Serum procalcitonin and C-reactive protein levels as markers of bacterial infection in patients with liver cirrhosis: a systematic review and meta-analysis Kuan-Ho Lin, Feng-Lin Wang, Meng-Shu Wu, Bing-Yan Jiang, WeiLiang Kao, Hsiao-Yun Chao, Jiunn-Yih Wu, Chien-Chang Lee PII: DOI: Reference:

S0732-8893(14)00209-0 doi: 10.1016/j.diagmicrobio.2014.03.029 DMB 13620

To appear in:

Diagnostic Microbiology and Infectious Disease

Received date: Revised date: Accepted date:

27 October 2013 1 March 2014 5 March 2014

Please cite this article as: Lin Kuan-Ho, Wang Feng-Lin, Wu Meng-Shu, Jiang Bing-Yan, Kao Wei-Liang, Chao Hsiao-Yun, Wu Jiunn-Yih, Lee Chien-Chang, Serum procalcitonin and C-reactive protein levels as markers of bacterial infection in patients with liver cirrhosis: a systematic review and meta-analysis, Diagnostic Microbiology and Infectious Disease (2014), doi: 10.1016/j.diagmicrobio.2014.03.029

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT Serum procalcitonin and C-reactive protein levels as markers of bacterial infection in patients with liver cirrhosis: a systematic review and meta-analysis Kuan-Ho Lin MD, 3Feng-Lin Wang MD, 3Meng-Shu Wu MD, 3Bing-Yan Jiang MD, 3Wei-Liang Kao MD, 3Hsiao-Yun Chao MD, 4Jiunn-Yih Wu MD, 5,6Chien-Chang Lee MD MSc

RI P

T

1,2

1

College of Medicine, China Medical University, Taichung, 40402, Taiwan Department of Emergency Medicine, China Medical University Hospital, Taichung, 40402, Taiwan 3 Department of Emergency Medicine, Chang Gung Memorial Hospital, Taoyuan,Chang Gung University College of Medicine, Taoyuan, Taiwan.

SC

2

MA NU

4

Department of Emergency Medicine, Chang Gung Memorial Hospital, Keelung, Taiwan and Chang Gung University College of Medicine, Taoyuan, Taiwan. 5 Department of Emergency Medicine, National Taiwan University Hospital Yunlin Brach, Yunlin, Taiwan

PT

ED

Running title: Value of PCT or CRP in liver function impairment patients Word count: 2727 Conflict of interest: None declared. Funding: None declared

AC

CE

Address correspondence to: Chien-Chang Lee MD, MSc Email: [email protected] Postal office: No 579 Sec 2 Yunlin Road, Douliou, Yunlin County 640, Taiwan Telephone: +8865532391 ext 2326 Fax: +88655341452 Jiunn-Yih Wu MD E-mail: [email protected] Postal address: 3F, No.3,Ln. 325, Sec. 2, Shipai Rd, Beitou Dist, Taipei City 11267, Taiwan Telephone: +88633281200 ext 2505 Fax: +88633287715

1

ACCEPTED MANUSCRIPT Abstract

T

Background

SC

RI P

The diagnostic value of procalcitonin (PCT) for patients with liver cirrhosis is unclear.

Methods

MA NU

We searched the PubMed, EMBASE, and Cochrane databases for studies published through December 2013 that evaluated the diagnostic performance of PCT for patients with acute or chronic liver disease with suspected systemic infection. We summarized the test performance characteristics by using forest

ED

plots, hierarchical summary receiver operating characteristic (HSROC) curves, and bivariate random

AC

Results

CE

PT

effects models.

Our search identified 230 citations, of which 10 diagnostic studies that evaluated 1144 patients and 435 bacterial infection episodes (32.1%) were ultimately included for analysis. The bivariate pooled sensitivity estimates were 79% (95% CI: 64%–89%) for PCT tests and 77% (95% CI: 69%–84%) for CRP tests. Pooled specificity estimates were higher for both PCT and CRP tests (PCT, 89% [95% CI: 82%–94%]; CRP, 85% [95% CI: 76%–90%]). The positive likelihood ratio for PCT (LR+ 7.38, 95% CI: 4.70-11.58) was sufficiently high to qualify PCT as a rule-in diagnostic tool, and the negative likelihood ratio for CRP was sufficiently low to qualify CRP as an acceptable rule-out diagnostic tool (LR- 0.23, 2

ACCEPTED MANUSCRIPT 95% CI: 0.13-0.41) in patients with no signs of infection. Available clinical evidence showed that PCT

RI P

T

has comparable accuracy to CRP for the diagnosis of systemic infection in patients with liver cirrhosis.

