Discussion

In this multicenter, retrospective cohort study, we described antimicrobial optimization based on negative results from BFP and respiratory culture. In our selected cohort of patients, we found that withholding or discontinuing antibiotics based on the negative results of both BFP and respiratory samples when clinically appropriate was not associated with increased in-hospital mortality, 30-day readmission due to PNA or recurrent PNA during index admission. Although the baseline characteristics were comparable between groups, there were a few clinical differences. WBC within 48 hours of antibiotic initiation was higher in the ATC group. Patients in the ATDW group were also more likely to have HAP, whereas patients in the ATC group were more likely to have CAP. Although the patients included in this study had similar baseline severity of illness evident from CCI, rates of ICU admission, and the rates of intubation at the time of inclusion, the distribution of qPBS favored the ATDW group ICU admission and/or intubation were identified as the only independent predictor of our composite primary outcome in multivariate regression analysis.

BFP includes 18 different bacterial species, including 3 atypical bacteria, and 8 different viruses (9 for BFP plus), and 7 different antimicrobial resistance genes. ⁴ BFP demonstrated high specificity for both sputum and BAL samples (>91%), ^{Reference Murphy, Fowler and Balada-Llasat7} with negative predictive value (NPV) ranging from 99.04–99.96% for common bacterial pathogens, when compared to the standard of care culturing method. ^{Reference Ginocchio, Garcia-Mondragon, Mauerhofer and Rindlisbacher19} Although cultures remain as the gold standard in the identification of bacterial respiratory tract pathogens, it may be difficult to accurately recover all pathogens in clinical samples due to multiple reasons—organisms being in a complex matrix, host immune response, and prior antibiotic usage. ^{Reference Murphy, Fowler and Balada-Llasat7} Furthermore, cultures are subject to laboratory interpretation by technicians examining cultures. ^{Reference Murphy, Fowler and Balada-Llasat7} On the other hand, the potential drawback of molecular methods is the detection of nonviable organisms and therefore positive results should be examined against clinical significance. ^{Reference Murphy, Fowler and Balada-Llasat7,Reference Ginocchio, Garcia-Mondragon, Mauerhofer and Rindlisbacher19} Given these characteristics, an absence of organism detection on both BFP and respiratory culture, can aid in antibiotic de-escalation and discontinuation when clinically appropriate.

Several studies have shown the potential for PCR-based methods to optimize antimicrobial therapy. In a randomized clinical trial of 208 patients, patients who were assigned to PCR-based molecular detection (Unyvero) had significantly shorter duration of inappropriate antibiotic therapy compared to patients assigned to traditional culture methods (47.1 h in PCR group vs 85.7 h in traditional culture method group). ^{Reference Darie, Khanna and Jahn11} Similarly, theoretical and simulation analyses have shown that PCR-based respiratory pathogen detection methods (BFP and Unyvero) have opportunities for de-escalation and potential antibiotic modifications, while reducing the duration of broad-spectrum antibiotic therapy. ^{Reference Monard, Pehlivan and Auger8–Reference Erich, Kilic and Palavecino10,Reference Guillotin, Poulain and Gaborit22} Two quasi-experimental studies showed feasibility of PCR-guided therapy implementation in the practical settings to reduce antimicrobial therapy duration in critically ill patients. ^{Reference Esplund, Taylor and Stone12,Reference Miller13} Negative BFP and respiratory culture results in our study led to antimicrobial de-escalation, discontinuation or not initiating antibiotics if clinically appropriate. Our study also adds additional finding of lower antibiotic-associated adverse events in ATDW group.

There is limited data evaluating the clinical outcomes of PCR-based antimicrobial therapies. In the previously mentioned randomized trial by Darie et al. ^{Reference Darie, Khanna and Jahn11} the authors did not find significant differences in clinical stability, length of stay, or mortality rates, although the duration of inappropriate antibiotics was significantly shorter in the PCR-based group. However, this study was limited to g organism infections, and the PCR samples were limited to BAL. Furthermore, resistance markers were also excluded in the PCR-based recommendations. Our study used a different PCR assay (BioFire Pneumonia) and included both sputum and BAL samples. Our results are consistent in that the reducing of antibiotic exposure was not associated with an increased signal for harm (length of stay, mortality, and readmission rates) in the targeted cohort of patients selected for our study who had no pathogens detected on both BFP and respiratory culture.

Our study has several limitations that need to be mentioned. Due to the nature of the study design, it is subjective to inherent retrospective bias. Patients in the ATC group potentially had higher severity of illness, which could have deterred clinicians from de-escalating or discontinuing antibiotics sooner. We tried to limit confounders by excluding patients who had other suspected infections (UTIs, SSTIs, etc.) and who had discordant results (MRSA PCR, atypical PCR, and blood cultures, etc.). All patients were required to have either sputum or BAL culture, with a suspicion of PNA on imaging. We conducted a regression analysis to identify predictors of our composite primary outcome. Nevertheless, the results of the study should be carefully interpreted due to the potential for residual confounding.

It is also possible that BFP would result in false-negative results, especially if the patient was receiving antibiotics for an extended period or if an organism is not part of the BFP panel. Non-detection or reduced sensitivity for Klebsiella aerogenes have previously been reported. ^{Reference Yoo, Huh and Shim20,Reference Nicolau-Guillaumet, Dortet, Jacquemin, Mourvillier, Muggeo and Guillard21} To minimize the rate of false negativity all patients who received more than 72 hours of antibiotics prior to BFP were excluded, and our results show that BFP samples were collected soon after development of PNA (median 1.2 d) and excluded patients who had positive cultures and PCRs. Although this approach strengthens the identification of true-negative BFP cases, it may also limit the ability to assess the real-time clinical utility of BFP’s rapid turnaround time. Specifically, by requiring culture confirmation, the benefit of early diagnostic information was not fully reflected in our outcome analysis. Lastly, we did not evaluate the clinical outcomes of patients who had antibiotics de-escalated to a narrower spectrum in the absence of resistant organisms, rather than discontinuing antibiotics. We saw that in the ATC group, 67.5% of anti-MRSA therapies, 25% of antipseudomonal therapy, and 50% of anti-atypical therapies were discontinued within 48 hours of PCR results. Further studies should evaluate the clinical impact of PCR-based de-escalation therapy rather than complete discontinuation of antimicrobials in patients with suspected PNA.