Our article has several take-home messages for the management of patients with COVID-19 pneumonia admitted to the ICU: First, approximately a quarter of patients with COVID-19 pneumonia admitted to the ICU have bacterial coinfection; second, a negative FA-PNEU result prevents the inappropriate empirical use of antibiotics in these patients as a stewardship strategy for COVID-19; and third, the overall concordance between FA-PNEU and culture was 90.1%, and it was between 92.7% and 100% when stratified by microorganisms.
Bacterial coinfection in critically ill COVID-19 patients occurred in 24.54% and 17.27% of patients tested by the FA-PNEU and conventional cultures, respectively. In the most recent meta-analysis  on the identification of bacterial coinfections by the FA-PNEU in ICU-hospitalized COVID-19 patients, four of the seven studies reported on the timing of specimen collection within the first 48 h of ICU admission. In total, 221 patients were included, and the pooled incidence of coinfections detected by the FA-PNEU was 33% (95% CI 0.25–0.41) and 18% by conventional cultures (95% CI 0.02–0.45) [7,8,9,10]; the incidence is higher than that reported in inpatient services, which ranges between 3.5 and 8% [4, 17]. The largest study by Kolenda et al. included 99 patients admitted to 3 ICUs in France, and the samples were taken in the absence of mechanical ventilation or within 48 h after mechanical ventilation was initiated; cultures identified 17 bacteria in 15 of 99 samples (15.1%).
The two most frequently detected bacteria were S. aureus (37.5%) and S. agalactiae (20%) by the FA-PNEU and S. aureus (34.78%) and K. pneumoniae (26.08%) by culture. Verroken et al.  reported the results of 32 respiratory samples from 41 COVID-19 patients in the ICU; the FA-PNEU identified 13/32 (40.6%) patients with a bacterial coinfection, where S. aureus (38.46–60% methicillin-sensitive), H. influenzae (23.07%), and Moraxella catarrhalis (15.38%) were the main pathogens identified. Kreitmann et al.  documented bacterial coinfection in 13 of 47 subjects (27.7%) from samples taken within 24 h of tracheal intubation, with three bacterial species representing ≥ 90% of those identified: S. aureus (69.2%, all methicillin-sensitive), H. influenzae (38.5%), and S. pneumoniae (23.1%). Kolenda et al. analyzed 99 patients with respiratory samples taken in the absence of mechanical ventilation or during their first 48 h; conventional cultures detected bacterial coinfection in 15%, with S. aureus (46.6%, all methicillin-sensitive), H. influenzae (26.66%), and S. pneumoniae (13.33%) being the most prevalent pathogens.
When comparing the aforementioned studies with our work, three aspects are worth highlighting: First, in all the studies, including ours, S. aureus was the most prevalent microorganism; second, in the present study, methicillin resistance was higher (40% of the FA-PNEU isolates had MecA/C/MREJ, while in the cultures, no methicillin resistance was found); and last, unlike other studies, K. pneumoniae was the second most prevalent microorganism in the current study by culture, with no ESBL or KPC resistance mechanisms.
Regarding the qualitative agreements between the FA-PNEU and conventional cultures, in our study, we found that the PPV was between 50 and 100% and lower for E. cloacae and S. aureus; the NPV was high (between 99.1% and 100%). Caméléna et al.  demonstrated that the results of the FA-PNEU are consistent (sensitivity 95%, specificity 99%, PPV 82%, and NPV 100%) with those of conventional culturing for bacterial pathogens of 96 samples from 43 intubated patients with suspected bacterial coinfection or superinfection; S. aureus, as opposed to our study, did have a good PPV (91%). Kolenda et al.  reported a FA-PNEU sensitivity of 100%, since all isolated bacteria in cultures were also detected using the FA-PNEU, with a specificity of 98.7%; the lowest specificity was for H. influenzae (< 88.4%), and the specificity for S. aureus was 93.5%.
In our study, tests for S. aureus had a sensitivity of 100%, a PPV of 53.3%, a specificity of 93.1%, and an NPV of 100%, since 6 of the 9 patients with FA-PNEU-positive and culture-negative microorganisms were positive for S. aureus. Moreover, in 6 FA-PNEU samples, the MecA/C/MREJ resistance mechanism was detected and not identified by conventional cultures. Fontana et al.  used the FA-PNEU to assess coinfection in 152 respiratory specimens from COVID-19 inpatients; 23 of them required assisted ventilation in the ICU. The most representative species was S. aureus in both BAL (21; 16 mecA positive) and sputum (27; 14 mecA positive), with the majority being mecA positive (30/44, 62%). Although most of the patients were not in the ICU, their results are consistent with our findings.
Concerning the quantitative agreement, in our study, microorganisms in cultures of ETA samples with > 105 CFU, 84.21% had a count of ≥ 105 copies/mL in FA-PNEU testing; of the culture-negative samples, 40.9% had microorganisms with a count < 105 copies/mL in FA-PNEU testing. In the study of Kolenda et al.  among 16 bacteria reported in cultures, 15 (93.8%) showed ≥ 106 copies/mL using the FA-PNEU; in contrast, among 26 bacteria detected using the FA-PNEU yet culture-negative, 20 (76.9%) had ≤ 105 copies/mL using the FA-PNEU. We can conclude that most positive samples in the FA-PNEU results, with negative cultures, have low DNA copies/mL. These findings raise the following questions: Is it possible that in these cases, it is not strictly a coinfection and rather a contamination by the endogenous flora? What is the clinical impact of this finding? While culturing remains the gold standard in the diagnosis of bacterial respiratory tract infections, it may be difficult to accurately recover all pathogens in clinical samples, as the organisms are in a complex matrix. In addition, culture results would be more affected by the host immune response and prior antibiotic usage. Antimicrobial therapy can impact bacterial growth, leading to negative cultures but to persistent positive FA-PNEU results, which is not able to distinguish dead from viable bacteria. For this reason, we excluded patients who, at the time of respiratory sample collection, had received any dose of empiric antimicrobial therapy. Distinguishing colonizing organisms from pathogens remains a challenge because levels of bacteria below the culture threshold can provide positive results in the FA-PNEU. However, previous studies have described that the bacterial burden could be overestimated by FilmArray compared to culturing [19, 20].
