Performance and Impact on Antibiotic Prescriptions of a Multiplex PCR in a Real-Life Cohort of Critically Ill Patients with Suspected Ventilated Pneumonia: A Retrospective Monocentric Observational Study

Pulmonary multiplex polymerase chain reaction (m-PCR) allows rapid pathogen detection. We aimed to assess its impact on initial antibiotic prescriptions in ventilated patients with suspected pneumonia. Between November 2020 and March 2022,ventilated patients with suspected pneumonia hospitalized in our ICU who benefited from respiratory sampling simultaneously tested using conventional microbiological methods and m-PCR were included. The proportion of appropriate changes in the initial antibiotic therapy following m-PCR results was assessed. We analyzed 104 clinical samples. Of the 47 negative m-PCR results, 16 (34%) led to an appropriate antibiotic strategy: 8 cessationsand 8 lack of initiation. Of the 57 positive m-PCR results, 51 (89%) resulted in an appropriate antibiotic strategy: 33 initiations, 2 optimizations, and 9 de-escalations. In the multivariate analysis, a positive m-PCR was associated with an appropriate antibiotic change (OR: 96.60; IC95% [9.72; 960.20], p < 0.001). A higher SAPS II score was negatively associated with an appropriate antibiotic change (OR: 0.96; IC95% [0.931; 0.997], p = 0.034). In our cohort, a positive m-PCR allowed for early initiation or adjustment of antibiotic therapy in almost 90% of cases. A negative m-PCR spared antibiotic use in onethird of cases. The impact of m-PCR results was reduced in the most severe patients.


Introduction
Ventilator-associated pneumonia (VAP)is among the most frequent infections in intensive care units (ICUs) [1].They induce significant morbidity and mortality [2,3].The COVID-19 pandemic resulted in a high incidence of Acute Respiratory Distress Syndrome (ARDS), requiring ventilation support and leading to high mortality [4,5].Patients with SARS-CoV-2 infection complicated with ARDS are at high risk of bacterial pulmonary coinfection or superinfection [6][7][8].A prolonged length of sedation and mechanical ventilation, relative immunosuppression induced by a viral infection, and corticosteroid prescription are all factors favoring the occurrence of VAP [9][10][11].
In the case of VAP, early prescription of an appropriate antimicrobial therapy could reduce mortality [12].Rapid recognition of VAP is, therefore, essential to improving prognosis.Traditionally, diagnosis of VAP is based on clinical suspicion, new or progressive clinical signs such as fever or hypoxemia are not specific toVAP, and dis radiologic infiltrates is random.In patients with severe COVID-19, the dia is even more challenging since substantial overlap exists between the symptoms and signs of COVID-19 with secondary infections [14,15].This co to frequent initiation of empirical large-spectrum antimicrobial therapy wh microbiological results.A challenge in diagnosing pulmonary coinfection tion in critically ill patients with COVID-19 is reducing the time from sam ogen identification.Pulmonary multiplex polymerase chain reaction (m-P fective and rapid pathogen detections and could help clinicians choose ta crobial therapy.Its performance, as well as its use in nosocomial pulmonar the ICU, were evaluated in combination with expert opinion [16].This strated the theoretical potential of m-PCR in reducing unnecessary antibio comparing hypothetical m-PCR-driven antibiotic prescriptions by ex ture-driven antibiotic prescriptions by clinicians unaware of m-PCR resu data on m-PCR real-life use remains limited, especially in the context of pandemic.Its use was evaluated in COVID-19 patients suspected of commu pneumonia (CAP), hospital-acquired pneumonia (HAP), or VAP, showing portion of de-escalation compared with those observed in studies simulati of m-PCR [17].We conducted a retrospective monocentric study to assess initial antibiotic prescription of m-PCR in ventilated patients with suspecte

Demographic and Clinical Data
During the study period, 158 m-PCRs were performed (Figure 1).Am were excluded from the analysis.The main reason for exclusion was the simultaneous microbiological culture.Table 1 summarizes the patients' characteristics.They were mainly men (68%), with a mean age of 632 ± 11 years.Comorbidities were frequent, with 41% of the patients suffering from diabetes and 43% from arterial hypertension.Almost three-quarters of our patients were suffering from SARS-CoV-2 infection.Suspected diagnosis was a CAP in 15%, a HAP in 39%, and a VAP in 46% of cases.The death rate in the ICU was 54%.

