Procalcitonin as a Biomarker for Bacterial Infections in Patients With Liver Cirrhosis in the Emergency Department
Presented at the 13th International Conference on Emergency Medicine, Singapore, May 2010.
This study was funded by the Chang-Gung Memorial Hospital Research program (No. CMRP 370501).
Supervising Editor: Sandy Bogucki, MD, PhD.
Abstract
ACADEMIC EMERGENCY MEDICINE 2011; 18:121–126 © 2011 by the Society for Academic Emergency Medicine
Objectives: The objective was to determine the diagnostic accuracy of procalcitonin measurement for bacterial infections in patients with all causes of liver cirrhosis.
Methods: The authors conducted a cross-sectional study of 98 patients with cirrhosis treated in the emergency department (ED) of Chang-Gung Memorial Hospital, Taiwan. Serum procalcitonin levels and other clinical information were obtained concurrently. Patients were assigned to a sepsis or nonsepsis group after the medical records were reviewed by two emergency physicians blinded to the study. Receiver operating characteristic (ROC) curve analysis was conducted to determine the sensitivity, specificity, likelihood ratio, and suggested cutoff values. The diagnostic accuracy of the C-reactive protein (CRP) level was also determined for comparison.
Results: A total of 98 patients were enrolled for analysis in 1 year. Twenty-seven patients (27.6%) were assigned to the sepsis group. Eleven patients (11.2%) had positive blood cultures. The areas under the ROC curves for procalcitonin and CRP in predicting sepsis were 0.89 (95% confidence interval [CI] = 0.77 to 0.92) and 0.81 (95% CI = 0.72 to 0.89), respectively (p = 0.11). The cutoff that maximized Youden’s index was 0.49 ng/mL for procalcitonin and 24.7 mg/L for CRP. At these cutoffs, the sensitivity and specificity were 81.5 and 87.3% for procalcitonin and 80.0 and 80.3% for CRP. These results suggest that procalcitonin measurement shows at least an equivalent diagnostic accuracy to CRP measurement.
Conclusions: Procalcitonin provided satisfactory diagnostic accuracy in differentiating bacterial infections in patients with all causes of liver cirrhosis in the ED. A cutoff value of 0.5 ng/mL is suggested for clinical use.
Bacterial infections occur in 32% to 34% of hospitalized patients with cirrhosis and in approximately 45% of those admitted with gastrointestinal bleeding.1 Cirrhosis is an independent risk factor for sepsis and suggests a poor outcome in the setting of sepsis.2,3 Bacterial translocation and immune system dysfunction are two causative factors contributing to the pathogenesis of bacterial infections in cirrhotic patients.4
According to the definition of sepsis published by the International Sepsis Definitions Conference in 2001, the diagnostic criteria for sepsis include documented or suspected infection, general variables, inflammatory variables, hemodynamic variables, organ dysfunction variables, and tissue perfusion variables.5 Clinical variables, such as tachycardia, tachypnea, altered mental status, and edema are often mimicked by the common complications of liver cirrhosis, such as variceal bleeding, hepatic encephalopathy, and hypoalbuminemia. This makes the clinical diagnosis of sepsis in cirrhotic patients more difficult in the emergency department (ED). C-reactive protein (CRP) and procalcitonin are two plasma biomarkers included among the inflammatory variables in the diagnosis of sepsis. Assessment of CRP as an acute-phase reactant is widely performed in daily practice; however, there is limited evidence regarding procalcitonin assessment for cirrhotic patients in the ED.
Procalcitonin is a 116-amino-acid prohormone of calcitonin, which is normally synthesized in C cells of the thyroid gland. Procalcitonin was first described as a factor of interest in sepsis by Assicot and colleagues in 1993.6 The diagnostic value of procalcitonin measurement has been studied in various diseases and patient groups. Despite the presence of some controversy, an elevated serum procalcitonin level is believed to provide better diagnostic accuracy than an elevated CRP level in predicting bacterial infections.7 Plasma levels of CRP and procalcitonin are similar in critically ill patients with and without liver cirrhosis, despite concerns regarding the ability of the liver to produce these proteins.8 However, the diagnostic power of procalcitonin measurement compared to CRP measurement for sepsis in patients with cirrhosis has not been well established in the literature.
