*3.4. Etiology of Bacterial Aspiration Pneumonia*

Of the 54 patients with BAP, 71 different bacterial strains were identified. They are presented in Table 3 with their antimicrobial susceptibility. Fifteen (28%) patients had a polymicrobial BAP. The most represented bacteria were *Staphylococcus aureus* (18; 33%), *Haemophilus influenzae* (13; 24%), *Streptococcus pneumoniae* (10; 19%) and *Klebsiella pneumoniae* (9; 17%).

**Table 3.** Bacterial strains identified by endotracheal aspirate in patients with bacterial aspiration pneumonia.



**Table 3.** *Cont.*

Of the 54 patients with BAP, 71 different bacterial strains were identified. Antimicrobial susceptibility was defined, and \* corresponds to ACA-resistant bacteria. Abbreviations: ESBL, extended spectrum beta-lactamase; ACA, amoxicillin-clavulanic acid.

### *3.5. Antibiotic Therapy*

A total of 158 patients (70%) were treated by antibiotic therapy: 138 (87%) for suspicion of BAP and 20 (13%) for documented or suspected infection other than BAP.

Among patients with suspicion of BAP, 45 patients (33%) did not have an aspiration syndrome and 120 (87%) were treated by ACA.

Among the 54 patients with BAP, 13 (24%) patients presented at least one bacterium with no ACA-susceptibility. These patients more frequently had a colonization with ESBLproducing bacteria (*p* = 0.01) and received more antibiotics during the 3 months before ICU admission (*p* < 0.01) (Supplementary Materials Table S1). Among them, initial antibiotic therapy was inappropriate in 11 patients (85%). None presented an acute respiratory syndrome distress, or a septic shock or death related to respiratory infection.

The median duration of treatment for BAP was 7 days (5–7). In comparison with BAP, patients with pneumonitis did not have a shorter duration of antibiotic therapy (*p* = 0.46).

#### *3.6. Outcomes of BAP Versus Pneumonitis*

Patients with BAP, as compared with those with pneumonitis, did not have a longer duration of MV (*p* = 0.18) and ICU stay (*p* = 0.18). Three-month mortality was 17% versus 8%, respectively, in patients with BAP and pneumonitis (*p* = 0.20). Twenty-four patients (44%) in the BAP group versus ten (27%) in pneumonitis group had a 3-month poor functional outcome (*p* = 0.06) (Table 4).


**Table 4.** Comparison of patients' outcomes according to the diagnosis of BAP or pneumonitis.

Categorical variables were expressed as number (percentage) and compared by a chi-squared test or Fisher's exact test when specified by \*. Continuous variables were expressed as median (inter-quartile range), and a *t*-test was performed (Welch or Wilcoxon tests, as appropriate). Missing values (BAP; pneumonitis): 3-month outcomes, 3 (3;0). Abbreviations: BAP, bacterial aspiration pneumonia; ICU, intensive care unit.

#### **4. Discussion**

In the present study, we found the following: (1) Half the patients with generalized convulsive SE requiring mechanical ventilation had an aspiration syndrome and onequarter suffered from BAP. (2) No clinical and laboratory characteristics allowed for the separation of BAP from pneumonitis. In particular, biomarkers related to inflammatory response such as serum CRP, PCT, and albumin were not associated with BAP and rather were modified by aspiration syndrome, including pneumonitis. (3) Up to one-quarter of bacteria isolated from tracheal samples was not sensitive to ACA, with no clinical consequence (i.e., septic shock, acute respiratory distress syndrome). (4) Finally, we did not find that BAP was associated with a poor functional outcome and death at 3 months compared to pneumonitis.

Aspiration syndrome is a frequent condition occurring in cases of impairment of consciousness [14]. In comparison with patients requiring MV for coma (epilepsy was the etiology of coma in 14%), we found a higher frequency of aspiration syndrome and BAP [7]. This result could be related to a higher risk of aspiration during SE. Indeed, in our study, persistent convulsions at admission were associated with aspiration syndrome. In another study of patients requiring MV for coma (13% had a convulsive SE), the authors found a higher frequency of aspiration syndrome (81 patients, 79%) and more BAP (45 patients, 44%) than in our cohort [15].

As previously described, aspiration syndrome was responsible for a longer duration of MV and ICU stay, without impact on ICU mortality [7].

