Accuracy of routine biomarkers and blood leucocytes count to assist diagnosis of COVID-19-associated pneumonia in adult patients visiting the emergency department

Background To investigate whether routine biomarkers and blood count could assist diagnosis of COVID-19-associated Methods This monocentre retrospective study enrolled 254 patients with nasopharyngeal RT-PCR for SARS-COV-2, routine biomarkers (D-dimers, brinogen, C-reactive protein, procalcitonin, NTpro-BNP, cTnT-hs) and blood cell counts. Sensitivity and specicity were evaluated. An adjudication committee classied diagnostic probability as certain, probable, unlikely, and excluded, based on all available data, then distributed in 2 categories: high (certain and probable) and low probability (unlikely and excluded). Results Between 25 th of February and 15 th of April, 2020, 254 of 388 patients could be analyzed. The adjudication committee classied 46 patients as denite, 18 as probable, 64 as unlikely, and 126 as excluded, corresponding to 64 high and 190 low probability. High and low probability patients differed for brinogen (P<0.0005) and white blood cell counts, notably leucocytes (P=0.0015), neutrophilic (P=0.0036), lymphocytes (P=0.0057), eosinophilic (P=0.027), and basophilic (P<0.001) counts. In a multivariate analysis, basophilic count < 25/µL (OR 3.048 [95%CI; 1.34-6.919]), neutrophilic count < 4000 /µL (OR 5.525 [95%CI; 2.20-13.855], and brinogen > 3g/L (OR 6355 [95%CI; 2.01-20.079] were independently associated with the diagnosis. Negative predictive values were 0.98 and 0.93 combining brinogen ( < 3g/L) and eosinophilic count ( < 80/µL), and brinogen and basophilic count ( < 25/µL), respectively. count, showed interesting performance for the diagnosis COVID-19 associated pneumonia. Combining brinogen with either eosinophilic or basophilic count was helpful to exclude the diagnosis.


Introduction
In December 2019, the onset of an outbreak related to SARS-CoV-2 occurred, an unknown coronavirus detected in January 2020 [1,2] and responsible for a disorder termed COVID-19. Since then, COVID-19 has spread worldwide and is responsible for an unprecedented pandemic with major threat on global health, and social and economic stability [3].
COVID-19 has a large spectrum of symptoms. Most patients seem to experience mild or moderate u-like disorder with cough, fever, and shortness of breath [4]. More severe presentations may occur and some patients develop an acute pneumonia that can lead to an acute respiratory distress syndrome (ARDS) [5,6]. Clinical signs and symptoms are poorly speci c, and an early diagnosis is crucial to isolate the patients and treat them in dedicated units. Positive diagnosis mostly relies on detection of the virus by RT-PCR in the upper respiratory tract [7,8] and suggestive images on chest CT-scan [9]. However, RT-PCR may be falsely negative in the early phase and CT abnormalities might be equivocal or lack speci city. Moreover RT-PCR results and CT-scan may be delayed and/or unavailable. As shortage of resources may occur during such a healthcare crisis [10], special attention should be paid to usual laboratory analysis in the management of patients suspected of COVID-19 infection during the phase of uncertain diagnosis. Several changes in biomarkers have been reported in COVID-19 infection: changes in leucocytes counts, notably lymphocytic and eosinophilic counts, elevation of D-dimers and cardiac biomarkers (since coronavirus may have cardiac and endothelial tropism and induce thrombosis and cardiovascular abnormalities). Professional and scienti c societies and healthcare organizations have endorsed the use of routine biomarkers and blood cell count to assist physicians at bedside [11][12][13][14]. Whereas numerous studies reported the prognostic value of such laboratory data, a limited number were dedicated to diagnosis. Therefore, while facing the outbreak, evaluating the role of routine biomarkers and leucocytes to guide diagnosis toward COVID-19 or alternative diagnoses is mandatory to better select patients since the Emergency Departments (ED) and x hospital organization [15].

