Abstract
Background/Aim: To assess the diagnostic performance of reverse transcriptase polymerase chain reaction (RT-PCR), low-dose chest computed tomography (CT), and serological testing, alone and in combinations, as well as routine inflammatory markers in patients evaluated for COVID-19 during the first wave in early 2020. Patients and Methods: We retrospectively analyzed data of all patients who were admitted to the emergency department due to fever and/or respiratory symptoms. CT scans were rated using the COVID-19 Reporting and Data System (CO-RADS) suspicion score. True disease status (COVID-19 - positive vs. negative) was adjudicated by two independent clinicians. Receiver-operating characteristic curves and areas under the curves were calculated for inflammatory markers. Sensitivities and specificities were calculated for RT-PCR, CT, and serology alone, as well as the combinations of RT-PCR+CT, RT-PCR+serology, CT+serology, and all three modalities. Results: Of 221 patients with a median age of 72 years, 113 were classified as COVID-19 positive. Among 180 patients from which data on CT and RT-PCR were available, RT-PCR had the highest sensitivity to detect COVID-19 (0.87; 95%CI=0.78-0.93). Notably, the addition of CT in the analysis increased sensitivity to 0.89 (95%CI=0.8-0.94), but lowered specificity from 1 (95%CI=0.96-1) to 0.9 (95%CI=0.83-0.95). The combination of RT-PCR, CT and serology (n=60 patients with complete dataset) yielded a sensitivity of 0.83 (95%CI=0.61-0.94) and specificity of 0.86 (95%CI=0.72-0.93). Conclusion: RT-PCR was the best single test in patients evaluated for COVID-19. Conversely, the routine performance of chest CT adds little sensitivity and decreases specificity.
In December 2019, the outbreak of a novel lung disease - COVID-19, caused by the newly discovered virus SARS-CoV-2, was observed in Wuhan, China (1). The first case in Switzerland was confirmed on February 25, 2020 in the Italian-border canton Ticino (2). Within few months, the caseload rose to over 8,000 confirmed cases and 66 deaths. The incidence of COVID-19 in Switzerland was comparable to most Western-European countries, with 496.5 confirmed cases per 100,000 habitants (Germany 264.7, France 449.9, Italy 448.4), but a relatively low death-toll was seen in our country, with 200.9 deaths per 1,000,000 inhabitants (Germany 111.6, France 471.2, Italy 586.2) (3, 4).
Due to its proximity to the densely populated greater Zürich area, the canton Aargau and especially the Cantonal Hospital of Baden were hit relatively hard, bearing a canton total of over 1,100 cases and nearly 400 patients at our hospital in the first three months of the pandemic. The further course of the pandemic with the ensuing waves is well known and has been described elsewhere (5, 6).
The clinical facets of COVID-19 have led to diagnostic challenges since the beginning of the pandemic. While reverse transcriptase polymerase chain reaction (RT-PCR) remains the mainstay in the diagnosis, false-negative test results may be caused by a decrease in the viral load in the upper respiratory tract, as the virus descends to the lower airways during the course of the disease. This reduces the sensitivity of the nasopharyngeal swabs obtained in routine clinical settings to an average of only 65-75% (7-9), depending on the relation to the start of symptoms (10, 11).
Characteristic patterns on computed tomography (CT) scans of the lungs can be detected in patients diagnosed with COVID-19 with high sensitivity, but low specificity (12-14). Since April 2020, serological tests detecting antibodies against SARS-CoV-2 have complemented the clinical, radiological and molecular biological diagnostic tools. Serum antibodies are first detectable approximately 7-14 days after the initial disease manifestation, limiting the clinical relevance of serological testing during the early disease state (15). Nevertheless, CT scan and serology have been propagated as important diagnostic tools since early in the pandemic (16).
The clinical course of COVID-19 is often mild, but vulnerable patients may need hospitalization due to rapidly progressing disease (17). These patients often show clinical symptoms and distinct radiological changes, despite initially negative RT-PCR tests from nasopharyngeal samples (7, 18).
