Airway Pseudomonas Aeruginosa Density in Mechanically Ventilated Patients: Clinical Impact and Relation to Therapeutic Ecacy of Antibiotics

Background: The bacterial density of Pseudomonas aeruginosa (PA) is closely related to its pathogenicity. We evaluated the effect of airway PA density on the clinical course of mechanically ventilated (MV) patients and the therapeutic ecacy of antibiotics. Methods: We retrospectively analyzed data of intensive care unit (ICU) MV patients with PA isolated from endotracheal aspirates (ETAs). Patients were divided into three groups according to the peak PA density during ICU stay: low-PA ( ≤ 10 4 cfu/mL), moderate-PA (10 5 ‒ 10 6 cfu/mL), and high-PA ( ≥ 10 7 cfu/mL) groups. The relationship between PA density and weaning from MV, risk factors for high-PA isolation, and antibiotic ecacy were investigated using multivariate and propensity score-matched analyses. Results: Four-hundred-and-sixty-one patients were enrolled. Patients with high-PA had higher inammation and developed more severe respiratory infections. High PA was independently associated with few ventilator-free days on day 28 (P < 0.01) and increased ICU mortality (P = 0.047). Risk factors for high PA were prolonged MV (odd ratio [OR] 3.07 95% condence interval [CI] 1.35 ‒ 6.97), non-antipseudomonal cephalosporins (OR 2.17, 95% CI 1.35 ‒ 3.49), hyperglycemia (OR 2.01, 95% CI 1.26 ‒ 3.22) during ICU stay, and respiratory diseases (OR 1.9, 95% CI 1.12 ‒ 3.23). Isolation of normal respiratory ora was associated with lower risks of high PA (OR 0.43, 95% CI 0.26 ‒ 0.73). Propensity score-matched analysis revealed that antibiotic therapy for patients with ventilator-associated tracheobronchitis (VAT) improved weaning from MV only in the high-PA group. Conclusions: High PA signicantly affected the clinical course of MV patients and could be a good therapeutic indicator for VAT treatment.

between PA density and clinical features of patients undergoing MV, showing patients with high PA density were likely to have more virulent strains [10]. The present study aimed to investigate the effect of PA density on the clinical course and therapeutic e cacy of antibiotics in MV patients. We also evaluated the risk factors for a high PA density.

Study Design and Population
This was a retrospective cohort study performed at Kumamoto University Hospital, a tertiary care teaching hospital in Japan. We included all ICU patients with PA isolated from respiratory specimens during their ICU stay between January 2004 and December 2019. Patients younger than 15 years or requiring MV for less than 24 hours were excluded from the study.

Data Collection And De nitions
The patients were divided into three groups according to the peak density of PA in endotracheal aspirates (ETA) obtained during their ICU stay: low PA density (low-PA) (≤ 10 4 cfu/mL), moderate PA density (moderate-PA) (10 5-10 6 cfu/mL), and high PA density (high-PA) (≥ 10 7 cfu/mL) groups (Fig. 1). These cutoff points for the three groups were chosen as a bacterial density of 10 5-10 6 cfu/mL in ETA is frequently used as the threshold for diagnosing VA-LRTI [4,12,13]. For patients in whom the same peak density was detected multiple times, the rst event was included in the analysis.
Disease severity on ICU admission and at peak PA density in isolates was assessed using Acute Physiology and Chronic Health Evaluation (APACHE) II score [15], Sequential Organ Failure Assessment (SOFA) score [16], and simpli ed Clinical Pulmonary Infection Score (CPIS) [17]. Patients' comorbid conditions were assessed using Charlson comorbidity index [18]. Patients were considered to have hyperglycemia if blood glucose exceeded 200 mg/dL on two or more occasions during the ICU stay.
In the microbiological analysis of ETA, serial dilutions of samples were plated on blood, MacConkey and chocolate agar. The numbers of colonies were counted after incubation, and pathogens were identi ed by automated VITEK system (bioMérieux, Tokyo, Japan). Antibiotic sensitivities were determined using microdilution methods [11]. The number of neutrophils in ETA was semi-quanti ed from Gram staining samples.
Normal respiratory ora (NRF) was considered positive if at least one of the following microorganisms was isolated: α-hemolytic streptococci, coagulase-negative staphylococci, non-pathogenic Neisseria species, Corynebacterium species, and Candida species [19]. Antibiotic therapy was de ned as appropriate when PA showed in vitro susceptibility to empiric antibiotics (administered within 24 hours of sampling for culture) for patients with VAP or targeted antibiotics (started or changed based on culture results) for patients with VAT [4,20].
Clinical outcomes of this study were ventilator-free days (VFDs) on day 28, successful weaning rates, and ICU mortality. VFDs at day 28 were de ned as the number of days that patients were both alive and free from MV during the 28 days after the peak density of PA was isolated (Fig. 1).