In a previous study, procalcitonin was compared to CRP, interleukin-6, and tumor necrosis factor-α in patients with decompensated liver cirrhosis.9 Procalcitonin had the best diagnostic value among these biomarkers. The patients in this study were all enrolled in the medical department. The serum concentration of procalcitonin might be affected by the various treatments provided before the blood sampling. The aim of the present study was to investigate the sensitivity and specificity of procalcitonin measured in the ED before various treatments occurred in patient with liver cirrhosis.
Methods
Study Design
This cross-sectional diagnostic study was conducted in the ED of Chang-Gung Memorial Hospital, Taiwan. The study was approved by the hospital’s institutional review board. The Standards for Reporting of Diagnostic Accuracy studies (STARD) criteria were used.10
Study Setting and Population
Patients admitted to the ED with previously documented liver cirrhosis were included for screening. Patients were considered to have liver cirrhosis by abdominal sonography, performed at Chang-Gung Memorial Hospital, regardless of Child-Pugh score. The result of the sonography must have been available before the ED visit. A total of 143 patients were screened between May 2008 and April 2009. Patients were excluded if they received any antibiotic treatment in the 7 days before admission to the ED, had documented hepatocellular carcinoma, and were receiving chemotherapy or declined to provide informed consent. Ninety-eight patients were enrolled in the final study.
Study Protocol
Patients were screened for infection after admission to the ED. After providing informed consent, patient age, sex, and vital signs were recorded. Blood samples were obtained for complete blood count, CRP assay, procalcitonin assay, and blood culture. Other blood tests or examinations were ordered according to each patient’s clinical condition. Diagnoses on discharge were recorded. The serum procalcitonin concentration was not available to the ED physicians providing clinical care until at least 24 hours after the specimen was obtained. All tests were performed by a quality-controlled central laboratory at Chang-Gung Memorial Hospital.
Determination of the Diagnosis
The patients were divided into a sepsis and a nonsepsis group after a medical chart review. The diagnoses on discharge and all laboratory results except procalcitonin were reviewed, and the diagnosis was made by the agreement of two ED physicians blinded to the procalcitonin results and to the study. If there was controversy between the two physicians, a third physician would make the determination. Positive blood culture was defined as any reported positive culture result regardless of the bacterial strain.
Data Analysis
Demographic data are shown as means and 95% confidence intervals (CIs). The F-test was used to evaluate the variance of means between two groups. Student’s t-test was used to compare the means of continuous variables between two groups. The Wilcoxon rank-sum test was used to compare procalcitonin and CRP between groups. Diagnostic accuracy, sensitivity, specificity, and likelihood ratio were determined by analysis of receiver operating characteristic (ROC) curves. The suggested cutoff value was determined by the Youden index. Sensitivity and specificity are expressed as percentages. All analyses were conducted by STATA 9 (StataCorp, College Station, TX) and a p-value of <0.05 was considered significant.
Results
Data from a total of 98 patients were included in the overall analysis. Demographic data are listed in Table 1. There were 27 patients in the sepsis group, and 71 in the nonsepsis group. Distributions of age and sex were similar between the two groups, as were body temperature and heart rate at presentation. The sepsis group had significantly higher respiratory rates and lower systolic blood pressures. White blood cell count, blood urea nitrogen, and international normalized ratio (INR) were also statistically different between the two groups. The percentages meeting the systemic inflammatory response syndrome (SIRS) criteria were 66.7% for the sepsis group and 22.5% for the nonsepsis group (p < 0.001). The mean procalcitonin value was 7.9 ng/mL in the sepsis group and 0.4 ng/mL in the nonsepsis group (p < 0.001). The mean CRP value was 65.5 mg/L in the sepsis group and 19.6 mg/L in the nonsepsis group (p < 0.001).