Therefore, adjunct diagnostic modalities are used in emergency departments (EDs) to facilitate rapid and accurate decisions on therapeutic and hygiene measures, potentially leading to extensive and superfluous examinations.
This study aimed to evaluate the diagnostic performance of SARS-CoV-2 RT-PCR, low-dose chest CT and serology in patients admitted with respiratory symptoms, including one patient in whom COVID-19 was suspected upon routine testing prior to an operation. Additionally, we analyzed the performance of inflammatory markers in the diagnosis of COVID-19.
Patients and Methods
We analyzed data obtained during the first COVID-19 wave in Switzerland, i.e., between February 28, 2020, when the first patient was admitted to our hospital, and May 28, 2020, when the Swiss federal authorities lifted all decrees pertaining to the SARS-CoV-2 pandemic. We included all patients ≥18 years of age admitted to Kantonsspital Baden (KSB) who qualified for COVID-19 diagnostic testing, either due to respiratory symptoms and/or fever of unclear origin (admission to emergency department; n=220); or upon routine evaluation prior to an operative procedure (n=1). All patient-related data were extracted from the electronic medical record system used at KSB (KISIM Version 5.1.0.3) and entered into the REDCap data capture platform (19, 20).
Diagnostic tests included low-dose CT with or without angiography of the chest, measurement of serum SARS-CoV-2 antibodies, and SARS-CoV-2 RT-PCR in swabs from the upper respiratory tract. RT-PCR in serum samples was performed using kits manufactured by Qiagen, Ann Arbor, MI, USA (Qiastat©; Neumodx©), Cepheid, Sunnyvale, CA, USA (Genexpert©), DiaSorin, Saluggia, Italy (Liaison MDX©), Euromed Swiss, Frauenfeld, Switzerland (Elitech Elite Ingenius©), Roche Diagnostics, Rotkreuz, Switzerland and TIB/MOLBIOL/Roche z 480© assay, performed by the Institute of Virology, University of Zürich on Cobas 6800© or Cobas 8800©). Cutoff values (cycle threshold) were used according to specifications of the manufacturer. Serum antibody testing was performed on Cobas e411 using the anti-SARS-CoV-2 Elecsys assay (Roche Diagnostics). This assay uses a recombinant antigen representing the nucleocapsid protein. Complete blood count was analyzed on Hematology Analyzers XN-1500 or XN-3000 (Sysmex, Horgen, Switzerland). Plasma C-reactive protein (CRP), lactic dehydrogenase (LDH), aspartate aminotransferase (ASAT), alanine aminotransferase (ALAT) and D-Dimer, quantified as fibrinogen-equivalent units (FEU), were measured on Cobas 6000 c501 (Roche Diagnostics). All assays were performed according to the instructions of the manufacturer. Low-dose CT or CT-angiography scans were assessed using CO-RADS, a published and validated score for suspicion of pulmonary involvement, with a score of 1, 2 or 3 (very low, low and equivocal suspicion, respectively) considered negative, and a score of 4 or 5 (high and very high suspicion, respectively) rated positive (21-25).
The true disease status was adjudicated to all patients by two independent experts, board-certified in pulmonology and infectious diseases, respectively, on the basis of complete patient records including the electronic hospital charts, all laboratory results, CT scans and radiological reports, discharge reports, and the patient’s epidemiological context. In case of disagreement, a third expert board-certified in infectious diseases was involved and the final diagnosis established.
The study was approved by the competent ethics committee (Ethikkommision Nordwest- und Zentralschweiz, EKNZ; project-ID 2020-01159) without written informed consent by the patients, under article 34 of the Swiss Law on Human Research (Humanforschungsgesetz HFG).
Statistical analyses. For descriptive analyses, all patients of the cohort were included and missing values were ignored. Only diagnostic modalities performed within 10 days before and 2 days after hospital admission were considered. Continuous variables were summarized as mean±standard deviation (SD) or median with interquartile range (IQR); categorical variables were summarized as frequency (%). Normal variables were compared using one-way ANOVA; continuous non-normal variables were compared using Kruskal-Wallis Rank Sum test. Categorical variables were compared using Chi-Squared test.