In future studies, we will try to give some answers by comparing the results of the cultures and FA-PNEU with samples from the lung microbiome through the extraction of DNA and RNA for sequencing of the same samples for metagenomics and metatranscriptomics, which will allow us to determine the functional profiles of the virulence and resistance genes of microorganisms and differentiate expressed human genes. These strategies that examine the host response provide opportunities to rethink what defines true pneumonia and lung coinfection, so some propose a reconceptualized view of pneumonia, in which the development of pneumonia is believed to result from disruption of the complex homeostasis of a microbial ecosystem interacting with multiple complex growth conditions .
We did not observe differences between inflammatory biomarker levels and pulmonary coinfection defined by the samples that were positive and negative for both techniques, the FA-PNEU and culturing. However, although the sample was not enough to make statistical comparisons, we found that in culture-negative patients with positive FA-PNEU results with a count of < 105 copies/mL, the inflammatory response was lower than that in patients with positive FA-PNEU results with a count of > 105 copies/mL, which could suggest that these patients have more colonization than infection. There are few studies evaluating whether acute phase reactants are useful as predictors of coinfection in patients with COVID-19. Bolker et al.  found that the risk factors for respiratory bacterial coinfection upon hospital admission were nursing home stay, severe COVID-19, and leukocytosis; the other inflammatory markers within 72 h of admission (procalcitonin, CRP, IL-6, and ferritin) were not predictors. Mason et al. , in a retrospective cohort study of patients with community-acquired pneumonia and patients with COVID-19, proposed that in COVID-19, the absence of both leukocytosis and an antibiotic-related decrease in C-reactive protein can exclude bacterial coinfection.
We did not find that coinfection by culture or by the FA-PNEU was associated with an increase in mortality. Another Latin American study conducted by Soto et al. evaluated ninety-three hospitalized patients with a diagnosis of COVID-19 who were analyzed with the FA-PNEU. Coinfection was evidenced in 38 (40.86%) cases, and no association with mortality was found (OR 1.63; 95% CI 0.45–5.82) .
Based on only a few studies without defined information on sampling strategies, a bacterial or fungal coinfection rate of 8% in COVID-19 patients was estimated, but 72% of all these reported COVID-19 patients received (empiric broad-spectrum) antibiotic therapy . In a study from Germany, of 135 analyzed cases, most patients received antimicrobial therapy within 24 h of admission (n = 109, 80.7%), and 46.0% of severely ill patients admitted to the ICU developed coinfections . In our data, of the 110 cases analyzed, 61 (55.45%) received at least one dose of empirical antibiotics after LRT sample collection, and when the treating physicians knew the result of the FA-PNEU, the therapy was modified in 58 (95.1%) of the 61 patients; furthermore, in 37 (97.4%) of the 38 cases in which the FA-PNEU result was negative, the antibiotics were suspended. The high NPV of the FA-PNEU allows us to quickly suspend or not start antibiotics as a strategy for improving antimicrobial stewardship in COVID-19.
A limitation of our study is that the samples were collected either by endotracheal aspirate or by mini-BAL and not through BAL. However, we did this for several reasons: First, the latest US guidelines recommend noninvasive sampling (endotracheal aspiration) with semiquantitative cultures to diagnose VAP and non-VAP; there is no evidence that invasive microbiological sampling with quantitative cultures improves clinical outcomes compared with noninvasive sampling with either quantitative or semiquantitative cultures. Second, noninvasive sampling can be performed more rapidly than invasive sampling, with fewer complications and resources; it is our usual practice due to the lack of availability of pulmonary professionals 24 h a day, and this applies even more during the COVID-19 pandemic. Third, the ETA samples were collected properly, and the respiratory samples were not processed when a quality level of 0 or 1 was detected by the Murray/Washington criteria, based on the number of squamous cells and neutrophils per field. Fourth, only results with growth ≥ 104 CFU/mL for a mini-BAL fluid sample or ≥ 105 CFU/mL for an ETA sample were considered; we did not take into account growth of < 105 for an ETA sample to avoid overestimating the prevalence of coinfection. Fifth, the BioFire® FilmArray® Pneumonia Panel technique has been validated for ETA samples 
There are several strengths in this study. First, the high NPV of the FA-PNEU was demonstrated; therefore, we can conclude that if we find a negative result, bacterial coinfection is practically excluded. Novy et al. proposed an algorithm for the rational use of the FA-PNEU in critically ill ventilated COVID-19 patients; this would allow 65.6% of antibiotic sparing for bacterial coinfection and better adequacy of empirical antibiotic therapy . Second, our investigation is one of the studies with the largest number of included patients with the aim of assessing the diagnostic concordance of the FA-PNEU with culturing in subjects with COVID-19 pneumonia admitted to ICUs. Third, this is the first Latin American study with this purpose; most of the studies have been carried out in Europe. Finally, we analyzed whether inflammatory response markers were associated with pulmonary coinfection and the changes in antibiotic management according to the results of the FA-PNEUM and cultures.