| Characteristic | Valid data | Sepsis group, n = 27 (95% CI) | Nonsepsis group, n = 71 (95% CI) | p-value |
|---|---|---|---|---|
| Age (yr) | 98 | 62 (57–67) | 60 (57–63) | NS |
| Male/female | 98 | 15/12 | 51/20 | NS |
| Bacteremia (%) | 98 | 41 | 0 | |
| Vital signs | ||||
| Body temperature (°C) | 98 | 37.1 (36.7–37.5) | 36.8 (36.6–37.0) | NS |
| Heart rate (beats/min) | 98 | 97.7 (91.7–103.7) | 94.6 (89.6–99.6) | NS |
| Respiratory rate (breaths/min) | 98 | 20.4 (19.3–21.5) | 18.4 (18.1–18.8) | 0.001 |
| Systolic blood pressure (mm Hg) | 98 | 118.5 (108.1–129.0) | 138.2 (131.0–145.4) | 0.003 |
| WBC/DC | ||||
| WBCs (cells/μL) | 98 | 11,248 (7,546–14,950) | 6897 (5603–8191) | 0.01 |
| Seg (%) | 98 | 77 (73.5–82) | 72 (69.2–74.8) | NS |
| Band (%) | 98 | 2.3 (0.6–4.0) | 0.2 (0.1–0.3) | 0.016 |
| SIRS criterion met (%) | 98 | 66.7 (48.9–84.4) | 22.5 (12.8–32.2) | <0.001 |
| Biochemistry | ||||
| BUN | 73 | 42.1 (23.7–60.5) | 21.7 (16.9–26.4) | 0.035 |
| Creatinine | 95 | 1.8 (1.1–2.6) | 1.5 (1.1–2.0) | NS |
| AST | 48 | 69 (46–92) | 78 (58–97) | NS |
| ALT | 88 | 90 (−1–181) | 42 (35–48) | NS |
| Total bilirubin | 67 | 5.1 (2.74–7.6) | 4.0 (2.6–5.4) | NS |
| INR | 56 | 1.9 (1.5–2.2) | 1.4 (1.3–1.5) | 0.005 |
| Inflammatory markers | ||||
| CRP (mg/L) | 98 | 65.5 (40.6–90.5) | 19.6 (12.8–26.2) | <0.001 |
| Procalcitonin (ng/mL) | 98 | 7.9 (0–15.9) | 0.4 (0.12–0.67) | <0.001 |
- ALT = alanine aminotransferase; AST = aspartate aminotransferase; CRP = C-reactive protein; DC = differential count; INR = international normalized ratio; NS = not significant; SIRS = systemic inflammatory response syndrome; WBC = white blood cell count.
The effectiveness of procalcitonin and CRP measurement in predicting sepsis and bacteremia in the ED setting was evaluated by the assessment of ROC curves. The area under the ROC curve, sensitivity, specificity, positive and negative likelihood ratios, and suggested cutoff values for each marker were calculated (Table 2). The areas under the ROC curve for procalcitonin and CRP in predicting sepsis were 0.89 and 0.81, respectively (p = 0.11, Figure 1A). The areas under the ROC curve for procalcitonin and CRP in predicting bacteremia were 0.86 and 0.72, respectively (p = 0.06, Figure 1B). For both sepsis and bacteremia, the suggested cutoff values for procalcitonin and CRP were 0.49 and 24.7 mg/L, respectively. Sensitivity and specificity ranged from 68% to 90.9%. The positive and negative likelihood ratios of each test to predict sepsis and bacteremia are reported in Table 2.
| AUC (95% CI) | Suggested cutoff | Sensitivity, % (95% CI) | Specificity, % (95% CI) | LR+ (95% CI) | LR− (95% CI) | |
|---|---|---|---|---|---|---|
| Procalcitonin vs. sepsis | 0.85 (0.77–0.92) | 0.49 ng/mL | 81.5 (0.62–0.93) | 87.3 (0.77–0.94) | 6.43 (3.97–10.39) | 0.21 (0.10–0.47) |
| CRP vs. sepsis | 0.81 (0.72–0.89) | 24.7 mg/L | 80.0 (0.59–0.93) | 80.3 (0.68–0.89) | 4.07 (2.55–6.49) | 0.25 (0.11–0.55) |
| Procalcitonin vs. bacteremia | 0.86 (0.77–0.92) | 0.49 ng/mL | 90.9 (0.59–0.99) | 75.9 (0.65–0.84) | 3.77 (0.12–2.16) | 0.12 (0.02–0.78) |
| CRP vs. bacteremia | 0.72 (0.61–0.81) | 24.7 mg/L | 72.7 (0.39–0.94) | 68.0 (0.56–0.78) | 2.27 (0.24–4.18) | 0.40 (0.15–1.07) |
- AUC = area under the curve; CRP = C-reactive protein; LR = likelihood ratio; ROC = receiver operating characteristic.