Methods
We designed a single center, retrospective, observational study conducted from February 25 th . 2020 to April 15 th 2020 in the Emergency Department (ED) of Princess Grace Hospital, a general hospital in Monaco. The study was monitored by the research department of our Hospital. No funding was obtained for this study. The ethical board held for the study approved the protocol and waived the need of a written informed consent for inclusion. The protocol was registered in the clinicaltrial.gov website under the MONACOVID-Biomarkers acronym (NCT04401241). The National laws for Ethics did not require informed consent. The study complied with STARD recommendations [16].

Objectives
The primary objective was to assess sensitivity of routine biomarkers and blood cell count for diagnosis of COVID-19-associated pneumonia in low and high probability groups for COVID-19-associated pneumonia classi ed by an adjudication committee.
Secondary objectives were i) to compare values of routine biomarkers and blood cell count amongst the four different categories of level of certainty; ii) to assess performance of combined routine biomarkers and blood cell count for diagnosis of COVID-19associated pneumonia; iii) to assess whether usual biomarkers and blood cell count were associated with diagnosis of COVID-19associated pneumonia using sensitivity analyses in prede nite subgroups chosen a priori; 1) when comparing de nite versus excluded patients; 2) when comparing high vs. low probability patients, excluding bacterial infection (respiratory and extrarespiratory) in the low probability COVID-19-associated pneumonia group; 3) when comparing de nite versus excluded probability patients, excluding bacterial infections (respiratory and extra-respiratory) in excluded patients.

Adjudication committee
An adjudication committee consisting of two independent senior experts in infectious diseases and pneumology retrospectively assigned the probability of COVID-19-associated pneumonia diagnosis using the 4-level Likert scale, based on data collected from baseline on standardized case report forms, results of SARS-Cov-2 speci c RT-PCR, of low dose chest CT-Scan, and full access to all available data including patients' discharge summary. To note, the adjudication committee was blinded for classi cation by an independent senior radiologist of low dose chest CT-scan. A Likert scale allowed distribution of patients in four categories: 1) absence of COVID-19-associated pneumonia referred as excluded; 2) unlikely COVID-19-associated pneumonia; 3) probable COVID-19-associated pneumonia; and 4) de nite COVID-19-associated pneumonia. After adjudication committee classi cation, patients were distributed in two groups: low probability of COVID-19-associated pneumonia (excluded and unlikely) and high probability of COVID-19-associated pneumonia (probable and de nite).
For this study, the gold standard was the diagnosis assessed by the adjudication committee. Alternative diagnoses were established for low probability COVID-19-associated pneumonia and classi ed as i) non-COVID viral respiratory tract infection; ii) non-COVID bacterial respiratory tract infection; iii) respiratory tract disorder at the exclusion of infectious diseases; iv) extrapulmonary infectious diseases; and v) miscellaneous.

Study population
For the study purpose and to ensure quality of the nal adjudication committee diagnosis, we selected consecutive adults (18 years of age and above) visiting the COVID-19 dedicated ED who presented with clinically suspected COVID-19-associated pneumonia and had SARS-Cov-2 RT-PCR. Clinically suspected COVID-19-associated pneumonia was based on physician's judgment and ful lling the following criteria: new onset of symptoms concordant with viral infection (at least one of the following: sweat; chills; myalgia; temperature ≥38 °C or <36 °C or perception of fever; loss of smell and/or taste; diarrhea) and symptoms of an acute lower respiratory tract disorder (at least one of the following: cough; sputum production; respiratory rate ≥20 per minute; dyspnea; chest pain; altered breathing sounds at auscultation). We selected patients whose presentation required a blood sample for global and severity assessment.

Patient management and usual biomarkers and blood cell count
Patients' management in the COVID-19-dedicated ED was based on local protocoled practices based on a collegial multidisciplinary decision if they presented with suspected COVID-19-associated pneumonia. Recorded baseline data consisted of demographic data (age; gender), medical history of coexisting conditions; treatments; symptoms; clinical ndings and predetermined laboratory tests including: SARS-Cov-2 speci c RT-PCR obtained on nasopharyngeal swab; low-dose chest CT-scan; standard blood analysis (complete blood count; hemostasis; metabolic panel; creatinine; blood urea nitrogen; liver enzymes); D-dimers (Vidas. Biomérieux); Procalcitonin (PCT); C-reactive protein (CRP); high sensitive cardiac troponin T; NT-pro-brain natriuretic peptide (NT-proBNP, all on Cobas, Roche diagnostics).

