C-Reactive Protein or Procalcitonin Combined with Rhinorrhea for Discrimination of Viral from Bacterial Infection in Hospitalized Adults of Non-Intensive Care Medical with Lower Respiratory Tract Infection

Background: Whether procalcitonin (PCT) or C-reactive protein (CRP) combined with some clinical characteristics can better distinguish viral from bacterial infection is not clear. The aim was to assess the ability of PCT or CRP combined with clinical characteristics to distinguish between viral and bacterial infections in hospitalized non-intensive care unit (ICU) adults with lower respiratory tract infection (LRTI). Methods: This was a post-hoc analysis of a randomized clinical trial previously conducted among LRTI patients. The ability of PCT, CRP, and PCT or CRP combined with clinical characteristics to discriminate between viral and bacterial infection were estimated by portraying receiver operating characteristic (ROC) curves among patients with only vial or typical bacterial infection . Results: In total, 209 patients (virus 69%, bacteria 31%) were included in this study. When using CRP or PCT to discriminate between viral and bacterial LRTI, the optimal cut-off point were 22mg/L and 0.18ng/ml, respectively. When the optimal cut-off for CRP ( ≤ 22ml/L) or PCT ( ≤ 0.18ng/ml) combined with rhinorrhea was used to discriminate viral from bacterial LRTI, the AUCs were 0.81 (95% CI, 0.75–0.87) and 0.80 (95% CI, 0.74– 0.86), respectively. When CRP ≤ 22ml/L, PCT ≤ 0.18ng/ml and rhinorrhea were combined, the AUC was 0.86 (95% CI, 0.80–0.91), which was statistically signicant higher than that when CRP( ≤ 22mg/L) or PCT ( ≤ 0.18ng/mL) was combined with rhinorrhea (p=0.0107 and p=0.0205). Conclusions: Either CRP ≤ 22mg/L or PCT ≤ 0.18ng/mL combined with rhinorrhea could help distinguish viral from bacterial infection in hospitalized non-ICU adults with LRTI. When rhinorrhea was combined together, discrimination ability can be further improved.


Background
Lower respiratory tract infection (LRTI) is the most common infectious disease that may cause death, with about 3.0 million deaths worldwide in 2020. [1] Viral infection is one of the most important causes of LRTI.
Identifying the pathogens involved timely is essential for antibiotics treatment, as detection delay may potentially result in antimicrobial resistance. Antimicrobial resistance can cause corresponding nancial burden and environmental pollution, especially when antibiotics are inappropriately prescribed to patients with viral infection. [2] Although some novel molecular diagnostic or culture-independent assays offer enhanced opportunities to identify respiratory pathogens, researchers are still pursuing much simpler, faster, and cheaper ways to identify different pathogens. Serum markers such as C-reactive protein (CRP) and procalcitonin (PCT), which can help guide antibiotic use in LRTI patients, have been studied most often. [3] , [4] But whether PCT or CRP could distinguish viral or bacterial infection is a controversial issue. [5] , [6] , [7] , [8] Furthermore, most of the studies have been hampered by an incomplete etiologic approach, because only a limited number of infectious agents have been assessed or techniques with low sensitivity have been used. [9] , [10] Consequently, those studies have not reported reliable information on the use of biomarkers for differentiating bacterial from viral LRTI.
Though some overlaps exist in symptoms and clinical presentation between bacterial and viral infection, viral infection has its own characteristic, such as headache, generalized muscle pain and rhinorrhea. One recently study showed that combination of clinical symptoms and blood biomarkers can distinguish bacterial from viral community-acquired pneumonia in children. [11] However, few studies have been conducted among adult LRTI so far.
The aim of this study was to assess whether PCT or CRP combined with clinical characteristics could distinguish between viral and bacterial infections using comprehensive and sensitive methods of etiologic classi cation in hospitalized non-intensive care unit (ICU) adults with LRTI.

Methods
This was a post-hoc analysis of a randomized controlled trial (RCT) that had been published previously. [12] The RCT took place between October 2017 and July 2018 in the China-Japan Friendship Hospital (CJFH), Beijing, China (clinicaltrials.gov identi er: NCT03391076). The study was approved by the ethics committee of CJFH (2017-29). Written informed consent was obtained from each participant after meeting the inclusion criteria.

