Agreement of respiratory viruses' detection between nasopharyngeal swab and bronchoalveolar lavage in adults admitted for pneumonia: a retrospective study

Objectives The COVID-19 pandemic has highlighted the high diagnostic accuracy of the nasopharyngeal swab (including in intensive care unit (ICU) patients). This study aimed to compare nasopharyngeal swab and bronchoalveolar lavage (BAL) results for non-SARS-CoV-2 viruses in patients with suspected pneumonia. Methods A retrospective analysis was performed in one French academic hospital on consecutive adults from 2012 to 2018 and tested nasopharyngeal swab and BAL within 24 hours by using multiplex PCR. The agreement in pathogen detection between nasopharyngeal swab and BAL was evaluated. Results Patients were primarily men (n = 178/276, 64.5%), with a median age of 60 years (IQR: 51–68 years). Of the 276 patients, 169 (61%) were admitted to the ICU for acute respiratory distress. We detected at least one respiratory virus in 34.4% of the nasopharyngeal swabs (n = 95/276) and 29.0% of BAL (n = 80/276). Two or more viruses were detected in 2.5% of the nasopharyngeal swabs (n = 7/276) and 2.2% of BAL (n = 6/276). Rhinovirus/enteroviruses were the most frequently detected viral group in 10.2% (n = 29/285) of the nasopharyngeal swabs and 9.5% (n = 27/285) of BAL, followed by influenza A, detected in 5.6% (n = 16/285) of the nasopharyngeal swabs and 4.9% (n = 14/285) of BAL. Overall agreement was 83.7% (n = 231/276 (95% CI [78.7%, 87.7%])) (i.e. same pathogen or pathogen combination was identified in the nasopharyngeal swab and BAL for 231 patients). Rhinovirus/enterovirus (n = 29/231) and respiratory syncytial virus (n = 13/231) had the lowest agreement of 62.1% (n = 18/29 (95% CI [42.4%–78.7%])) and 61.5% (n = 8/13 (95% CI [32.3%–84.9%])), respectively). Conclusions There was a good agreement between nasopharyngeal swabs and BAL in detecting respiratory viruses among adult patients with suspected pneumonia. However, these data still encourage BAL in the case of a negative nasopharyngeal swab.


Introduction
Pneumonia remains a leading infectious cause of hospitalization and death worldwide [1]. Pathogen detection is essential for diagnosis and to guide antimicrobial treatments to avoid antibiotic resistance. Syndromic testing has expanded our detection capabilities in the recent years, with broader pathogen detection and shorter turnaround times [2e5]. Multiplex PCR panels can be performed both on nasopharyngeal swabs and bronchoalveolar lavages (BAL). However, the efficiency-to-invasiveness ratio is difficult to assess. The question of testing BAL by mPCR for virus detection after having tested nasopharyngeal swabs in patients admitted to intensive care unit (ICU) for suspected pneumonia is still unclear. Few paediatric studies have shown poor agreement between nasopharyngeal swab and BAL for detecting respiratory viruses and variable diagnostic accuracy based on the viral species [6,7]. For example, nasopharyngeal swab had a good concordance with BAL for adenoviruses detection (kappa ¼ 0.561, 95% CI [0.321e0.801], p < 0.001), but a poor concordance for rhinovirus/ enterovirus detection (kappa ¼ 0.398, 95% CI [0.218e0.578], p < 0.001). Meanwhile, the COVID-19 pandemic has demonstrated that a nasopharyngeal swab might be adequate for an accurate diagnosis, even in ICU admitted patients, despite some differences between the two respiratory areas [8]. Thus, nasopharyngeal swab and BAL SARS-CoV-2 tests have reasonably high concordance in patients with respiratory failure when tested within 5 days since symptoms onset (88.6%) [8]. However, especially at the later disease stages, some discordances lead intensivists to perform a BAL in the case of a negative nasopharyngeal swab. Nevertheless, the viral detection agreement for non-SARS-CoV-2 respiratory viruses between nasopharyngeal swab and BAL is still poorly described in adult populations.
The objective of this study was to describe the agreement between the results of paired nasopharyngeal swab and BAL for non-SARS-CoV-2 viruses in adult patients with suspected pneumonia.

