Outcomes of Stenotrophomonas maltophilia hospital-acquired pneumonia in intensive care unit: a nationwide retrospective study

Background There is little descriptive data on Stenotrophomonas maltophilia hospital-acquired pneumonia (HAP) in critically ill patients. The optimal modalities of antimicrobial therapy remain to be determined. Our objective was to describe the epidemiology and prognostic factors associated with S. maltophilia pneumonia, focusing on antimicrobial therapy. Methods This nationwide retrospective study included all patients admitted to 25 French mixed intensive care units between 2012 and 2017 with hospital-acquired S. maltophilia HAP during intensive care unit stay. Primary endpoint was time to in-hospital death. Secondary endpoints included microbiologic effectiveness and antimicrobial therapeutic modalities such as delay to appropriate antimicrobial treatment, mono versus combination therapy, and duration of antimicrobial therapy. Results Of the 282 patients included, 84% were intubated at S. maltophilia HAP diagnosis for duration of 11 [5–18] days. The Simplified Acute Physiology Score II was 47 [36–63], and the in-hospital mortality was 49.7%. Underlying chronic pulmonary comorbidities were present in 14.1% of cases. Empirical antimicrobial therapy was considered effective on S. maltophilia according to susceptibility patterns in only 30% of cases. Delay to appropriate antimicrobial treatment had, however, no significant impact on the primary endpoint. Survival analysis did not show any benefit from combination antimicrobial therapy (HR = 1.27, 95%CI [0.88; 1.83], p = 0.20) or prolonged antimicrobial therapy for more than 7 days (HR = 1.06, 95%CI [0.6; 1.86], p = 0.84). No differences were noted in in-hospital death irrespective of an appropriate and timely empiric antimicrobial therapy between mono- versus polymicrobial S. maltophilia HAP (p = 0.273). The duration of ventilation prior to S. maltophilia HAP diagnosis and ICU length of stay were shorter in patients with monomicrobial S. maltophilia HAP (p = 0.031 and p = 0.034 respectively). Conclusions S. maltophilia HAP occurred in severe, long-stay intensive care patients who mainly required prolonged invasive ventilation. Empirical antimicrobial therapy was barely effective while antimicrobial treatment modalities had no significant impact on hospital survival. Trial registration clinicaltrials.gov, NCT03506191


Background
Stenotrophomonas maltophilia is one of the 10 most frequently isolated pathogens responsible for hospitalacquired pneumonias (HAPs) in intensive care unit (ICU) patients in western countries [1,2], representing approximately 5% of positive pulmonary samples. Previous studies identified several risk factors for developing S. maltophilia HAP in critically ill patients, such as prolonged ICU hospitalization associated with invasive procedures, extended periods of mechanical ventilation, or exposure to broad-spectrum antibiotics [3][4][5]. Therefore, S. maltophilia pneumonia occurs preferentially in patients with the poorest prognosis [6,7]. However, these studies were conducted from heterogeneous and small cohorts of patients. The severity of S. maltophilia HAP and antimicrobial therapy modalities were sparsely reported [3,4]. Hence, data are lacking to draw recommendations on the optimal therapeutic strategies against S. maltophilia pneumonia.
We undertook a large nationwide multicenter retrospective study with the main objective to demonstrate that modalities of antibiotic therapy, including empirical antimicrobial choice, whether a combination therapy was used, or the duration of the therapy, would influence in-hospital mortality. Secondary objectives were to describe the characteristics of ICU patients with S. maltophilia HAP and to draw prognostic factors of these pneumonias.

Design of the study and setting
The medical records of patients who experienced S. maltophilia pneumonia from January 2012 to January 2017 were collected from 25 ICUs of the French Society of Anaesthesia & Intensive Care Medicine (SFAR) and AZUREA networks [8]. Participating centers and casemixes are listed in Additional file 1.
The collected data involved both ICU and hospital stays. Follow-up was stopped either after hospital discharge or death, whichever occurred first.

Eligibility criteria
All patients aged over 18 years who were admitted to the participating ICUs and presenting with a documented diagnosis of S. maltophilia pneumonia during their ICU stay were eligible.

Source and method of selection
The patient's files were extracted through French hospital discharge database containing individual records of all hospital stays using International Classification of Disease (ICD-10) for the terms "Stenotrophomonas maltophilia" and "pneumonia." In addition, ICU medical charts were cross-checked with microbiology laboratoryspecific information systems to ensure exhaustivity.
Each medical record was analyzed by local investigators to determine if clinical, biological, and/or radiological signs of S. maltophilia HAP were present, thus excluding respiratory tract colonizations (defined as a positive respiratory sample without clinical, biological, and/or radiological signs of S. maltophilia pneumonia). In case of uncertainty, consensus was obtained between local infectious disease specialists and study coordinators (PG, AB) to clarify S. maltophilia HAP cases.

Definitions
Pneumonia was defined as follows: (i) new or progressive lung infiltrate, (ii) temperature > 38°C or < 36.5°C, leukocyte count > 12,000 μl −1 or < 4000 μl −1 , purulent endotracheal aspirate or sputum, (iii) positive respiratory sample (see below), and (iv) decline in oxygenation [9,10]. HAP was defined as a pneumonia not incubating at the time of hospital admission and occurring 48 h or more after admission. Ventilator-associated pneumonia (VAP) was defined as a pneumonia occurring 48 h or more after tracheal intubation [9].
The clinical cure of S. maltophilia pneumonia was defined by the absence of pneumonia criteria 48 h after antimicrobial therapy cessation. Treatment failure was defined as a failure of first-line treatment or death attributable to S. maltophilia pneumonia. Recurrence was defined as the onset of new pneumonia criteria associated with a positive respiratory sample with S. maltophilia after the initial pneumonia was considered successfully cured.
Empirical antimicrobial therapy was defined as the first agents prescribed for the initial treatment of HAP (effective or not on S. maltophilia) finally diagnosed as being caused by S. maltophilia. Empirical antimicrobial therapy was considered as effective if the S. maltophilia strain cultured from the respiratory sample was susceptible to at least one of the antimicrobial agents. Combination therapy was defined as the administration of at least two antimicrobial agents a priori (before S. maltophilia HAP has been confirmed, usually within 48 h) or a posteriori (after S. maltophilia HAP has been confirmed) effective on the S. maltophilia strain for more than 24 h.