For the evaluation of inflammatory markers, only patients with values available for all markers of interest were analyzed. Empirical receiver-operating characteristic (ROC) curves and areas under the curves (AUCs) with 95% confidence intervals (CI)s were computed for all inflammatory markers to estimate their ability to discriminate between COVID-19 diseased and non-diseased patients.
For the comparative analysis of diagnostic tests, patients with all three test modalities performed in the period specified above were analyzed. Sensitivity and specificity with 95% CIs were calculated for RT-PCR, CT, and serology alone, as well as the combinations of RT-PCR+CT, RT-PCR+serology, CT+serology, and all three modalities (considered positive if at least one of the respective tests gave a positive result). Since serology was available for a small portion of the cohort, a sensitivity analysis was performed including all patients with RT-PCR and CT in order to compare the two modalities on a larger number of patients.
Statistical analyses were performed using R, an open-source free software available under the GNU General Public License, version 4.1.0.
Results
Figure 1 shows an overview of the study. Patient characteristics are shown in Table I. Of 221 patients, 113 were classified as COVID-19-positive and 108 as Covid-19 negative by the independent panel. Sixty-seven of the patients in each group were males (62% in the non-COVID-19 and 59.3% in the COVID-19 group). Median (IQR) age was 73 (57.5, 82) years in the non-COVID and 69 (53, 81) years in the COVID-19 group (p=0.160).
The comorbidities were analyzed in both groups. There were no significant differences between the COVID-negative and positive groups for cardiovascular disease, prior surgery, chronic infection, chronic kidney disease, immunosuppression, diabetes, hypertension, and smoking.
However, there were significant differences between COVID-19-negative and positive patients in chronic airway disease (40% vs. 14.4%, respectively; p<0.001), malignancy (27.5% vs. 12.6%, respectively; p=0.016) and autoimmune diseases (15.7% vs. 2%, respectively; p=0.002).
The body mass index (BMI) was also significantly different between the groups. The COVID-positive patients had a higher BMI (median 27.14 kg/m2 vs. 24.69 kg/m2, respectively; p=0.001). Heart rate was slightly but significantly higher in the COVID-negative than in the COVID-positive group (mean 95.15 bpm vs. 89.68 bpm, respectively; p=0.045), but systolic and diastolic blood pressure (BP) upon admission showed no significant between-group difference.
Symptoms and signs at presentation. For our analysis we inquired symptoms and signs according to the questionnaire issued by the Swiss Federal Office of Public Health (FOPH). As shown in Table II, there were no significant differences between the COVID-19-negatives and COVID-19-positives in the following items: cough, sinusitis/rhinitis, myalgia, diarrhea, acute respiratory distress syndrome, and confusion. Asymptomatic presentations were also not significantly different.
The non-COVID-19 group significantly differed from the COVID-19 group in the following items: fever (44.4 vs. 69.7%, respectively; p<0.001), dyspnea (54.6 vs. 29.2%, respectively; p<0.001), chest pain (16.7 vs. 6.2%, respectively; p=0.026), pharyngeal pain (2.8 vs. 13.3%, respectively; p=0.009) and headache (3.7 vs. 12.5%, respectively; p=0.033).
Laboratory findings at presentation. Previous studies have shown that certain laboratory parameters may indicate COVID-19 (26). Results of CRP, D-dimer, neutrophile and lymphocyte count, LDH and liver transaminases were available for 104 of the 221 patients (47%). Table III shows the comparative analysis in the whole cohort. There were no significant differences between COVID-19-negatives and positives for CRP (median 52.3 mg/l vs. 58.30 mg/l, respectively; p=0.609) and D-dimer levels (median 811 vs. 505 μg/l, respectively; p=0.137). In contrast, and in line with previous descriptions, higher serum concentrations of LDH (median 294 vs. 237 U/l, p<0.001), ASAT (median 39 vs. 27.5 U/l, p<0.001), and ALAT (median 32.5 vs. 23 U/l, p<0.001) were present in COVID-positive than negative patients. Neutrophil (median 8.03 vs. 4.41×103/μl, p<0.001) and lymphocyte counts (median 1.1 vs. 0.82×103/μl, p=0.015) were significantly lower in patients with COVID-19. When only patients were considered in which all laboratory parameters were available, results were similar (Table IV). Figure 2 shows the receiver-operated characteristics of the laboratory parameters in this subgroup. The highest AUC was observed for the neutrophil count (0.79, 95%CI=0.70-0.88), suggesting acceptable discrimination ability. The AUCs (95%CI) of the other parameters were as follows: CRP, 0.54 (0.42-0.65); LDH, 0.67 (0.65-0.78); ASAT, 0.68 (0.57-0.78); ALAT, 0.69 (0.59-0.79); lymphocyte count, 0.62 (0.51-0.73); D-Dimers, 0.57 (0.45-0.68).