Study design and participants
We retrospectively reviewed the electronic medical records of consecutive adult patients (age 18 years old) evaluated at Bichat-Claude Bernard University Hospital in Paris between 2012 and 2018 for suspected pneumonia. Those patients had multiplex PCR testing on both nasopharyngeal swab and BAL samples within 24 hours.
Patient electronic medical records were reviewed, and the following information was obtained: age, sex, immunosuppression, chronic respiratory disease, chronic renal failure, diabetes, cirrhosis and chronic heart failure, ICU admission hospital length of stay, and death during hospitalisation.

Procedures
Nasopharyngeal swabs were collected with a flocked swab. Nasopharyngeal swab and BAL samples were placed in the Virocult MWE (Sigma, St Louis, MO, USA) transport medium and transported to the laboratory at room temperature within 24 hours. Different multiplex PCR panels were used during the study period to decrease the turnaround time. Because none of the tests differentiated between rhinovirus and enterovirus, these viruses were grouped as rhinoviruses for the purposes of this analysis. Included viruses in the analysis are influenza (A and B), Respiratory syncytial virus (RSV), metapneumoviruses, coronaviruses (HKU1, NL63, 229E, and OC43), parainfluenza (1 to 4), rhinoviruses, and adenoviruses. The various tests have been reported to have similar performances [2,9] as confirmed by our internal comparisons and method validations. Their reliability was also assessed throughout the study period by monthly internal controls and biannual QCMD controls (Quality Controls for Molecular Diagnostics, Glasgow, UK). Nasopharyngeal swab was routinely performed at the emergency department on the patients' arrival. By contrast, BAL was performed either by the ICU team for the patient admitted to ICU or the pneumology team when he or she was admitted to a medical ward.

Outcomes
The primary outcome was the percentage of agreement between BAL and nasopharyngeal swab; then, we conducted an evaluation in two ways. First, the binary (positive vs. negative) agreement between nasopharyngeal swab and BAL led to 276 paired binary results compared between nasopharyngeal swab and BAL. The double-positive results had to be consistent regarding identified pathogens.
Second, the evaluation took into consideration the identified pathogens. For this purpose, given that some samples had several identified pathogens, the results were dissociated. Each result corresponds to a unique pathogen, leading to 285 results from the 276 original samples.

Statistical analysis
The agreement of binary results was assessed by the Mc Nemar test and Cohen's kappa coefficient. The following grid was used to interpret Cohen's kappa coefficient: <0 ¼ Poor agreement, (0e2) ¼ Slight agreement, (0.21e0.40) ¼ Fair agreement, (0.41e0.60) ¼ Moderate agreement, (0.61e0.80) ¼ Substantial agreement, and (0.81e1.00) ¼ Almost perfect agreement. Because the BAL sampling should provide a more accurate etiological diagnosis of lower respiratory tract infection than the detection of nasopharyngeal portage, BAL was considered the reference standard for sensitivity and specificity evaluations.
The statistical analysis was performed with R version 3.6.1 (R Core Team (2020). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/).

Ethics
The study was approved by the CHU de Nîmes ethics committee IRB number: 22.03.10.

Patient characteristics
A total of 276 patients were included between 2012 and 2018. Nine patients had two pathogens detected, and therefore, we analysed 285 paired results.

Microbiological findings
Among the 80 patients with a positive result on nasopharyngeal swabs, 85.0% (n ¼ 68/80) also had a pathogen identified in the BAL sample, whereas among the 196 patients with a negative result on nasopharyngeal swabs, only 13.8% (n ¼ 27/196) had a positive result in the BAL. Among the 95 patients with a positive result on BAL, 71.6% (n ¼ 68/95) were also positive on nasopharyngeal swabs, whereas among the 181 patients with a negative result on BAL, 6.6% (n ¼ 12/181) were found positive on nasopharyngeal swabs (Fig. 1).

Agreement between nasopharyngeal swab and BAL
There was a significant difference between the two types of samples when using dissociated results (p ¼ 0.025) with a kappa of 0.67 ([0.58e0.77]). When considering the BAL sample as the reference standard for pneumonia diagnosis, the diagnostic performances of nasopharyngeal swabs were 71.6% 95% CI (61.4%e 80.4%) and 93.4% 95% CI (88.7%e96.5%) for sensitivity and  specificity, respectively (Table 2). Overall, 85.9% (n ¼ 237/276) of the results were consistent between nasopharyngeal swabs and BAL in positivity status and 83.7% (n ¼ 231/276) of the results were consistent between nasopharyngeal swabs and BAL in the exact same pathogen detected, including 169 patients negative in both tests (Fig. 1 Fig. 2).

