquisition of data: J.P.C.,O.B., D.W., E.M., T.T.; analysis and interpretation of data: V.D.M.,J.P.C., E.M., T.T.; drafting of the manuscript: V.D.M., J.P.C., M.S.;Statistical analysis: V.D.M.; critical revision of the manuscriptfor important intellectual content: J.P.C., T.T., E.M., V.D.M.

vol. 56 j 1299–1304 1303

Research Article

References

[1] Vanlemmens C, Di Martino V, Milan C, Messner M, Minello A, Duvoux C, et al.Immediate listing for liver transplantation versus standard care for Child-Pugh stage B alcoholic cirrhosis: a randomized trial. Ann Intern Med2009;150:153–161.

[2] Malik R, Mookerjee RP, Jalan R. Infection and inflammation in liver failure:two sides of the same coin. J Hepatol 2009;51:426–429.

[3] Leithead JA, Ferguson JW, Bates CM, Davidson JS, Lee A, Bathgate AJ, et al. Thesystemic inflammatory response syndrome is predictive of renal dysfunctionin patients with non-paracetamol-induced acute liver failure. Gut2009;58:443–449.

[4] Rolando N, Wade J, Davalos M, Wendon J, Philpott-Howard J, Williams R. Thesystemic inflammatory response syndrome in acute liver failure. Hepatology2000;32:734–739.

[5] Vaquero J, Polson J, Chung C, Helenowski I, Schiodt FV, Reisch J, et al.Infection and the progression of hepatic encephalopathy in acute liverfailure. Gastroenterology 2003;125:755–764.

[6] Shawcross DL, Davies NA, Williams R, Jalan R. Systemic inflammatoryresponse exacerbates the neuropsychological effects of induced hyperam-monemia in cirrhosis. J Hepatol 2004;40:247–254.

[7] Cazzaniga M, Dionigi E, Gobbo G, Fioretti A, Monti V, Salerno F. The systemicinflammatory response syndrome in cirrhotic patients: relationship withtheir in-hospital outcome. J Hepatol 2009;51:475–482.

[8] Thabut D, Massard J, Gangloff A, Carbonell N, Francoz C, Nguyen-Khac E,et al. Model for end-stage liver disease score and systemic inflammatoryresponse are major prognostic factors in patients with cirrhosis and acutefunctional renal failure. Hepatology 2007;46:1872–1882.

[9] American College of Chest Physicians/Society of Critical Care MedicineConsensus Conference: definitions for sepsis and organ failure and guide-lines for the use of innovative therapies in sepsis. Crit Care Med1992;20:864–874.

[10] Castelli GP, Pognani C, Meisner M, Stuani A, Bellomi D, Sgarbi L. Procalcitoninand C-reactive protein during systemic inflammatory response syndrome,sepsis and organ dysfunction. Crit Care 2004;8:R234–R242.

[11] Chauveau P, Level C, Lasseur C, Bonarek H, Peuchant E, Montaudon D, et al.C-reactive protein and procalcitonin as markers of mortality in hemodialysispatients: a 2-year prospective study. J Ren Nutr 2003;13:137–143.

[12] Lavrentieva A, Kontakiotis T, Lazaridis L, Tsotsolis N, Koumis J, Kyriazis G,et al. Inflammatory markers in patients with severe burn injury. What is thebest indicator of sepsis? Burns 2007;33:189–194.

[13] Wanner C, Metzger T. C-reactive protein a marker for all-cause andcardiovascular mortality in haemodialysis patients. Nephrol Dial Transplant2002;17:29–32, [discussion 9–40].

[14] Park WB, Lee KD, Lee CS, Jang HC, Kim HB, Lee HS, et al. Production of C-reactive protein in Escherichia coli-infected patients with liver dysfunctiondue to liver cirrhosis. Diagn Microbiol Infect Dis 2005;51:227–230.

[15] Bota DP, Van Nuffelen M, Zakariah AN, Vincent JL. Serum levels of C-reactiveprotein and procalcitonin in critically ill patients with cirrhosis of the liver. JLab Clin Med 2005;146:347–351.

1304 Journal of Hepatology 2012

[16] Charlson M, Szatrowski TP, Peterson J, Gold J. Validation of a combined co-morbidity index. J Clin Epidemiol 1994;47:1245–1251.

[17] Rosenbloom AJ, Pinsky MR, Bryant JL, Shin A, Tran T, Whiteside T. Leukocyteactivation in the peripheral blood of patients with cirrhosis of the liver andSIRS. Correlation with serum interleukin-6 levels and organ dysfunction.JAMA 1995;274:58–65.

[18] Albillos A, de la Hera A, Gonzalez M, Moya JL, Calleja JL, Monserrat J, et al.Increased lipopolysaccharide binding protein in cirrhotic patients withmarked immune and hemodynamic derangement. Hepatology2003;37:208–217.

[19] Heikkila K, Ebrahim S, Rumley A, Lowe G, Lawlor DA. Associations ofcirculating C-reactive protein and interleukin-6 with survival in womenwith and without cancer: findings from the British Women’s Heart andHealth Study. Cancer Epidemiol Biomarkers Prev 2007;16:1155–1159.

[20] Cirera I, Bauer TM, Navasa M, Vila J, Grande L, Taura P, et al. Bacterialtranslocation of enteric organisms in patients with cirrhosis. J Hepatol2001;34:32–37.

[21] Perez-Paramo M, Munoz J, Albillos A, Freile I, Portero F, Santos M, et al. Effectof propranolol on the factors promoting bacterial translocation in cirrhoticrats with ascites. Hepatology 2000;31:43–48.

[22] Thevenot T, Cervoni JP, Monnet E, Sheppard F, Martino VD. Is this really theend of beta-blockers in patients with cirrhosis and refractory ascites?Hepatology 2011;53:715–716.

[23] Coant N, Simon-Rudler M, Gustot T, Fasseu M, Gandoura S, Ragot K, et al.Glycogen synthase kinase 3 involvement in the excessive proinflammatoryresponse to LPS in patients with decompensated cirrhosis. J Hepatol2011;55:784–793.

[24] Arvaniti V, D’Amico G, Fede G, Manousou P, Tsochatzis E, Pleguezuelo M,et al. Infections in patients with cirrhosis increase mortality four-fold andshould be used in determining prognosis. Gastroenterology2010;139:1246–1256, 56 e1–5.

[25] Jalan R, Mookerjee R. Systemic hemodynamic, hepatic blood flow and portalpressure in patients with cirrhosis and multiorgan failure: the role ofsympathetic activation. Hepatology 2008;48:1077A.

[26] Mookerjee RP, Malaki M, Davies NA, Hodges SJ, Dalton RN, Turner C, et al.Increasing dimethylarginine levels are associated with adverse clinicaloutcome in severe alcoholic hepatitis. Hepatology 2007;45:62–71.

[27] Jalan R, Schnurr K, Mookerjee RP, Sen S, Cheshire L, Hodges S, et al.Alterations in the functional capacity of albumin in patients with decom-pensated cirrhosis is associated with increased mortality. Hepatology2009;50:555–564.

[28] Fernandez-Esparrach G, Sanchez-Fueyo A, Gines P, Uriz J, Quinto L, VenturaPJ, et al. A prognostic model for predicting survival in cirrhosis with ascites. JHepatol 2001;34:46–52.

[29] Kim WR, Biggins SW, Kremers WK, Wiesner RH, Kamath PS, Benson JT, et al.Hyponatremia and mortality among patients on the liver-transplant waitinglist. N Engl J Med 2008;359:1018–1026.

vol. 56 j 1299–1304