Comparative analysis of diagnostic modalities at presentation. Table V shows the results of RT-PCR, low-dose CT and serology for the whole cohort and according to true disease status.
Analysis in patients in which all three modalities were performed. Results from all three modalities (RT-PCR, serology and low-dose CT) were available for 60 of 221 patients (27%). Results from comparative analyses in this subgroup are shown in Table VI and Figure 3A. Eighteen of these 60 patients were in the COVID-19 group. Out of these, 7 had both positive PCR and CT but negative serology; in 6, all three test modalities were positive, and in 3 patients all test modalities were negative. Two patients had a negative PCR and a negative CT, but a positive serology.
As specified in Figure 3A, this results in a sensitivity of 0.72 (95%CI=0.49-0.88) for PCR and CT alone or in combination, a specificity of 1 (95%CI=0.92-1) for PCR, and a specificity of 0.88 (95%CI=0.75-0.95) for CT alone or in combination with PCR. The sensitivity of serology alone is 0.44 (95%CI=0.25-0.66) with a specificity of 0.98 (95%CI=0.88-1). The combinations of serology+PCR, serology+CT and serology+PCR+CT all had a sensitivity of 0.83 (95%CI=0.61-0.94). The specificity of the combined PCR and serology was 0.98 (95%CI=0.88-1), and that of serology combined with CT or CT+PCR was 0.86 (95%CI=0.72-0.93). Thus, performing a low-dose CT in patients with negative RT-PCR does not increase sensitivity for this subgroup. However, the sample size in this analysis was small.
Table VII summarizes the sensitivities, specificities as well as the positive and negative predictive values of the diagnostic modalities alone and in combination.
Sensitivity analysis in patients with RT-PCR and CT. Data on RT-PCR and CT, but not serology, were available for 180 patients (81%). Results from comparative analyses in this subgroup are shown in Table VIII and Figure 3B.
PCR alone had a sensitivity of 0.87 (95%CI=0.78-0.93) and a specificity of 1 (95%CI=0.96-1). CT alone had a sensitivity of 0.77 (95%CI=0.67-0.85) and a specificity of 0.9 (95%CI=0.83-0.95). Adding CT to PCR marginally increases the sensitivity to 0.89 (95%CI=0.8-0.94), while the specificity was reduced to 0.9 (95%CI=0.83-0.95).
Discussion
In this retrospective analysis, we studied a cohort of 221 patients evaluated for SARS-CoV-2 at the Kantonsspital Baden during the first COVID-19 wave from February 28 through May 28, 2020.
In our cohort, chronic respiratory disease was more common in COVID-negative than positive patients. This seemingly surprising finding can be explained by selection bias resulting from our hospital’s guideline for sign/symptom-triggered testing for COVID-19 at the emergency department, i.e., the presence of fever and/or respiratory symptoms. The same bias could also pertain to patients with preexisting cancer and autoimmune disease, but that these disorders were more common in the non-COVID group could also be a chance finding.
Since the turnaround time for the RT-PCR results was up to 24 h during the early phase of the Sars-CoV-2- pandemic and the sensitivity of RT-PCR in respiratory tract samples was known to be below 1, a low-dose CT of the chest was routinely performed in all symptomatic patients and prior to planned surgical procedures. This routine has continued at our institution until now. In about one third of the patients, a serological test was also performed. We evaluated whether such additional tools improved diagnostic accuracy and found that the benefit resulting from adding CT to RT-PCR was small, with only a minimal gain in sensitivity, accompanied with a slight loss of specificity.