Impact of m-PCR Results on Antibiotic Prescription and Additional Modifications after Culture Results
The impact of m-PCR results on antibiotic prescription is summarized in Table 4.In the case of a negative m-PCR, an appropriate antibiotic strategy was applied in 16 (34%) cases, with a lack of antibiotic initiation in 8 cases and an antibiotic interruption in 8 cases.Regarding positive m-PCRs, 51 (89%) led to an appropriate antibiotic prescription, including 33 appropriate initiations, 9 escalations, 7 de-escalations, and 2 optimizations.In 6 cases, antibiotherapy following positive m-PCR results was inappropriate.Three S. aureus pneumonia and one polymicrobial pneumonia (H.influenzae, S. aureus, M. catarrhalis) were treated with piperacillin-tazobactam without de-escalation.Amoxicillin-clavulanic acid was continued in one case despite the detection of gene mec A/C and in one episode of polymicrobial pneumonia despite the detection of P. aeruginosa. Figure 2 shows initial antibiotic prescriptions, changes following m-PCR results, and additional modifications after culture results.Four patients died between day 1 and day 2 after the onset of pneumonia, preventing antibiotic change depending on microbiological culture results.On day 2, in cases of a negative m-PCR, obtaining the results of culture and antibiogram never led to the cessation of antibiotics, despite a negative culture in 75.4% of cases.In cases of positive m-PCR, antibiotic therapy was modified after cultureresults in 17.5% of cases.In 4 cases, antibiotic change was motivated by resistance to initial probabilistic treatment.In 2 cases, antibiotic escalation was applied despite bacterial sensitivity to ongoing treatment.In 4 cases, negative standard culture motivated antibiotic de-escalation.
Antibiotics 2023, 12, x FOR PEER REVIEW 5 of 13 cases.Regarding positive m-PCRs, 51 (89%) led to an appropriate antibiotic prescription, including 33 appropriate initiations, 9 escalations, 7 de-escalations, and 2 optimizations.In 6 cases, antibiotherapy following positive m-PCR results was inappropriate.Three S. aureus pneumonia and one polymicrobial pneumonia (H.influenzae, S. aureus, M. catarrhalis) were treated with piperacillin-tazobactam without de-escalation.Amoxicillin-clavulanic acid was continued in one case despite the detection of gene mec A/C and in one episode of polymicrobial pneumonia despite the detection of P. aeruginosa. Figure 2 shows initial antibiotic prescriptions, changes following m-PCR results, and additional modifications after culture results.Four patients died between day 1 and day 2 after the onset of pneumonia, preventing antibiotic change depending on microbiological culture results.On day 2, in cases of a negative m-PCR, obtaining the results of culture and antibiogram never led to the cessation of antibiotics, despite a negative culture in 75.4% of cases.In cases of positive m-PCR, antibiotic therapy was modified after cultureresults in 17.5% of cases.In 4 cases, antibiotic change was motivated by resistance to initial probabilistic treatment.In 2 cases, antibiotic escalation was applied despite bacterial sensitivity to ongoing treatment.In 4 cases, negative standard culture motivated antibiotic de-escalation.
A. In case of negative PCR :

Factors Associated with Appropriate Antibiotic Strategy after m-PCR Results
Table 5 summarizes the univariate and multivariate analysis of factors associated with an appropriate antibiotic strategy following m-PCR results.In univariate analysis, factors associated with appropriate antibiotic strategy following m-PCR results were history of chronic respiratory insufficiency (p = 0.05), lower SAPSII score (p = 0.03), absence of prior antimicrobial therapy within one month (p < 0.01), and a positive m-PCR (p < 0.01).In multivariate analysis, identification of at least one bacteria by m-PCR (OR: 96,60; IC95% [9.72, 960.20], p < 0,001) and a lower SAPSII score (OR: 0.96; IC95% [0.931, 0.997], p = 0.034) were significantly associated with an early appropriate change ininitial probabilistic antibiotic therapy following the results of m-PCR.

