Diagnostic accuracy of C-reactive protein for bacterial infections in acute exacerbation of chronic obstructive pulmonary disease: A metaanalysis

Ze-Hui Lin, Yin-Ji Xu

1. Department of pulmonary and critical care medicine, The Second Clinical College of Guangzhou University of Chinese Medicine, Guangzhou 510006, China

2. The Second Affiliated Hospital of Guangzhou University of Chinese Medicine, Guangzhou 510000, China

Keywords:C-reactive protein Acute exacerbation of chronic obstructive pulmonary disease Bacterial infections

ABSTRACT Objective: The clinical diagnostic accuracy of C-reactive protein (CRP) for bacterial infections in acute exacerbation of chronic obstructive pulmonary disease (AECOPD) is still controversial. This study evaluated its accuracy through Meta-analysis. Methods: Studies on the diagnostic value of CRP for bacterial infections in AECOPD were searched form Web of Science, PubMed, Cochrane library, EMBASE, WANFANG DATA, CNKI and China Biology Medicine disc. The included studies were evaluated according to QUADAS-2 assessment tool. Stata 12.0 software was used for meta analyses to summarize the sensitivity and the specificity of the included studies, and the heterogeneity test was conducted. The symmetric receiver operating characteristic curves (SROC) was drawn and the area under the curve (AUC) was calculated, and the funnel plot was drawn to determine publication bias. Results: 664 studies were initially identified, 20 of which met the inclusion criteria (9 in English and 11 in Chinese). Meta-analysis results showed that the summary sensitivity and specificity of CRP in diagnosing bacterial infections in AECOPD were 0.84, 95%CI(0.77-0.90) and 0.76, 95%CI(0.67-0.82), respectively. The AUC of SROC was 0.87, 95%CI (0.83-0.89). Conclusion: CRP has high accuracy, sensitivity and specificity in distinguishing AECOPD bacterial infections from nonbacterial infection.

1. Introduction

Chronic obstructive pulmonary disease, a common preventable and treatable disease characterized by persistent respiratory symptoms and limited airflow, is the fourth leading cause of death in the world and is expected to become the third leading cause of death by 2020[1]. acute exacerbating event refers to a patient with COPD who has worsening symptoms and requires additional treatment. AECOPD is an important event in the management of COPD, and the primary cause of high hospitalization rate, high mortality rate and deterioration of health status in COPD patients. Among them, respiratory virus infection, bacterial infection and environmental pollution are important causes of AECOPD [2, 3]. Therefore, in the treatment of AECOPD, clinicians should firstly identify the AECOPD patients who with bacterial infection, and initiate empirical antibiotic treatment according to the condition. At present, the sputum bacterial culture is useful in identifying bacterial infectious AECOPD, however, sputum culture requrie a long time and good technical conditions, those disadvantages limit the this examination in patients with clinically guide AECOPD antibiotic treatment in the practical application value, especially for outpatient service of AECOPD patients, is not recommended to do sputum bacterial culture [4, 5]. C-reactive protein (CRP), as an acute protein reflecting the inflammatory state of the body, has been proved to be associated with the onset and course of acute infectious inflammatory diseases such as community-acquired pneumonia and AECOPD [6], C-reactive protein can be used as a clinical indicator for the identification of bacterial infection. However, the accuracy of CRP in the diagnosis of AECOPD bacterial infection remains controversial in clinical practice. Therefore, this study evaluated its accuracy through meta-analysis, providing a reference for clinicians to identify patients with acute bacterial exacerbation in the treatment of AECOPD

2.Methods

2.1 Retrieval of literature

The Web of Science, PubMed, Cochrane library, EMBASE, wanfang database, CNKI and Chinese biomedical literature database were searched from the database construction to October 29, 2019. English database search used the combination of subject words and free words, such as the search formula used in PubMed database is: ((((Bacterial infections) OR Bacterial infections[MeSH Terms])) AND ((CRP) OR C-reactive protein)) AND ((((Pulmonary Disease, Chronic Obstructive[MeSH Terms]) OR AECOPD) OR acute exacerbation of chronic obstructive pulmonary disease)). China clinical trial registry, US clinical trial registry and WHO international clinical trial registration platform were manually searched.

2.2 Selecting of the literature

The Endnote software was used to select the articles, download the full text for reviewing when necessary. The inclusion criteria were: ①The diagnostic method must include serum CRP; ②Gold standard is sputum bacteria culture; ③The study type was diagnostic test. The exclusion criteria were: ①Among repeatedly published literature, only the most detailed research report was included; ②Unable to extract effective data indicators of the literature.