SC

Conclusions

Compared with patients with normal liver function, both PCT and CRP tests have acceptable accuracy

MA NU

for diagnosing bacterial infection among patients with liver cirrhosis.

PT

ED

Keywords: C-reactive protein; chronic liver disease; procalcitonin

Introduction

CE

The liver plays an important role in host defenses against infections through the production of all

AC

leukocyte lineages from resident hematopoietic stem cells, the production of acute phase proteins and complement, and the activation of cytotoxicity against microbe-infected cells upon bacterial stimulation. Patients with chronic liver dysfunction may thus be more to susceptible to infections.1 In cirrhotic patients, the translocation of bacteria or bacterial products, such as endotoxins and/or bacterial DNA through the intestinal wall usually leads to systemic inflammatory reactions with renal and hepatic failure, encephalopathy, circulatory collapse, and ultimately death.2 Early identification of bacterial infections in patients with chronic liver disease is sometimes challenging. The clinical presentation is usually vague and atypical and the leukocyte count is difficult to interpret due to chronic hypersplenism 3

ACCEPTED MANUSCRIPT with pancytopenia.3 In this context, the biomarkers for acute inflammatory response, such as C-reactive

T

protein (CRP) and procalcitonin (PCT) may prove useful in the early stages of sepsis before positive

SC

RI P

microbiological identification can be obtained.4,5

Both CRP and PCT are well known acute phase reactant proteins. CRP has long been known to play a

MA NU

role in the complement system and has important opsonization properties in relation to bacterial infection.6 CRP is synthesized mainly in the liver, while PCT is produced by most parenchymal tissues throughout the body during acute bacterial infection.7 Both CRP and PCT have been reported to change

ED

in chronic liver diseases and cirrhosis.8 Altered serum levels of PCT and CRP may affect diagnostic

PT

accuracy in septic patients, but only a few studies to date have investigated the diagnostic value of these

CE

biomarkers in patients with liver disease, and their diagnostic accuracy and the cut-off values were

AC

highly variable.9-16 Therefore, we intended to search the available evidence and pool the qualified studies to determine the diagnostic and discriminative role of serum PCT and CRP in patients with chronic liver disease.

Methods Search strategy and selection criteria We searched 3 electronic databases (Medline, EMBASE, and Cochrane) for studies published through December 2013 with the following MeSH terms and free text: procalcitonin, liver cirrhosis, chronic 4

ACCEPTED MANUSCRIPT liver disease, and liver. We did not set any date or language restrictions for these searches. We checked

T

the reference lists of relevant review articles. The selection was performed independently by 2 reviewers.

RI P

Discrepancies between the reviewers were resolved by a consensus meeting with the third and fourth

SC

coauthor.

MA NU

The title and abstract of the studies were screened in the first round, and potentially relevant articles were retrieved for full-text review in the second round. For inclusion, the studies were required to fulfill the following criteria: (1) Contain a study population of adult patients (>18 year old) with liver cirrhosis

ED

or liver disease; (2) include the results of a PCT or CRP test; (3) use infection as one of the endpoints;

PT

and (4) include calculations for sensitivity, specificity, or odds ratio or have sufficient data to construct a

CE

2×2 contingency table. We excluded case reports, case series, review articles, editorials, and clinical

AC

guidelines. Two authors independently assessed all the titles and abstracts to determine whether the inclusion criteria were satisfied. Full-text articles were retrieved if any of the reviewers considered the abstract suitable. The study inclusion and exclusion process is summarized in Figure 1.