(A) Areas under the ROC curve for procalcitonin and CRP in predicting sepsis. (B) Areas under the ROC curve for procalcitonin and CRP in predicting bacteremia. CRP = C-reactive protein; ROC = receiver operating characteristic.
Discussion
Early diagnosis and timely treatment of bacterial infections and sepsis result in improved outcomes for patients. In the presence of septic shock, each hour delay in the administration of effective antibiotics is associated with a measurable increase in mortality.11 Early resuscitation of patients with severe sepsis or septic shock improves patient outcome.12 Clinical diagnosis of sepsis by SIRS criteria at the time of the visit is relatively easy for physicians; however, it is more difficult to identify or to exclude occult infections in the absence of SIRS criteria, especially in cirrhotic patients. Our results suggest that procalcitonin, as an emerging biomarker for bacterial infections, is able to differentiate bacterial infections in a satisfactory manner. Procalcitonin and CRP were the two biomarkers compared here; CRP was chosen as the best indicator for sepsis in cirrhotic patients among several acute-phase proteins.13 If clinical parameters and CRP are currently the best diagnostic tools available for the diagnosis of sepsis, procalcitonin could provide further value in the ED. Further validation is needed to confirm its utility.
When assessing vital signs, patients in the sepsis group had similar body temperatures and heart rates compared to the patients in the nonsepsis group. Despite the fact patients in the sepsis group had statistically higher respiratory rates and lower systolic blood pressures, neither parameter is strong enough to confirm or to exclude the diagnosis of bacterial infection. One-third of the patients in the sepsis group did not meet the criteria for SIRS, which suggests that it is difficult to make the diagnosis of bacterial infection with existing clinical parameters in cirrhotic patients. More laboratory tests or image studies might be necessary to confirm or exclude the diagnosis.
The suggested cutoff values for procalcitonin and CRP were derived by maximizing Youden’s index. The cutoff value of 0.49 ng/mL for procalcitonin was similar to the 0.5 ng/mL of the general population. The cutoff value of CRP was 24.7 mg/L in our study. This was similar to the previous report.14 These suggest that impaired liver function does not affect the production of procalcitonin from other tissues.
An ED physician is likely to prescribe antibiotics more aggressively to immunocompromised patients. There is no single diagnostic test that was able to exclude bacterial infections completely, especially in patients with liver cirrhosis. Beyond the diagnostic tools currently available in daily practice, procalcitonin provided relatively better diagnostic accuracy in predicting or excluding sepsis and bacteremia. The use of procalcitonin might contribute to improved patient safety and fewer unnecessary antibiotic prescriptions in the future.
Limitations
Patient outcomes were not analyzed in this study. An ideal biomarker for bacterial infections should provide information for the clinical decisions about the prescription or discontinuation of antibiotics. The suggested cutoff value derived from Youden’s index needs to be validated for safety and accuracy as well. Further study to determine if procalcitonin could be used safely as a guide for antibiotic treatment should be conducted in the future.
The grouping of the patients in this study could be biased because there was no standard blinding procedure for the physicians providing clinical management. Not all of the patients received thorough workups for bacterial infections. If the physician was aware of an elevated procalcitonin level clinically, more studies might be conducted to identify an infection focus. This bias was minimized because the procalcitonin result was not available the same day as ED admission. The clinical management was thus determined in advance of the procalcitonin result.
Conclusions
We found that the diagnostic accuracy of procalcitonin is at least as good as that of C-reactive protein measurement in differentiating bacterial infections in patients with all causes of liver cirrhosis. The suggested cutoff value is 0.49 ng/mL, which is close to the current reference value of 0.5 ng/mL. A cutoff value of 0.5 ng/mL is suggested for clinical use.
Acknowledgments
The authors appreciate the efforts in statistical analysis provided by Professor Chee-Jen Chang from the Clinical Informatics and Medical Statistics Research Center (CIMS), Chang-Gung University, Taoyuan, Taiwan.