Discussion
This study aimed to evaluate the impact of m-PCR on the real-life antibiotic management of mechanically ventilated patients suspected of pneumonia in a polyvalent intensive care unit.It demonstrated an impact of m-PCR on antibiotic strategy in nearly twothirds of cases.A positive m-PCR result allowed for the initiation or early adjustment of antibiotic therapy in almost 90% of cases.A negative m-PCR result spared antibiotic use in over onethird of cases.
In our study, m-PCR results were compared withthose of standard culture obtained 48 to 72 h after sample collection.We found a sensitivity of 50% and a specificity of 55%.The positive and negative predictive values were 29% and 75%, respectively.The concordance between m-PCR and culture was 55%, consistent with the concordance of 56% reported by Crémet et al. in a similar population [18].The concordance improved with the bacterial load.A >90% concordance was reported for culture values above 10 6 /mL [19].Higher performances have been reported in similar studies in patients with and without COVID-19,with sensitivities and specificities exceeding 80% [20][21][22].Our study revealed that m-PCR led to the identification of additional bacteria compared withstandard culture.Accordingly, Monard et al.found nearly twice as much microbiological documentation with m-PCR compared with culture [21].Buchan et al. demonstrated a 94.8% increase in thenumber of detected bacteria, while Lee et al. reported a 70.3% increase [21,23].
The false positive results of m-PCR mostly revealed H. influenzae and S. aureus, as reported by others [23,24].Crémet et al. showed that, for over 50% of m-PCR positive for H. influenzae, either the culture turned positive after using an enriched medium, or this bacterium was finally overgrown in culture by other pathogens from the commensal flora [18].S. aureus, H. influenzae, M. catarrhalis, and P. aeruginosa were the bacteria most frequently associated with false-positive m-PCR results.Murphy et al. found similar results on 845 BAL and 846 sputum and ETA [19].Standard culture has limitations since the culturing techniques are based on the detection of dominant pathogens.Minor bacteria and exigent pathogens such as H. influenzae can be missed.Given the retrospective nature of our study and its real-life conditions, no additional microbiological tests were performed in case of discordant results.The concordance between m-PCR and culture is also influenced by prior antibiotic use, which can produce negative cultures.Our results showed that 47% of false-positive m-PCR results involved patients who had received at least one dose of an antibiotic active against the identified bacterium within the 24 h preceding the sampling.Similar findings were reported by Buchan et al., with 49% of false positive m-PCR being from patients who had received antibiotics within 72 h [20].Taking into account bacteria isolated in culture at rates below the threshold of significance also reduces the discordance between the two techniques [16,17].The FilmArray ® pulmonary panel provides semiquantitative results expressed in copies/mL ranging from 10 4 to ≥10 7 copies/mL.Within this range, the concordance between m-PCR and culture is accurate within a 0.5 log difference [19,25].The culture of ETA samples often shows negative results for PCR-positive samples at 10 4 copies/mL.This raises the question of the contribution of m-PCR in patient management when positive at a low concentration with a negative culture, suggesting potential contamination of the sample with oral flora.Concordance is also improved when the results of repeated cultures surrounding m-PCR are taken into account [18,26].In our population, the majority of patients underwent ETA rather than bronchoalveolar lavage (BAL), increasing the m-PCR false positive results.The m-PCR in ETA showed a sensitivity of 57%, a specificity of 53%, a positive predictive value of 29%, and a negative predictive value of 79%.This high negative predictive value could encourage antibiotic deescalation in cases of negative m-PCR in ETA samples [27].We described 14 false negative m-PCR results involving H. alvei, Aspergillus sp., and S. maltophilia that are not included in the m-PCR panel.These results confirm that m-PCR should not be performed alone, as culture remains necessary to detect m-PCR false negativesand to perform antimicrobial susceptibility testing.
The real-life impact of the m-PCR in our ICU was evaluated by the proportion of antibiotic strategies aligned with the m-PCR results.For positive m-PCR results, 89% of patients had appropriate antibiotic strategies, including 58% initiations, 16% escalations, and 12% de-escalations.On day 2, antibiotic therapy was modified based on standard culture results in only 17.5% of cases.For negative m-PCR results, the absence of antibiotic initiation or discontinuation occurred in over onethird of our patients.Previous studies have demonstrated the potential of m-PCR in reducing unnecessary antibiotic treatment, reporting significant reductions in the duration and overall use of antibiotics.Most of these studies simulated the impact of m-PCR results by comparing the antibiotics prescribed in practice by clinicians unaware of the m-PCR results to those chosen in theory by experts informed of the m-PCR results.The anticipated proportion of antibiotic de-escalation when using m-PCR was around 40% in the literature [16,21,[28][29][30].For instance, Guillotin et al. showed only 37% of predicted broad-spectrum antibiotic therapies when using m-PCR compared with 88% when following clinical guidelines [28].Buchan et al. reported a potential de-escalation or discontinuation of antibiotic therapy based on m-PCR results in 48% of patients, resulting in an average saving of 6.2 antibiotic days/patient.Two prospective randomized studies evaluated the impact of m-PCR on antibiotic use [20].Darie et al. found a reduction in the duration of inappropriate antibiotic therapy by 38.6 h [31].However, the conclusions of this study were limited by the mild severity of the patients and the low rate of bacterial documentation [32].Farthouk et al. assessed the impact of m-PCR coupled with PCT.They did not demonstrate a significant reduction in antibiotic use, although it suggested a possible antibiotic-sparing effect [33].
The proportion of antibiotic de-escalation when using m-PCR was only 12% in our real-life cohort.Our results were comparable to those reported by Maataoui et al., who found 11% de-escalation after m-PCR results in cases of SARS-CoV-2-related pneumonia in a retrospective cohort [17].Similarly, the DIANA study, which evaluatedantibiotic de-escalation in infected intensive care unit patients, found 16% de-escalation [34].No deleterious impact of de-escalation was observed on clinical recovery.Tabah et al. reported that de-escalation was more often applied in patients with an already favorable clinical course [30].
A negative m-PCR led to antibiotic sparing by discontinuation or absence of antibiotic initiation in 34% of our patients.Maataoui [17,32].In our cohort, a high SAPSII score was significantly associated with the lack of consideration for m-PCR results.A negative culture confirming the m-PCR result did not lead to additional discontinuation of antibiotics on days 2 and 3, highlighting the reluctance of physicians to discontinue antibiotics in the most criticallyill patients.Maataoui et al. also found that patient severity encouraged the continuation of antibiotics in half of the cases for at least 48 h, despite the high NPV of a negative m-PCR [17].Conversely, a negative m-PCR in patients with minor symptoms could safely result in the absence of antibiotic therapy.The algorithm proposed by Novy et al. in cases of a negative m-PCR showed antibiotic sparing in 65% of samples.They suggested discontinuing empirical antibiotic therapy if the m-PCR is negative and the patient does not present any severity criteria, such as septic shock or ARDS, with no Gramnegative bacteria observed on direct examination [35].This strategy requires considering the local ecological risk in each ICU.
Our study has several limitations.We could not include all m-PCRperformedwhen concomitant cultures were lacking.The study was monocentric, which prevents the extension of results to other centers with different ecologies and antibiotic strategies.Due to the observational approach, factors related to the physicians, patients, and type of infection were not controlled.The suspicion of pneumonia remained at the discretion of the attending physician.The presence of other infectious foci indicating the continuation of antibiotic therapy wasnot collected, nor were clinical criteria such as clinical improvement.Samples were sometimes collected on weekends without a microbiology team available to provide rapid results, explaining the large variation in the turnaround time for m-PCR classification of pneumonia (ventilator-associated pneumonia, VAP; hospital-acquired pneumonia, HAP; community-acquired pneumonia, CAP), prior antimicrobial therapy within one month, duration of hospital and ICU stay, and severity of illness were collected.Antimicrobial prescriptions were recorded on the day before and day 0, day 1, day 2, and day 3 after m-PCR performance for suspected pneumonia.Antimicrobial changes after the results of m-PCR (d0-d1) and after the results of culture and susceptibility testing (d2-d3) were analyzed.All patients were follow-up until death or release from the ICU.