Microbiological samples
Nasopharyngeal swabs were collected at admission and placed in a Middle Virocult MWE (Sigma®) transport medium. Samples were kept at room temperature and sent to the laboratory immediately after collection. For the presence of SARS-COV-2, swabs were sent to french referent centres for virological analysis (Nice, Marseille, Paris). Routine microbiological examinations were performed at the discretion of the emergency physicians and included: presence of in uenza A and B viruses and respiratory syncytial virus (RSV) A and B on nasopharyngeal swabs; blood culture; urine antigens for Streptococcus pneumoniae and Legionella pneumophila type I; serodiagnosis of Mycoplasma pneumoniae. These results were available to the adjudication committee.
Low dose chest CT-scan data and COVID-19-associated pneumonia diagnosis classi cation Multidetector low dose thoracic CT-scan was performed for each individual patient and interpretation was recorded using a standardized report form. The low dose chest CT-scans were independently reviewed by a senior radiologist, blinded from other data, and classi ed for probability of COVID-19-associated pneumonia as 1) excluded; 2) unlikely; 3) probable; and 4) de nite.

Statistical analysis
Baseline and follow-up characteristics were described by means and standard deviations (SD) or by median and interquartile range (IQR) for continuous variables with normal or with skewed distribution, respectively, and by percentages for categorical variables. Chi-square or Fisher exact tests were performed when appropriate for qualitative variables. The Student or Mann-Whitney tests were used to compare baseline characteristics and study outcomes between study groups for continuous variables with skewed distributions.
The distribution of values for usual biomarkers and blood cell count were determined in the different populations of patients using boxplots. The performances of current laboratory data in predicting de nite COVID-19-associated pneumonia were evaluated by sensitivity analysis (de nite vs excluded COVID-19-associated pneumonia). CRP was evaluated at several cut-off points of 20 mg/L, 50 mg/L, and 100 mg/L, values used in previous studies [17]. Several cut-off points for PCT were chosen at the level of 0.10 μg/L [17] and at the two levels for suspected bacterial infection 0.25 μg/L and 0.50 μg/L. Cut-offs for cTnT-hs were 14 ng/L and 50 ng/L, as stated by the manufacturer. Cut-offs for D-dimers test were 500 µg/ml and age-adjusted threshold [18]. Sensitivity, speci city, positive predictive values (PPVs), negative predictive values (NPVs), and likelihood ratio were calculated. Receiver operating characteristic (ROC) curves were drawn, area under the curve AUC was computed and optimal cut-off was identi ed by the maximization of the Youden's index, comparing laboratory results values in patients with excluded COVID-19-associated pneumonia and de nite COVID-19-associated pneumonia. From these optimal cut-offs for laboratory results, sensitivity analyses were performed combining cut-offs.
A multivariate logistic regression model was built to identify factors associated with having high probability of COVID-19associated pneumonia as compared to having an excluded COVID-19-associated pneumonia diagnosis. All variables with a p value of < 0.15 in the bivariate analysis were entered into a multivariate logistic regression with a backward stepwise approach; the discrimination was evaluated by the C-index and its 95 % con dence interval (95 % CI) and the calibration was evaluated by the Hosmer Lemeshow goodness-of-t test.
All tests were two-sided, and p-values below 0.05 were considered to denote statistical signi cance. All statistical analyses were performed using SAS 9.1. (SAS Institute, Cary, NC, USA).