Study Population
The inclusion criteria of the RCT study were as follows: hospitalized patients aged ≥ 18 years who were preliminarily diagnosed as having radiographically con rmed community acquired pneumonia (CAP), acute exacerbation of chronic obstructive pulmonary disease (AECOPD), or acute exacerbation of bronchiectasis were recruited on the day of hospitalization. Patients were excluded if they were < 18 years old, pregnant, had hospital acquired pneumonia, or lung tuberculosis. We also excluded immunosuppressive patients. In addition, patients with any other condition that may have increased serum PCT levels were also excluded. For this post-hoc analysis, Patients who did at least one bacterial and one viral test could be recruited in this study.
patients without CRP or PCT testing results, or without bacterial or viral pathogens detected were further excluded.
PCT and CRP Measurement PCT or CRP concentrations were measured in the clinical laboratory of CJFH within 24 hours of admission. CRP was measured using the high sensitive-CRP Kit (i-Reader, China). The upper and lower detection limits were 200 mg/L and 1 mg/L, respectively. PCT was measured by use of the PCT Kit (i-Reader, China), with a detection limit of 0.01 ng/mL.

Pathogen Testing
Bacterial testing included quali ed sputum (de ned as squamous cells < 10 per low-power eld; polymorphonuclear leukocytes > 25 per low-power eld, or the ratio between the 2 < 1:2.5), lower respiratory tract samples and histological biopsy samples such as endotracheal aspiration, bronchoalveolar lavage uid, and

Statistical Analysis
Patients were divided into two groups according to pathogen detection results. Those with bacteria detected and negative mycobacterial/fungal tests results, regardless of viral or atypical bacteria results, were classi ed into bacteria group. The other patients only with viruses detected were classi ed into virus group.
Baseline characteristics were expressed as number (proportion) or median (interquartile range) respectively and compared by χ2 test or Mann-Whitney U test where appropriate. We then assessed the predictive performance of CRP, PCT and PCT combined with CRP for discriminating viral from bacterial infection by portraying receiver operating characteristic (ROC) curves, respectively. Optimal cut-points for CRP and PCT were de ned as the point on the ROC curve that has the maximum Youden index. Furthermore, according to the optimal cut-points, the performance of PCT ≤ 0.18 ng/L, CRP ≤ 22 mg/L, PCT ≤ 0.18 ng/L and CRP ≤ 22 mg/L combined with signi cant clinical features to discriminate viral and bacterial infection were evaluated. Sensitivity, speci city, positive predictive value (PPV), negative predictive value (NPV) and their 95% con dence intervals (CIs) were calculated. The areas under the curve (AUCs) and 95% CIs were estimated and compared to determine the different discriminations of models. A two-sided α less than 0.05 was considered statistically signi cant for all statistical tests. Statistical analyses were performed by the SAS software, version 9.4 (SAS Institute Inc.), unless otherwise indicated.

Results
Between Oct 16, 2017 and Jul 13, 2018, we recruited 800 patients in the previous RCT study. After excluding 129 patients without PCT or CRP, 39 patients with mycobacterial/fungal detected, and 423 patients with no pathogens detected, 209 patients were included in the current analysis. Of these patients, the viral group accounted for 69% and the bacteria group accounted for 31%.
Categorical variables were compared using χ 2 tests, and continuous variables were compared using Wilcoxon rank-sum test or Student's t-test.