Differentiation BAP from pneumonitis is the keystone of antibiotic management in order to avoid useless prescriptions and the emergence of resistant bacteria. In line with a previous study, we found no relevant difference between BAP and pneumonitis, including previous use of a proton pump inhibitor [7,16]. Serum levels of acute-phase proteins of inflammation were not modified by BAP. PCT increases under various inflammatory conditions, especially in the case of bacterial infections [17,18]. Some studies tried to use PCT to distinguish pneumonitis from BAP with no significant difference [7,19]. Only one study found that serum levels of CRP and PCT increased in cases of BAP with a poor diagnostic value [15]. However, they compared BAP versus no BAP (including pneumonitis and absence of aspiration syndrome), and the results could be biased by the presence of patients without aspiration syndrome who do not have increased serum levels of CRP and PCT [15]. Interestingly, early elevation of PCT is rather associated with poor neurological outcome in acute brain injury, such as SE [20,21], post-cardiac arrest syndrome [22], and stroke [23]. Currently, only microbiological culture of the tracheal sample differentiates these two entities with a result obtained in 2–3 days [24]. One elegant method to rapidly differentiate pneumonitis from BAP is to use polymerase chain reaction (PCR)-based techniques [25,26]. The results could be obtained in a few hours, reducing antimicrobial consumption in the ICU and breaking the vicious circle of multidrug-resistant bacteria emergence [27]. Furthermore, in the case of SE, some experimental and clinical data suggest that antibiotic therapy could increase the risk of symptomatic seizures in combination with renal dysfunction, brain lesion, and epilepsy [28].

Microbiology of BAP with a predominance of *Staphylococcus aureus*, *Haemophilus influenzae*, and *Streptococcus pneumoniae* is in line with a previous study [29].

Infections in SE are already known as factors associated with poor outcome [5]. The impact on MV duration and ICU stay is uncertain, in contrast to aspiration syndrome, but could be increased in BAP [7,15]. Further, previous studies reported the absence of significant relationship between BAP and ICU mortality [7,15]. However, our study follow-up at 3 months highlighted a possible higher mortality and morbidity in BAP versus pneumonitis. A large prospective study is needed to confirm these results.

Antibiotic therapy was frequently initiated for suspicion of BAP. However, one-third of patients did not have aspiration syndrome. Two reasons could explain this early antibiotic treatment: the presence of macroaspiration [30] and the elevation of inflammatory biomarkers such as serum levels of CRP and PCT. We found that macroaspiration, defined by history of vomiting before or during intubation, was not predictive of aspiration syndrome; and modifications of acute-phase proteins of inflammation, such as CRP and PCT were associated with aspiration syndrome rather than BAP.

One-quarter of patients with BAP had at least one bacterium that was non-susceptible to ACA. ACA, as alternative to ceftriaxone, is recommended for treatment of respiratory infection occurring during the first 5 days after admission [31]. Antibiotics before ICU admission and colonization with ESBL-producing bacterium are well known factors associated with ACA resistance [31]. In another study of early onset VAP (2–5 days after mechanical ventilation starting, corresponding to ICU admission) in a French neuro-ICU, the authors found 36% of patients with at least one bacterium resistant to ACA, mainly in patients with antibiotics before admission [32]. These data are in line with previous studies [31]. It seems important to determine new specific factors associated with ACA resistance and to develop a PCR-based method to rapidly identify antimicrobial resistance.

The median duration of antibiotic therapy for suspicion of BAP was 7 days, in line with the actual recommendation [33]. However, in the case of pneumonitis, we did not show a shorter duration, probably due to the delay in obtaining the result of microbiological culture and the fear of false-negative results [30]. In one study of patients with coma requiring MV, discontinuation of antibiotic therapy when no microorganism was found did not increase morbidity or mortality and tended to decrease the time with antibiotics during first 8 days [7]. The excellent predictive value of PCR-based methods would be interesting to use early on to differentiate pneumonitis from BAP, in order to avoid antibiotic treatment or to allow early withdrawal in patients with negative results and so decrease antibiotic consumption. This strategy is being studied in an ongoing randomized controlled trial (NCT03763799).

Our study has some strengths. Patients were similar in their clinical presentation with only GCSE. Post-anoxic SE was excluded due to the heterogeneity of their management [8]. The monocentric design contributed to a homogeneous management, especially in SE treatment and BAP diagnosis. Baseline characteristics and the main results were in line with previous studies [7,15]. Infection was considered only when confirmed by culture.

Our study also has some limitations. Monocentric design contributes to the small size of the cohort and results should be applied only for generalized convulsive SE. However, our main results are similar to those of a study including different causes of coma [7]. The long duration of data collection could impact the results. However, the methodology to determine infection did not change during the study. The primary outcome was the incidence of bacterial aspiration pneumonia, and a potential modification of bacterial ecology could not affect this variable. We did not have a control group to compare incidence of BAP between patients with SE and patients with impaired consciousness not due to SE. No anaerobic microorganisms were identified by microbiological cultures. They may play a role in BAP; however, they are often mixed with aerobic bacteria [29].