Statistical Analysis
Differences among three groups were evaluated using the Kruskal-Wallis test for continuous variables and Chisquare test for categorical variables.
The impact of airway PA density on outcomes was investigated by multivariate linear regression analysis for VFDs at day 28, and by logistic regression analysis for ICU mortality, after adjusting for baseline patient characteristics. Multivariate logistic regression analysis was also used to identify risk factors for high-PA in MV patients. All variables marginally signi cant in the univariate analysis (P < 0.15) and those variables associated with colonization or VAP due to PA (age, respiratory disease, APACHE II at ICU admission, hyperglycemia during ICU stay, presence of NRF or other pathogenic bacteria in the airway, antibiotic therapy during ICU stay) were included in a logistic regression model, based on previous research [21,22,23,24].
Propensity score-matched analysis was performed to investigate the relationship between antibiotic therapy and outcomes. The propensity score was estimated using logistic regression analysis with the following 10 variables: age, sex, admission category, comorbid respiratory diseases, APACHE II score at ICU admission, SOFA, CPIS, albumin level, chest X-ray ndings, and length of MV when peak density of PA was isolated. These variables were selected based on previous reports and clinical signi cance, suggesting a potential relationship with both antibiotic therapy assignment and outcomes [17,25,26,27]. Patients who received appropriate antibiotic therapy (AAT) were matched 1:1 with those who did not, by using the nearest-neighbor matching method with a caliper of 0.2. After matching, clinical outcomes were compared with the Wilcoxon signed-rank test and Fisher's exact test. Kaplan-Meier analysis was also used to compare the time to successful weaning from MV. This propensity score-matched analysis was performed separately for each PA density group.
A P-value < 0.05 was considered to be statistically signi cant. All statistical analyses were performed using SPSS Statistics 23.0 (SPSS Inc., Chicago, USA).

Baseline Characteristics of the Study Population
Of the 3,465 screened patients, 461 patients were enrolled for analysis and classi ed into three group as follows: low-PA (n = 171), moderate-PA (n = 149), and high-PA (n = 141) groups (Fig. 1). The baseline characteristics are shown in Table 1. No signi cant differences were observed among the three groups in terms of demographic characteristics, admission category, and illness severity parameters. Almost half of the patients were admitted to the ICU for reasons such as respiratory failure, pneumonia, and septic shock. The most common comorbid condition was malignancy. Patients in the high-PA group were more likely to have respiratory disease (P < 0.01). The details of respiratory disease were as follows: chronic obstructive pulmonary disease (COPD), 42%; interstitial lung disease, 31.8%; bronchial asthma, 21.6%; bronchiectasis, 12.5%. Laboratory ndings at ICU

Clinical Features And Outcomes
The clinical features at the time-point when the peak density of PA was detected are presented in Table 2. The frequency of VA-LRTI was less than 50% in the low-PA group, whereas it was about 80% in the moderate-and high-PA group. VAT accounted for about 70-80% of VA-LRTI in each group, and the frequency of VAP tended to be higher in high-PA group (P < 0.01). SOFA scores did not differ signi cantly among the three groups, whereas the CPIS tended to increase as the PA density increased. Levels of systemic in ammation markers, including white blood cell (WBC) count and C-reactive protein (CRP), were also higher in the high-PA group than in other groups (P < 0.01). Chest X-rays showed that pulmonary in ltrate lesions were more diffuse in the high-PA group. In the analysis of ETA, the number of neutrophil cells increased as the PA density increased (P < 0.01). Regarding pathogenic bacteria besides PA, counts of Stenotrophomonas maltophilia were slightly lower in the high-PA group. VFDs at 28 days and ICU mortality were worse in the higher density groups.

Antibiotic Therapy And Outcomes
VFDs and ICU mortality in patients without VA-LRTI were similar in all groups, with median values of 20-26 days and 4-13%, respectively, and were not signi cantly different in patients receiving AAT or inappropriate antibiotic therapy (IAAT). Association of antibiotic therapy with outcomes in patients with VA-LRTI are shown in Table 4. In patients with VAT, VFDs in low-and moderate-PA groups did not vary signi cantly with appropriateness for antibiotic therapy. On the other hand, in the high-PA group, VFDs of patients who received IAAT were markedly lower than the VFDs of those who received AAT. Patients with VAP had shorter VFDs, particularly in the moderateand high-PA groups. ICU mortality tended to be decreased in VAP patients in the low-and moderate-PA groups who received AAT, as compared to those who did not (0-16.7% vs. 45.5-46.2%). Patients with high-PA VAP had a very high mortality rate, even with AAT (41.7%). We performed separate propensity score matching for VAT patients within each density group (Table 4, right panel). After matching, although not statistically signi cant due to the small numbers, AAT was also associated with an improvement in VFDs, but only in the high-PA group (median 0 vs. 17 days, P = 0.06), and not in the low-and moderate-PA groups. Kaplan-Meier analyses among matched patients showed that weaning from MV was almost identical between AAT and IAAT in both the lowand moderate-PA groups, while, in the high-PA group, a lower and later incidence of weaning success were observed with IAAT as compared to with AAT (Fig. 2).