Results
Characteristics of the population During the study period, 388 patients with suspected COVID-19-associated pneumonia and nasopharyngeal swab for SARS-Cov-2 RT-PCR visited the dedicated ED. One hundred and thirty four patients could not be included since routine biological results were missing for 128 and 6 were under 18 years of age ( Figure 1). Two hundred and fty four adults patients with available routine laboratory results were selected for analysis (Table 1) Performance of routine biomarkers and blood leucocytes count for diagnosis of COVID-19-associated pneumonia Among patients with high probability for COVID-19 associated pneumonia, PCT concentration was low, eosinophilic and lymphocytic counts were below normal range and levels of brinogen, D-dimers and CRP were increased. Lymphopenia (<1000 / µL) was observed in 26.0%, eosinopenia (<10 /µL) in 28.0%, high brinogen concentrations (>3.42 g/L) in 55.3%. When comparing high and low probability patients, a signi cant decrease was observed in nearly each leucocytes neutrophilic, lymphocytic, basophilic and eosinophilic counts; brinogen measurement was signi cantly elevated in high probability patients ( Table 2). Distribution of these 6 variables amongst the 4-level Likert scale categories are reported in Figure 2. Global performance of these parameters remained modest and none had an AUC above 0.80. Other parameters did not differ between the high and low probability groups.
Amongst all parameters, a drop in basophilic count showed valuable characteristics to ascertain diagnosis (Table 3). Indeed, positive likelihood ratios were 6.94 and 5.96 at a cut-off of 0 and 10 /µL, respectively. Performance for positive diagnosis was lower with other leucocytes counts (Table 4). To rule out the diagnosis, high neutrophilic and lymphocytes counts were of interest.
Negative likelihood ratio was 0.46 for both neutrophilic count above 7500 /µl and lymphocytes count above 2500 /µl. Results for other parameters are detailed in Supplementary data. When comparing patients with de nite or excluded COVID-19 associated pneumonia, similar results were observed (Table 3). A multivariate analysis showed interesting individual characteristics of basophilic counts, neutrophilic counts and brinogen measurement at the Youden's index value to help positive diagnosis of COVID-19 associated pneumonia ( Table 5).
Performance of routine biomarkers and blood leucocytes count for diagnosis of COVID-19-associated pneumonia, excluding bacterial infection To better assess sensitivity, we compared patients with high and low probability when bacterial infections were removed from the study population (Table 3) (Table 3).

Discussion
Here we report the accuracy of leucocytes counts and routine biomarkers to assist diagnosis of COVID-19 associated pneumonia in patients visiting the emergency department with pneumonia-like symptoms. The major strength of the present study is the control group to whom COVID-19 patients were compared. Our results highlighted the ability of basophilic count to exclude or con rm the diagnosis. We also observed that neutrophilic, lymphocytes and eosinophilic counts, and brinogen had interesting characteristics. Combining brinogen with either eosinophilic or basophilic count was helpful to exclude the diagnosis of COVID-19 associated pneumonia. Cardiac biomarkers and D-dimers test were of limited values at the emergency department.
Several scienti c societies and healthcare organisms have supported the use of common biological results to manage COVID-19 patients [8,[11][12][13][14]. The value of leucocytes counts to assist diagnosis of COVID-19 has been advocated since the beginning of the outbreak. Drop in lymphocytes and eosinophilic counts has been proposed to guide diagnosis process. Lymphopenia has been reported as the most frequent abnormality [19,20], observed in up to 82%. The decrease in eosinophilic count has also been underlined [21]. A study comparing patients with COVID-19 (n = 52) and LRTI (n = 53) reported lower eosinophils count in COVID-19 (0.02 /µL vs. 0.05 /µL, P value 0.004), and frequent eosinopenia (78.8% vs. 35.8%, P value < 0.001) [22]. In this study, leukocyte counts had no diagnostic value. Combining eosinopenia to clinical signs improved sensitivity and speci city (78.8% and 64.2%, respectively), as did lymphopenia (48.1% and 52.8%). While our results are in accordance with these results, we showed that leucocytes could also be of interest. We also reported that low basophilic count frequently occurred and had good diagnostic performance. This abnormality has not been reported so far. So far, we have been unable to provide speci c explanation. We believe that a drop in basophilic count should be investigated in other series of viral infection, including COVID-19. Combining eosinophilic count (≤ 80 /µl) and brinogen (≥ 3 g/L) could help excluding diagnosis of COVID-19 in patients with a u-like pneumonia. Indeed, we reported the good predictive values of brinogen measurements. Fibrinogen is associated with worse prognosis of COVID patients, as described for D-dimers test [23]. Activation of coagulation and major risk of venous thromboembolism are described in COVID patients, in relation with the so-called cytokines storm [24]. Therefore, increase in coagulation markers is foreseen. Interestingly acute brinous and organizing pneumonia (AFOP)-like pneumonia are described in SARS-Cov-2 infected patients [25]. Therefore it cannot be excluded that D-dimers increase could partly be related to the presence of brinogen in the lung compartment.
In COVID-19 patients, PCT is mostly below 0.25 µg/L [12,19]. In case of a superinfection, levels or PCT are usually measured above 0. 5 [27], we have been unable to detect any difference between groups for cardiac biomarkers [27].
We acknowledge that our results have limitations. For this study, we used retrospective data because of the urge of the sanitary situation and the relative resolution of the pandemic. This obviously leads to some bias. First, this method elicits missing data among laboratory measurements. Second, 128 adult patients analyzed for the presence of SARS-COV-2 using RT-PCR could not be included because no routine laboratory analysis was ordered. These patients signi cantly differ from the population analyzed in this study since none was admitted. Therefore we did not captured all the population with u-like syndrome but focused on patients with potential pneumonia. This explains that blood analysis was not ordered for some patients. Finally, we analyzed data available for attending physicians that obviously used them at bedside. Additionally, the adjudication committee may have utilized these laboratory results to classify patients for level of certainty. We previously reported such a methodology to classify community-acquired pneumonia [28]. Indeed, it was asked to adjudication committees to comment their decision for classi cation.
It appears that classi cation mostly relied on RT-PCR and CT-scan for positive diagnosis; the committees seldom quoted the use of routine biological parameters to assist their decision.