Endpoints
The main judgment criterion was the proportion of appropriate changes in initial probabilistic antibiotic therapy following m-PCR results.An appropriate change was defined as a lack of initiation or interruption of antibiotic therapy in the case of a negative m-PCR result; an appropriate initiation, escalation, optimization, or de-escalation of antibiotic therapy in the case of a positive m-PCR result.An appropriate initiation corresponded to the introduction of an effective antibiotic on the bacteria identified by m-PCR, not treated by probabilistic antibiotic therapy preceding the m-PCR result.Optimization was defined as the use of a 4th generation cephalosporin in place of a 3rd generation cephalosporin in case of detection of group 3 enterobacteria.De-escalation included switching from one beta-lactam to another one with a narrower spectrum and lighter selective pressure according to a six-rank consensual classification of beta-lactams [37].The proportion of inappropriate changes is defined by the rate of sampling for which the m-PCR result was not taken into account in the antibiotic strategy.
The secondary objectives of our study were to determine the factors significantly associated with an appropriate change in initial probabilistic antibiotic therapy following the results of them-PCRanalysis and to assess the concordance between the results of pulmonary m-PCR and those of conventional microbiological tests (gold standard).Detection by m-PCR of a bacteria not present in culture was considered false positive, and the absence on m-PCR of a bacteria present in culture was considered false negative.Cases where the cultural test detected a pathogen that could not be identified by the m-PCR were considered false negatives.Concordance was defined by complete qualitative agreement between m-PCR and conventional culture results.We particularly assessed the performance of m-PCR in the subcategory of ETA sampling.