Quality assessment The methodological quality of the selected studies was evaluated independently by two reviewers with a validated tool for the quality assessment of diagnostic accuracy studies (Quality Assessment of Diagnostic Accuracy Studies, QUADAS).17 Discrepancies were resolved by a consensus meeting with 5

ACCEPTED MANUSCRIPT

T

the third and fourth coauthors.

RI P

Data synthesis and analysis

SC

We calculated the pooled sensitivity, specificity, likelihood ratio, and diagnostic odds ratio (DOR) for the included studies. For diagnostic test studies, different thresholds were used in different studies, and

MA NU

there was a negative correlation between the sensitivity and specificity due to the threshold effects. Separate pooling of sensitivity and specificity may lead to biased results. Therefore, we used the bivariate model to adjust for the correlation between sensitivity and specificity and estimate the

ED

diagnostic accuracy parameters.18 The bivariate approach assumes a bivariate distribution for the

PT

logit-transformed sensitivity and specificity and uses a general estimating equation to adjust for the

CE

covariance between the correlated parameters.18 To compare the diagnostic performance between 2

AC

biomarkers, we calculated the area under the summary receiver operating characteristic (AUROC) curve and the diagnostic odds ratio (OR) to summarize the true- and false-positive rates of different diagnostic studies, irrespective of the cutoff points used in the various studies. To address zero observations in 2×2 contingency tables, 0.5 was added to each cell, reducing the performance in the small studies. To quantify the extent of between-study variations (that is, heterogeneity), we calculated the inconsistency index I2, which describes the variation of effect estimates that is attributable to the heterogeneity across the studies.19 Summary DORs were estimated by random (DerSimonian-Laird) or fixed (Mantel-Haenszel) effect models, depending on whether I2 was greater or lower than 50%.20 We 6

ACCEPTED MANUSCRIPT prespecified several additional analyses to examine the potential effects of various methodological

T

quality factors, adjust for covariates, and assess the robustness of our results. We defined a priori the

RI P

following potential relevant covariates: biomarker, type of underlying liver disease of the study

SC

population, and the type of infectious outcome. We constructed funnel plots and tested for publication bias by inspecting the symmetry of funnel plots, and assessed the potential for publication bias by

MA NU

performing Begg and Egger tests.21 We used Galbraith plots to identify potential outlier studies. Statistical analyses were conducted using STATA 11.0 (Stata Corp, College Station, TX, USA). All

ED

statistical tests were two-sided, and statistical significance was defined as a P-value less than 0.05.

PT

Results

CE

In total, 230 studies (excluding duplicates) were identified using the search strategy outlined earlier

AC

(Figure 1). After the first round screening of tile and abstracts, 216 nonrelevant studies, case reports, and reviews were excluded. 14 potential relevant studies were retrieved for full text evaluation, of which 6 further studies were excluded for varying reasons at this stage, leaving 8 that met the inclusion criteria. An update search identified additional 2 studies. Finally, 10 studies were included for analysis.

These studies included 1144 episodes of suspected infection, with 435 (38.0%) confirmed infection episodes. Table 1 summarizes the characteristics of the included studies and patients. The 10 included studies consisted of 9 studied PCT tests, 7 studied CRP tests, and 2 studied IL-6 tests. All studies studied 7

ACCEPTED MANUSCRIPT on cirrhotic patients. Three of the 9 studies used a case-control design, and the remaining used a cohort

T

design. Seven studies reported data using the standard PCT cutoff value (0.5 ng/mL). The number of

RI P

subjects in each study ranged from 20 to 368. The mean/median age was 54 to 63 years. Five studies

SC

used spontaneous bacterial peritonitis (SBP) as the primary endpoint, and the remaining used documented infection as the primary outcome. SBP was defined as an ascites sample with

MA NU

polymorphonuclear cell more than 250 cell/mm3 and compatible clinical symptoms. Systemic bacterial infection was defined either by positive blood culture results (microbiologically documented infection) with compatible clinical symptoms or by fulfilment of systemic inflammatory response syndrome

PT

ED

criteria with identified source of infection (clinically documented infection).