Discussion
This study assessed the agreement of respiratory virus identification using multiplex PCR between paired nasopharyngeal swab and BAL, obtained within 24 hours, in adults admitted for suspected pneumonia. The results showed a good agreement between nasopharyngeal swabs and BAL, although 14.1% (n ¼ 39/276) of cases showed discrepancies.
In a similar study conducted by Azadeh et al. in immunocompromised patients, the agreement between nasopharyngeal swabs and BAL was slightly lower, at only 70%. The positivity rates of both nasopharyngeal swabs (21%) and BAL (24%) were also lower than those in our study at 29.0% (n ¼ 80/276) and 34.4% (n ¼ 95/276), respectively [10]. Another fascinating finding is the difference of agreement between BAL and nasopharyngeal swabs according to the pathogen. For instance, RSV had one of the poorer agreements. To our knowledge, no previous study had reported these results, and further investigations are required. The rest of our findings are in line with the previous studies, mostly nasopharyngeal swabs and BAL results [10e13].   The most clinically relevant message is that BAL rarely triggers additional clinical measures to a positive nasopharyngeal swab result, although it could add important information regarding bacterial or fungal infection and therefore have a direct effect on antimicrobial prescription, as shown in Gadsby et al. previous study where molecular testing allowed a de-escalation in the number and/ or the spectrum of antimicrobials in 77% of the patients [14]. Indeed, recently developed BAL-designed mPCR assays now include the detection of bacterial infections, providing a benefit over nasopharyngeal swab-designed mPCR, which is mainly limited to intracellular bacteria. When the nasopharyngeal swab is negative, BAL testing is useful to obtain a more precise diagnosis as, for example, 20.8% (n ¼ 5/24) of influenza and 38.5 % (n ¼ 5/13) of RSV detected in BAL were not detected in paired nasopharyngeal swabs.
Recent findings during the COVID-19 pandemic confirmed that BAL is essential when the nasopharyngeal swab is negative among patients with a suspicion of a severe SARS-CoV-2 infection [15]. A study on critically ill patients also showed a significant gradient between nasopharyngeal swabs and BAL viral loads with median Ct values of 29 for NP and 24 for BAL [16]. The gradient of viral loads between nasopharyngeal swabs and BAL has not been described in other respiratory viruses and warrants exploration in future studies. Indeed, it could explain some of the discrepancies observed between nasopharyngeal swabs and BAL and helping to understand the disease stage and whether viral replication occurs preferentially in either the upper or lower respiratory tract.
Our study had some limitations. It was a single centre study and, therefore, only represents the specificity of our population that includes almost 40% of immunocompromised patients. Furthermore, it is limited by its retrospective design that prevented us from performing an adjudication of final pneumonia diagnosis, assessing the pathogenicity of detected viruses and avoiding inclusion bias. Finally, the quality of nasopharyngeal swab is more subject to the operator variability than BAL and can generate a false negative nasopharyngeal swab [17,18]. Another limitation to our study is also the change of mPCR tests over time. All changes were made with the aim of shortening the time required to obtain the results (48 working hours for the Respifinder® test, 24 hours for AnyplexTM, and less than 2 hours for the FilmArray® test). Despite that the same test was used for both NP and BAL, these changes may have introduced some bias into viral distribution despite the good reproducibility of our results across time periods. The different mPCR tests used have been reported to have similar performances [2,9,19] as confirmed by our local assessment of comparative performances before implementation. Their reliability was also assessed throughout the study period using internal and QCMD controls (Glasgow, UK).

Conclusion
Nasopharyngeal swabs and BAL testing present a good agreement to detect respiratory viruses despite discrepancies in 14.1% of the studied population. Wide BAL testing for respiratory viruses detection, along with classic bacteriological and fungal testing, appears useful when the nasopharyngeal swab is negative to add important information on respiratory viruses not detected in the nasopharyngeal swab for the diagnosis, prognosis, and patient management. Further prospective studies, allowing systematic BAL testing and multicentric design, should be useful to confirm and complete our observations.