The sensitivity of low-dose CT alone (72%) appears lower in our study than reported previously (13). However, in their metaanalysis, Xu et al. found considerable heterogeneity, with sensitivities outside Wuhan ranging between 61 and 98%. We also used a stringent cutoff value of ≥4 points for COVID positivity in the CO-RADS suspicion score.
Even though CT is readily available, which was a valuable asset especially early in the pandemic, it added little to confirming or rejecting the diagnosis of COVID-19 in our cohort. The value of CT lies mostly in establishing alternative diagnoses such as non-infectious diseases or pulmonary embolism. Combining clinical and laboratory data with CT metrics extracted by artificial intelligence might facilitate the diagnosis of COVID-19 in the future (27), but at present, the diagnostic accuracy of CT is limited.
Adding serology to either RT-PCR, CT or their combination yielded the highest sensitivity; in fact, 2 out of the 5 patients diagnosed with COVID-19 despite both negative RT-PCR and CT had a positive serology. Because our study uses data from the first wave, when dissemination was low (estimated 6% in May 2020 according to the FOPH) and no vaccine was yet available, the specificity of serology was very high. In the current situation, with a larger number of recovered and vaccinated individuals, the specificity would be much lower, making serological testing inappropriate for diagnosis of acute COVID-19. The development of antibody tests that distinguish between previous infection or vaccination and acute infection would be clinically valuable. The addition of routine inflammatory markers did not increase diagnostic accuracy, which could be expected, given their nonspecific nature.
Our study has several strengths and limitations. It is of relatively small size and only uses data collected in clinical routine in a single center setting, limiting generalizability of our findings. As the case definition for COVID-19 underwent frequent changes and amendments, some symptoms, such as anosmia, ageusia, abdominal discomfort or confusion, may be underrepresented. The data is inhomogeneous regarding serology results, as they were available in only 60 of the 221 patients. This potentially limits the generalizability of the findings of our analyses in the subgroup in which all three modalities were available. To optimize the reliability of the analyses, an independent panel of experienced clinicians adjudicated every patient in which the diagnosis was unclear from all available data.
As expected, RT-PCR proved to be the best single test to diagnose COVID-19 in patients admitted to the emergency department at our institution. Contrary to current practice, adding low-dose chest CT in this setting does not improve sensitivity enough to justify its routine use. We suggest that CT scans should be performed only in patients with negative RT-PCR but high clinical suspicion for COVID-19.
Acknowledgements
The Authors thank Janko Rakic, MD, Senior Resident in Pulmonary Medicine, and Michael Greiner, MD, Senior Resident in Infectious Diseases, both in the Department of Medicine, Kantonsspital, Baden, Switzerland, for reviewing the charts of all patients and deciding on their true disease status.
Footnotes
Authors’ Contributions
Benedict Gereke: Data capture and analysis; interpretation of data; first draft of manuscript; final version of manuscript. Andrée Friedl: Study design; interpretation of data; final version of manuscript. Tilo Niemann: Analysis of CT scans; interpretation of data; final version of manuscript. Romana Calligaris-Maibach: Laboratory analyses; final version of manuscript. Hans-Rudolf Schmid: Laboratory analyses; final version of manuscript. Chiara Vanetta: Study design; analysis plan; statistical analysis, interpretation of data; final version of manuscript. Jonas Rutishauser: Study design; data analysis; interpretation of data; final version of manuscript. Benedikt Wiggli: Study design; data analysis; interpretation of data; final version of manuscript.
Conflicts of Interest
The Authors declare no conflicts of interest regarding this study.
- Received January 12, 2022.
- Revision received February 16, 2022.
- Accepted February 21, 2022.
- Copyright © 2022, International Institute of Anticancer Research (Dr. George J. Delinasios), All rights reserved
This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY-NC-ND) 4.0 international license (https://creativecommons.org/licenses/by-nc-nd/4.0).