CE

We used the QUADAS tool for study quality assessment. Figure 2 provides an overall impression of the

AC

methodological quality of the studies. All blood draws were taken in close proximity to the confirmation of the diagnosis. All patients were verified by the same reference standard in all studies. None of the included studies explained patient withdrawals or reported uninterpretable results, and none of the studies reported whether physicians were blinded to the index test while verifying the outcomes by reference standards. Either procalcitonin or CRP was not incorporated as the diagnostic criteria for SBP or systemic bacterial infection and most of the studies required objective microbiological documentation as one of the reference criteria to make a final diagnosis of infection. Therefore, there is a low possibility of incorporation bias. (Table 1). 8

ACCEPTED MANUSCRIPT

T

Diagnostic accuracy indices

RI P

We constructed summary ROCs for both PCT and CRP (Figure 3). PCT had an area under the ROC of

SC

0.92 (95% Cl: 0.89-0.94). Pooled sensitivity and specificity estimates for PCT were 0.79 (95% Cl: 0.64-0.89) and 0.89 (95% CI: 0.82-0.94), respectively. The positive likelihood ratio (LR+, 7.38; 95% CI:

MA NU

4.70-11.58) of the PCT test was sufficiently high to qualify PCT as a rule-in test, and the low negative likelihood ratio (LR-, 0.23; 95% CI: 0.13-0.41) reduced the pretest probability to an acceptable level to safely justify the withholding of antibiotics, especially when patients were first screened by a laboratory

ED

test. CRP had an AUC of 0.87 (95% Cl: 0.84-0.90). The pooled sensitivity and specificity estimates of

PT

CRP were 0.77 (95% Cl: 0.69–0.84) and 0.85 (95% CI: 0.76-0.90), respectively. The positive likelihood

CE

ratio of CRP testing was 5.12 (95% CI: 3.08-8.54), and the negative likelihood ratio was 0.26 (95% CI:

AC

0.18-0.38). Overall, PCT had a slightly higher discriminative capability than CRP in differentiating infection from other causes of systemic inflammatory response syndrome among patients with liver cirrhosis. The diagnostic OR for CRP was 19.4(95% CI: 7.9-48.1), while the diagnostic OR for PCT was 24.0(15.5-37.5) (figure 4). We did not observe a substantial degree of heterogeneity for PCT (I2 = 38.0%; 95% CI: 0.0–71.5). We used Galbraith plots to explore sources of heterogeneity, but these plots did not show a potential source of heterogeneity in either the PCT or CRP meta-analysis. No significant evidence of potential publication bias was noted using the Egger’s test for funnel plot asymmetry (Table 2, Appendix Figure). 9

ACCEPTED MANUSCRIPT

T

Subgroup analysis

RI P

We performed a subgroup analysis on the studies reporting accuracy parameters for a standard PCT

SC

cutoff (0.5 ng/ml). The sensitivity and specificity remained nearly unchanged. Two major primary endpoints, bacterial infection and SBP, were used. PCT performed equally well in these two outcomes.

MA NU

The AUROC for bacterial infection was 0.91 (95% CI: 0.88-0.93), and for SBP was 0.92 (95% CI: 0.89-0.94).

ED

Discussion

PT

The results of this meta-analysis confirmed a comparable diagnostic accuracy for PCT testing in patients

CE

with liver cirrhosis than in patients with normal liver function. In our analysis, which included 1144

AC

patients with liver cirrhosis, we showed that PCT testing has an AUROC of 0.92 (95% CI: 0.89-0.94), a sensitivity of 0.79 (95% CI: 0.64-0.89), and a specificity of 0.89 (95% CI: 0.82-0.94) in diagnosing infection. These results are similar or superior to the results for normal adult populations (AUROC: 0.92, 95% CI: 0.89-0.94; sensitivity: 0.79, 95% CI: 0.64-0.89; specificity: 0.89, 95% CI: 0.82-0.94) reported in previous meta-analyses.22 The results are consistent when we restrict the analysis to patients with SBP, to patients with systemic infection. PCT also shows a strong performance on both positive and negative likelihood ratios, and is suitable as an important rule-in and rule-out ancillary diagnostic tool.