Statistical Analysis
Continuous variables were expressed as mean values ± standard deviation or as median (interquartile range), depending on the normality of their distribution.They were compared using the Student's test or the Mann-Whitney U test, as appropriate.Categorical variables were expressed as percentages and evaluated using the chi-square test and Fisher's test when appropriate.Differences between groups were considered to be significant for variables yielding a p-value ≤ 0.05.To determine the independent effect of the variables on the appropriate change ininitial probabilistic antibiotic therapy following the results of m-PCR, we performed a logistic regression analysis using the purposeful selection of covariates.PaO 2 /FiO 2 ratio, COVID-19 status, duration of mechanical ventilation, Simplified Acute Physiology Score (SAPS) II score, antibiotic treatment at the time of diagnosis of pneumonia, and all covariates with p < 0.2 in the unadjusted model were entered into the multivariate model.To assess the concordance between the results of m-PCR and those of conventional microbiological culture (reference method), the calculation of sensibility, specificity, positive predictive value (PPV), and negative predictive value (VPN) was performed, and a concordance analysiswas conducted usingthe Cohen Kappa test globally and in the group of patients who benefited from tracheal aspiration.All statistical analyses were performed using R-software ® , version 4.3.2.

Figure 2 .
Figure 2. Antibiotic prescriptions after m-PCR and conventional culture results: (A) In case of negative PCR and (B) In case of positive PCR.

Figure 2 .
Figure 2. Antibiotic prescriptions after m-PCR and conventional culture results: (A) In case of negative PCR and (B) In case of positive PCR.2.5.Factors Associated with Appropriate Antibiotic Strategy after m-PCR ResultsTable5summarizes the univariate and multivariate analysis of factors associated with an appropriate antibiotic strategy following m-PCR results.In univariate analysis, factors associated with appropriate antibiotic strategy following m-PCR results were history of chronic respiratory insufficiency (p = 0.05), lower SAPSII score (p = 0.03), absence of prior antimicrobial therapy within one month (p < 0.01), and a positive m-PCR (p < 0.01).In multivariate analysis, identification of at least one bacteria by m-PCR (OR: 96.60; IC95% [9.72, 960.20], p < 0.001) and a lower SAPSII score (OR: 0.96; IC95% [0.931, 0.997], p = 0.034) were significantly associated with an early appropriate change ininitial probabilistic antibiotic therapy following the results of m-PCR.

Table 4 .
Proportion of appropriate antibiotic strategies following m-PCR results.

Table 4 .
Proportion of appropriate antibiotic strategies following m-PCR results.

Table 5 .
Univariate and multivariate analysis of factors associated with an appropriate antibiotic strategy following m-PCR results.

Table 5 .
Univariate and multivariate analysis of factors associated with an appropriate antibiotic strategy following m-PCR results.
et al. and Posteraro B et al. reported similar results in patients suffering from COVID-19