Conclusion
Guidelines have endorsed the use of routine laboratory analysis to help physicians making a diagnosis of COVID-associated pneumonia. Our results underscored the signi cance of blood cell counts analysis, the notable value of brinogen, and advocate for the use of combined eosinophilic or basophilic count and brinogen measurement to exclude diagnosis. COVID-19 may become endemic, senior UN health o cial Mike Ryan declared on Wednesday, 13 rd of May, 2020. This announcement strengthens the need for physicians to get reliable results for data available in daily practice, and underlines the scope of the present study.

Declarations Ethical Approval and Consent to participate
The ethical board held for the study approved the protocol and waived the need of a written informed consent for inclusion. The National laws for Ethics did not require informed consent.

Consent for publication
All authors have critically read and commented on draft versions of the report, and approved the nal version.

Availability of supporting data
Extensive access to data is available on request.

Competing interests
The authors declare no competing interests.

Funding
Not applicable.
Authors' contributions YEC and FB developed the study design. CFM was responsible for laboratory data collection. YEC and FB had full access to the data and take responsibility for the accuracy of the data analysis. FB and performed the statistical analysis. All authors assisted with data interpretation. YEC performed the literature search and wrote the rst draft of the paper. All authors have critically read and commented on draft versions of the report, and approved the nal version.  Table 2. Comparison of leucocytes counts, C-reactive protein, procalcitonin, brinogen, D-dimers, N-terminal pro-brain natriuretic peptide, and high sensitive cardiac troponin T in patients with high or low probability for COVID-19 associated pneumonia. Table 3. Comparison of leucocytes counts, C-reactive protein, procalcitonin, brinogen, D-dimers, N-terminal pro-brain natriuretic peptide, and high sensitive cardiac troponin T in patients with high or low probability for COVID-19 associated pneumonia and excluded bacterial infections, and in patients with de nite or excluded COVID-19 associated pneumonia, with or without bacterial infections. Table 4. Performance of brinogen, lymphocytes, neutrophil, eosinophil, and basophil polymorphonuclear leucocytes, according to different cut-offs in the patients with high or low probability for COVID-19 associated pneumonia. Youden corresponds to the optimized cut-off for sensitivity and specificity. Abbreviations. Se: sensitivity; Sp: specificity; PPV: positive predictive value; NPV: negative predictive value; LR+: positive likelihood ratio; LR-: negative likelihood ration; CRP: C-reactive protein; PCT: procalcitonin; age-adjusted: age-adjusted cut-off over 50 years (age x10); cTnT-hs: high sensitive cardiac troponin T. Figure 1 Chart ow.

Figure 2
Distribution (boxplot) of leucocytes count, neutrophil count, eosinophil count, basophil count and brinogen measurement each level of diagnosis certainty of COVID-19-associated pneumonia according to diagnosis certainty classi cation (adjudication committee).