2.3 Data extraction

The data was extracted and examined separately by the two authors. The data we extracted including: ①General information: author, year of publication, country and region, sample size, etc. ②Diagnostic parameters: the cut-off value of CRP in the diagnosis of bacterial infection and the parameters in the four-cell table of diagnostic test included true positive value (TP), false positive value (FP), true negative value (TN) and false negative value (TN).

2.4 Quality evaluation

The second edition of the Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2) was used in the quality evaluation of included studies, including case selection, diagnostic methods to be evaluated, diagnostic gold standard, test process and progress, a total of 4 quality domains, 10 questions, and 3 extrapolated applicability evaluation questions. Each entry is judged by "yes", "no" and "uncertain".

2.5 Statistical analysis

Meta analysis were performed by Stata 12.0 software, Meta regression analysis was used to investigate the sources of heterogeneity, the symmetric receiver operating characteristic curve and map of funnel publication bias were plotted and we calculated the area under the curve (AUC) and Q index. P value less than 0.05 was considered statistically significant.

3. Results

3.1 Characteristics

A total of 664 articles were obtained in the initial examination, and a total of 20 articles were included after selection, including 9 English articles and 11 Chinese articles. The minimum sample size was 30 and the maximum was 210. Four articles did not provide the diagnostic threshold of CRP, and a total of 1801 AECOPD patients were included. The process of literature retrieval and selection is shown in Figure 1, and the basic information of each included studies is shown in Table 1.

Table 1. The basic characteristics of 20 included studies

Figue1. Flowchart of literature search and study selection.

3.2 Meta analysis

Combining the sensitivity and specificity of each study, the symmetric sensitivity of CRP in diagnosing AECOPD bacterial infection were 0.84, 95%CI (0.77-0.90), and the symmetric specificity values were 0.76, 95%CI (0.77-0.82). The sensitivity and specificity reported by each included institute, as well as the symmetric sensitivity and specificity, are shown in Figure 2. As can be seen from the forest map shown in Figure 2, the Q test for sensitivity showed that the P value was less than 0.01, indicating that the heterogeneity test between the included studies is statistically significant, and the I2 statistic is 84.76%, indicating that the heterogeneity is relatively large. In terms of specificity, the Q test of the meta-analysis showed that the P value was less than 0.01, indicating that the heterogeneity among included studies was statistically significant, with the I2 statistic being 83.99%.

Figure 2. Forest map of the combined sensitivity and specificity of CRP in diagnosing AECOPD bacterial infection.

3.3 Heterogeneity analysis, symmetric receiver operating characteristic curve

Considering the different diagnostic thresholds of each study, there was a certain threshold effect, therefore, SROC curve was drawn to evaluate the diagnostic accuracy, as shown in Figure 3A. The AUC was 0.87, 95%CI(0.83-0.89), and Q index was 0.79. We perforemed the Meta regression analysis to invesitge the threshold effect, regression analysis enrolled the research type, sample size, area, whether the subjects enrolled continuously or randomly, whether or not to use the same gold standard, is the gold standard can accurately distinguish between bacterial infections and five variables, the Meta regression analysis of symmetric sensitivity showed that, whether or not to use the same gold standard is a statistically significant factor, other variables’P values were lager than 0.05, in Meta regression analysis of symmetric specificity, all P value of the variables were greater than 0.05, indicating that there were no statistical significance.

3.4 Publication bias

The Deek funnel plot was drawn, and the linear regression test indicated that P value was 0.533, indicating no obvious asymmetry in the funnel plot. Combined with the Deek funnel plot, the straight line in the figure was the regression line and the included Angle with the horizontal axis was close to 90 , indicating that the possibility of publication bias was small, as shown in Figure 3B.

3.5 Literature quality evaluation

The results of literature quality evaluation indicated that in the case selection, there were two articles with high bias, ten articles with moderate bias, and eight articles with low bias. In terms of evaluation and diagnosis methods, five articles were introduced with high bias, three with moderate bias, and twelve with low bias. In terms of gold standard, one was introduced with high bias, one was introduced with medium bias, and the rest all belonged to low bias. In terms of the test process and progress,two articles introduced high bias in the process of diagnosis, four articles introduced medium bias, and fourteen articles belonged to low bias. The applicability of extrapolation in all literatures was of low bias, and patients with AECOPD were included in case selection, so the diagnosis was clear. In terms of diagnostic test methods, serum CRP was used, and sputum bacteria were selected for the gold standard.