Discussion
With etiologic detection approach covering relatively wide spectrum of pathogens in our study, we found that either CRP ≤ 22mg/L or PCT ≤ 0.18ng/mL combined with rhinorrhea could discriminate viral from bacterial infection in hospitalized non-ICU adults with LRTI, which has rarely been explored in adults. When CRP ≤ 22mg/L, PCT ≤ 0.18ng/mL and rhinorrhea were combined together, discrimination of viral from bacterial infection can be further improved.
For many years, physicians hoped to nd a marker that could help discriminating bacterial infection. CRP is an in ammatory marker, which was considered to be able to distinguish between viral and bacterial infections in 1990s. [13] , [14] But with the relative progress of detection technology, more studies thought CRP could not distinguish viral from bacterial infection. [11] , [15] , [16] Review these studies, most of them were conducted among paediatric patients, and the pathogen detection test have low sensitivity and covered limited pathogen. In this study of adults hospitalized with LRTI, we used RT-PCR and multiple nested PCR (FilmArray Respiratory Panel) testing, which were highly sensitive and accurate for the diagnosis of microbial etiology to detect viruses and atypical bacteria. Furthermore, the types of pathogens we detected were very comprehensive. Based on this, we suggested that our grouping was more accurate and the results were more credible than those of previous studies. We found the optimal CRP cut-off point was 22 mg/L, but which alone can not identify viral or bacteria infection in adult hospitalized LRTI patients.
PCT is a widely used and recognized biomarker of bacterial infection. Though PCT can guide antibiotic use in respiratory tract infections that had been widely adopted throughout the world, [17] some recently published studies found PCT could not distinguish viral and bacterial infections.
[8] , [18] Self`s study used sensitive and widely available etiological detection methods found no procalcitonin threshold perfectly discriminated between viral and bacterial pathogens.
[8] A meta-analysis found the sensitivity and speci city of PCT were 0.55 and 0.76 to distinguish viral from bacteria for CAP patients, which could be not reliable evidence either to mandate administration of antibiotics or to enable withholding such treatment in patients with CAP.
[18] Our result showed the optimal PCT cut-off point was 0.18 ng/mL and it may not be an ideal marker to distinguish viral or bacterial infections. And this viewpoint was consistent with Self`s study. Therefore, we thought using PCT alone to identify bacterial or viral infections and to guide the use of antibiotics should be cautious.
With increased interest in PCT research, many studies have shown that CRP is inferior to PCT in identifying bacterial or viral infections.
[6] , [19] , [16] In our study, we found that CRP is non-inferior to PCT in differentiating viral from bacterial infection in LRTI patients. Recently, one RCT found that CRP-guided prescribing of antibiotics for AECOPD resulted in a lower percentage of patients, with no evidence of harm. [4] Another study showed that the provision of PCT assay results in addition to usual care did not result in lower use of antibiotics than usual care among patients with suspected LRTI.
[20] Combined with our result, we need to further examine the importance of CRP in identifying viral infections and guiding antibiotic use for it is more available and cheaper than PCT.
The most important nding of this study was that CRP ≤ 22mg/L or PCT ≤ 0.18ng/mL combined with rhinorrhea could help to discriminate bacterial or viral infection, which was rstly reported among adults with LRTI in our consciousness. A study among children found that compared to CRP ≥ 72mg/L alone, CRP ≥ 72mg/L combined with symptoms (including rhinorrhea) could improve the speci city and PPV in discriminating bacterial from viral pneumonia. [11] Some reasons could explain that the CRP optimal cut-off point of our study is lower than Bhuiyan`s [11]. First, the types of patients and diseases were different between two studies. Second, the proportion of patients who received antibiotic therapy was high before hospitalization and onset of illness to admission was long (7days) in our study, which may in uence CRP value. [12] Though antiviral drugs and virus detection methods are limited clinically, clinicians should raise awareness if LRTI patient had low CRP or PCT combined with rhinorrhea and have more con dence in stopping or degradation of an antimicrobial drugs.
Our study has a number of limitations. Firstly, it is a reanalysis of a previous RCT, and not all enrolled patients received the FilmArray Respiratory Panel test. Secondly, quite a large proportion of patients who had no pathogen detected were excluded from current analysis although we did an etiology-based study. Thirdly, the study was conducted in general wards, without including patients from ICUs. Because of above limitations, extrapolation of our results should be carefully interpreted. We need further deep research to verify its accuracy in the future.   When using CRP to discriminate viral from bacteria LRTI, the area under the ROC curve was 0.77 (95% CI, 0.70-0.84), and the optimal CRP cut-off point was 22 mg/L (Figure 2A). Regarding PCT, the area under the ROC curve was 0.74 (95% CI, 0.66-0.82), and the optimal PCT cut-off point was 0.18ng/mL ( Figure 2B). When CRP (≤22mg/L) was combined with PCT(≤0.18ng/mL) to discriminate viral from bacteria LRTI, the area under the ROC curve was 0.77 (95% CI, 0.70-0.84) ( Figure 2C).