Discussion
In this study, we investigated the relationship between airway PA density and clinical outcomes in MV patients, as well as the effects of antibiotic treatment. To the best of our knowledge, no previous study has highlighted the importance of airway PA density in MV patients or identi ed the risk factors for high PA. We found that high-PA was associated with worse clinical outcomes. Prolonged MV, non-antipseudomonal cephalosporins, hyperglycemia, and respiratory diseases were associated with higher risk and nding NRF was associated with lower risk for high-PA. Moreover, AAT for VAT patients was related to improved weaning from MV, but only in the high-PA group.
The relationship between airway PA density and clinical course has been reported for various respiratory diseases. A previous analysis of 385 patients with bronchiectasis showed that bacterial density was directly correlated with airway and systemic in ammatory markers, such as myeloperoxidase activity, neutrophil elastase activity, and tumor necrosis factor-α [28]. Airway bacterial density was also associated with exacerbation frequency and health-related quality of life. In other studies of cystic brosis and COPD patients, the higher the density of airway bacteria, the more airway neutrophils were increased and activated [29,30]. We found that MV patients with high-PA had higher airway (neutrophils in the ETA) and systemic in ammation. The frequency of VA-LRTI was low in the low-PA group, while the high-PA group had more cases of VAP than other groups. Patients with high-PA may develop more severe respiratory infections, re ecting the higher in ammatory response in cases with high bacterial density. Our study found that the high-PA group had prolonged weaning from MV, and higher ICU mortality. These results were similar when adjusted for the severity scores by multivariate analysis. Thus, the pathogenicity of PA, particularly when present at a high density in the airway, is expected to have a signi cant impact on the in ammatory response and the clinical course in MV patients.
PA was isolated from the airway in 18% of patients admitted to the ICU at our institution. Deciding which of these patients should be treated with antibiotics is an important issue in daily practice, particularly for patients with VAT, as an intermediate stage between airway colonization and VAP. The rst randomized control trial of treatment for VAT demonstrated that patients who received AAT had higher VFDs (median 12 vs. 2 days) and lower ICU mortality (18% vs. 47%) than the no antibiotic group [31]. On the other hand, several studies reported no signi cant bene ts of AAT [32,33]. The discrepancy of these results could be partly due to differences in the de nition of VAT, causative bacteria, and antibiotic regimen used across studies. A recent guideline does not recommend routine use of antibiotics for VAT, given the inconsistent evidence of clinical bene t and the problem of adverse drug events [34]. The guideline suggests assessing the condition of individual patients and considering antibiotic therapy depending on disease severity. However, it is not yet known in which subgroups antibiotic therapy is particularly effective for VAT. In the present study, the number of airway neutrophils in VAT patients with high-PA was higher than that in other groups, suggesting that antibiotics may be particularly effective in these patients. Indeed, in our propensity score-matched analysis, AAT improved MV weaning outcomes only in VAT patients with high-PA. This relationship between PA density and the therapeutic effect of antibiotics was also supported by a recent study of patients with bronchiectasis [35].
We explored the risk factors associated with high-PA and identi ed ve signi cant variables: respiratory diseases at ICU admission, prolonged MV (> 28 days), non-antipseudomonal cephalosporins, hyperglycemia, and NRF during ICU stay. Respiratory diseases, duration of MV, and ineffective antibiotics against PA are well known risk factors for PA colonization and infection in MV patients [21,22,23]. We newly identi ed hyperglycemia as an important risk factor for high-PA. In vitro and animal experiments showed that, as the blood glucose increased, the glucose concentration in the airway surface liquid also increased, which promotes the growth of airway PA, and causes severe pneumonia [36,37]. In human trials, airway glucose concentration was affected by blood glucose level and the presence of respiratory disease. Critically ill patients receiving MV showed high airway glucose levels [38]. For these reasons, airways of MV patients with hyperglycemia may provide a more favorable environment for PA growth. We also found that isolation of NRF during the ICU stay was associated with a lower risk of high-PA. Bacteria have complex effects on each other. For example, Candida sp. downregulates quorum sensing of PA by farnesol [39], while oral commensal Streptococcus sp. suppresses the growth of PA by producing hydrogen peroxide [40]. Thus, NRF may have protective effects against the growth of PA.
Our study had several limitations. First, as it was a retrospective analysis from a single ICU, our ndings are subject to bias and may not be generalizable to all ICU patients. Second, 27.1% of our patients had been exposed to anti-pseudomonal antibiotics at the time of collecting the ETA sample, which may have affected the density of the isolated PA. However, the proportion of patients receiving anti-pseudomonal antibiotics was comparable among the three groups. Third, we were unable to evaluate antibiotic therapy in detail, including the regimens and length of administration, due to the small number of patients.

Conclusion
Airway PA density was associated with the clinical course and therapeutic e cacy of antibiotics used for patients on MV. To prevent ineffective antibiotic therapy and reduce antibiotic use, we suggest a strict control of risk factors for high-PA identi ed in this study and the use of targeted antibiotic therapy for patients with VAT, particularly when high-PA are isolated.