4. Discussion

AECOPD has a serious adverse impact on the quality of patients’life and economic burden of sociaty. The incidence and mortality of COPD in China are both high, which is related to the numerous inducing factors of AECOPD and the inadequate prevention and control. At the same time, it also indicates that there are many improvements space in the early identification and treatment of AECOPD. AECOPD is an important event in the clinical course of COPD, and also the main cause of the deterioration of health status and poor prognosis of COPD patients. Therefore, the standard treatment of AECOPD is a difficult and important task in clinic. Priority drugs for AECOPD treatment include bronchodilators, glucocorticoids, and antimicrobial agents. AECOPD upper respiratory tract infection is the most common pathogenic factors, the most common pathogenic factors for AECOPD are upper respiratory tract infection and tracheal - bronchial infection. Studies have reported that about 50% of AECOPD patients are associated with upper respiratory tract virus infection, which can trigger bacterial infection, and about 40-60% of AECOPD patients can isolate bacteria from sputum culture. Either bacterial infection triggered by upper respiratory virus infection or AECOPD directly induced by bacterial infection can lead to increased bacterial load in the airway of patients or the emergence of new infectious bacteria in the airway[27].

The specific immune response caused by bacterial infection and the inflammation mainly caused by neutrophils and CD8T lymphocytes led to the exacerbation of chronic respiratory tract inflammation in AECOPD. Bacterial, viral infection and air pollution can induce acute exacerbation. Infection and colonization of pulmonary viruses and bacteria are often accompanied by exacerbation of COPD airway inflammation. In clinical practice, when the patient’ history and clinical symptoms are integrated, serological laboratory examination is often required to help clinicians distinguish bacterial infectious AECOPD from non-bacterial infectious AECOPD.

CRP is a non-specific protein synthesized by the liver, various acute and chronic inflammatory factors, such as bacterial, viral, or fungal infections, stimulate macrophages and T cells to secrete interleukin-6 and other cytokines, Interleukin-6 stimulates liver epithelial cells to synthesize and release CRP, CRP binds phosphor choline on microorganisms, assists complement to bind foreign microorganisms, and enhances phagocytosis of macrophages, promotes phagocytosis of macrophages, and clears necrotic and apoptotic cells and bacteria [28, 29].

CRP is a commonly used clinical laboratory indicator to assess the severity of infection, which usually rises in 4-6 hours after inflammation occurs, peaks 1-2 days later, and gradually decreases with remission. In view of the biological role and characteristics of CRP level changes in inflammatory diseases, serum CRP levels are often used clinically as an aid in the identification of bacterial infectious AECOPD. Previous studies have reported that the accuracy of CRP in identifying AECOPD bacterial infection varies greatly, such as sensitivity ranging from 0.46 to 0.95, specificity ranging from 0.47 to 0.92. Most studies suggested that the specificity of CRP in distinguishing AECOPD bacterial infection was lower than the sensitivity, and its diagnostic threshold was difficult to be unified. Therefore, the clinical application of CRP in the identification of bacterial infection in patients with AECOPD is controversial, leading to many restrictions on the use of CRP in the treatment of bacterial infection with AECOPD.

Twenty diagnostic tests were included in this study. The results of meta-analysis indicated that serum CRP had a great accuracy in diagnosing bacterial infection in AECOPD. The area under the SROC reached 0.87, and the symmetric sensitivity and specificity were 0.84 and 0.76, respectively. The specificity of CRP in the diagnosis of AECOPD bacterial infections is lower than sensitivity, which is similar to the results of previous studies, we speculate that the reason mainly is that not only bacterial infection, chronic airway inflammation in AECOPD patients, other atypical pathogen infections, and viral infections are also the factors that lead to elevated serum CRP levels in AECOPD patients. The results of the quality evaluation indicated that, in terms of the four quality domains and three extrapolation applicability of the literature quality evaluation, most of included articles belong to low bias, but in terms of case inclusion, 9 of them were moderately biased, because they did not clarify whether the cases were included continuously or randomly.

We found that the included articles exists heterogeneity, Meta regression analysis suggests the heterogeneity did not come from non-threshold effect, since the diagnostic threshold of each study varied from 2.37 mg/L to 91.50 mg/L, indicating the presence of threshold effect, we believe that the threshold effect is the main reason for the heterogeneity of the included studies. This study was unable to determine the specific threshold of CRP in the differential diagnosis of bacterial infection AECOPD. Therefore, high-quality diagnostic tests are still needed in the future to determine the threshold of CRP in the diagnosis of AECOPD bacterial infection, so as to improve the application of CRP in the diagnosis and treatment of AECOPD.

In conclusion, in AECOPD, serum CRP level has a high accuracy in differentiating bacterial infection from non-bacterial infection, and the area under SROC reaches 0.87, with a high sensitivity and specificity. In the empirical antibiotic treatment of AECOPD, CRP may be a reliable laboratory indicator for distinguishing bacterial infectious AECOPD.