10

ACCEPTED MANUSCRIPT Unlike previous systemic reviews which universally found that PCT has a superior performance than

T

CRP in the diagnosis of systemic infection,22-25 we showed that the overall accuracy between PCT and

RI P

CRP in differentiating bacterial infections from other noninfectious causes of systemic inflammation in

SC

cirrhotic patients is nearly comparable. CRP is mainly synthesized by the liver in response to IL-6, while PCT is produced by nearly all tissue in response to TNF-alpha and other proinflammatory cytokines. It

MA NU

is believed that the production of CRP may be less suppressed in patients with impaired liver function during acute infection. However, this idea has been refuted by the observation that serum concentrations of CRP consistently elevate even in patients with Chid-Pigh C chronic liver failure regardless the

ED

presence of bacterial infection.26-29 Our results may provide evidence against the unselective use of PCT

PT

testing for identification of bacterial infection in patients with impaired function, given that the cost of a

AC

CE

PCT assay is double for a CRP assay.

In trying to make our findings more informative to clinicians, we calculated LRs, which can be used for the calculation of post-test probability.30 In the current pooled study population with a 38% prevalence (pretest probability) of bacterial infection, PCT has a LR+ of 7.38 and a LR- of 0.23 which can then translate into a positive post-test probability of 82% and a negative post-test probability of 12% (figure 5). Similarly, a CRP test with a LR+ of 5.12 and a LR- of 0.26 can derive a post-test probability of 76% and a negative post-test probability of 14% (figure 5). PCT has slightly better rule-in and rule-out value than CRP. However, considering that infection in patients with liver cirrhosis carries a high mortality 11

ACCEPTED MANUSCRIPT rate if left untreated, either PCT or CRP test is not recommended for use in isolation to exclude the

T

possibility of infection. Clinically, virtually all patients with liver cirrhosis who present signs of

RI P

infection will receive antibiotics. An algorithm that integrates the clinical information and PCT, which

SC

interprets gray-zone PCT results (for example, 0.25–0.5 ng/mL) in the context of clinical findings has shown to be able to safely shorten the duration of antibiotics in critically ill patients.31,32 It is therefore

MA NU

interested to know whether CRP can perform as well as PCT in guiding the use of antibiotics in cirrhotic patients with bacterial infection.

In addition to PCT and CRP, NGAL and bacterial DNA in peritoneal fluid have shown potential to

ED

contribute the diagnosis and management of SBP in patients with liver cirrhosis. Neutrophil

PT

gelatinase-associated lipocalin (NGAL) is a protein involved in iron metabolism and has been linked to

CE

inflammation. Early studies showed NGAL may have a high sensitivity and moderate specificity in

AC

differentiating bacterial from other cause of peritonitis. A small cross-sectional study showed a sensitivity of 0.96 and a specificity of 0.75 for the diagnosis of SBP. In contrast to NGAL, nucleic acid amplification tests showed a poor sensitivity but high specificity for diagnosing SBP. At this stage, nucleic acid amplification tests are better be used as a complementary test to detect SBP in patients with negative ascites culture results. With a vast array of potential diagnostic tools, clinicians should use these tests based on their characteristics. Our studies have both strengths and weaknesses. For the strengths, we used bivariate models rather than direct pooling of sensitivity and specificity of individual studies, which greatly reduced the influence of 12

ACCEPTED MANUSCRIPT threshold effects.18 Furthermore, by comparing the diagnostic accuracy between PCT and CRP, our

T

results provide the most compelling evidence to date that PCT and CRP performed equally well in

RI P

patients with liver cirrhosis. For the weaknesses, the validity of our meta-analysis depends on the quality

SC

of the selected studies. Three of the selected studies used the case-control design, which is prone to overestimate the performance of a test by excluding patients with undetermined diagnosis.33 Five studies

MA NU

used the old generation LUMItest with a reported detection limit of 0.5 ng/mL. Only two studies used the assays with higher precision with a reported detection limit up to 0.09 ng/mL.34 Values

Serum procalcitonin and C-reactive protein levels as markers of bacterial infection in patients with liver cirrhosis: a systematic review and meta-analysis.

The diagnostic value of procalcitonin (PCT) for patients with liver cirrhosis is unclear. We searched the PubMed, EMBASE, and Cochrane databases for s...
1MB Sizes 0 Downloads 3 Views