Data collection
Usual demographic variables were collected, including previous hospital stays and previous exposure to antimicrobial therapies (agents and durations). Simplified Acute Physiology Score II (SAPSII) and the Sequential Organ Failure Assessment (SOFA) score were assessed.
On the day of S. maltophilia HAP diagnosis, the SOFA score was collected, as well as the number and type of invasive devices inserted. The severity of hypoxemia was graded according to the Berlin acute respiratory distress syndrome (ARDS) criteria [11]. Requirements for highflow nasal oxygen therapy, non-invasive or invasive mechanical ventilation, or extracorporeal membrane oxygenation (ECMO) were reported. Empirical antimicrobial therapy and secondary adaptations were recorded, as were durations.

Antimicrobial susceptibility testing (AST)
S. maltophilia identification characteristics (date of isolation and type of respiratory tract sampling) and antimicrobial susceptibility testing were independently performed by each microbiology laboratory. AST was performed on isolates using disk diffusion or automated testing methods according to guidelines and breakpoints established by the European Committee on Antimicrobial Susceptibility Testing (EUCAST) [13].

Data management
Data were collected and managed using Research Electronic Data Capture (REDCap) software [14]. The database was approved by the institutional review board of the SFAR (IRB00010254-2015-010), which waived the need for signed informed consent of the participants, in accordance with the French legislation on noninterventional studies [15]. The study was declared on clinicaltrials.gov (NCT03506191). This manuscript was written in accordance with the STROBE statement for the reporting of observational studies in epidemiology.

Statistical analysis
The results are expressed as the number of patients (%) for categorical variables and mean (± standard deviation) or median [IQR] for continuous variables.
Prognostic factors associated with time to in-hospital death were studied using the Cox proportional hazard model. Time to in-hospital death was calculated from the diagnostic date of S. maltophilia to death. The follow-up was censored at discharge from the ICU and/ or the hospital. Baseline prognostic factors were age, SAPS II, mechanical ventilation at diagnosis, VAP, duration of mechanical ventilation before the diagnosis, SOFA score at diagnosis, bacteremia, mono/polymicrobial pneumonia, use of empirical antimicrobial therapy, and use of empirical antimicrobial therapy effective against S. maltophilia. Other antimicrobial therapyrelated variables were not defined as baseline and were thus entered in the model as time-dependent variables, including time elapsed between sample and effective antimicrobial therapy, use of effective combination antimicrobial therapy, and duration of effective antimicrobial therapy against S. maltophilia (monotherapy or combination therapy).
Baseline and time-dependent variables associated (p < 0.05) with outcome in the univariate analysis and that were present at the diagnosis were considered for the multivariate model, and the final model was selected using backward stepwise regression (p < 0.05). Hazard ratios (HR) were calculated accordingly with their 95% confidence intervals (CI).
We compared the time to in-hospital death between patients who received or not an empirical antimicrobial therapy effective against S. maltophilia using propensity score framework. The variables used for propensity score estimations were age, sex, SOFA score at diagnosis, SAPS II, and the ICU length of stay before pneumonia diagnosis. The two groups of patients were matched using a 1:1 nearest neighbor matching algorithm with replacement, using a caliper of 0.2 of the standard deviation of the propensity score on the logit scale [16]. Covariate balance between the two groups was assessed after matching, and we considered an absolute standardized difference (ASD) less than 0.1 as evidence of balance [17]. Then, time to in-hospital death was compared between matched groups using a Cox proportional hazard model. The 95% confidence intervals of the estimated hazard ratio (empirical antimicrobial therapy yes vs no) were estimated using robust standard error [18].
Significance was defined as p values < 0.05. Statistical tests were two-sided. Statistical analyses were performed using R 3.5.0 (R Foundation for Statistical Computing, http://www.R-project.org/, Vienna, Austria).

Description of Stenotrophomonas maltophilia hospitalacquired pneumonia
Characteristics of S. maltophilia HAP are described in Table 2. Briefly, 41.6% of S. maltophilia pneumonias were monomicrobial and 80.8% were VAP. Blood culture was concomitantly positive in only 7.1% of cases. Microbiological diagnosis methods for isolation of S. maltophilia are presented in Additional file 3: Table S2 in the online data supplement. Patients with S. maltophilia VAP had a duration of mechanical ventilation before the onset of pneumonia of 11 [5][6][7][8][9][10][11][12][13][14][15][16][17][18] days. S. maltophilia pneumonia-related septic shock was present in 123 patients (43.6%) within 48 h of S. maltophilia HAP (septic shock attributed to pneumonia by clinicians and without other identified cause in the post hoc analysis). Among these patients who developed septic shock, 38 (30.8%) did not receive initial empirical antimicrobial therapy. Forty-nine percent of patients fulfilled moderate or severe ARDS criteria.

Prognosis
Treatment failure occurred in 65 patients (23.1%). Recurrence of S. maltophilia pneumonia was diagnosed in   Table 4). The subsequent occurrence of a septic shock was significantly associated with an increased risk of death (HR = 3.070, 95%CI [1.9; 5.0], p < 0.0001).
Neither the duration of treatment nor the use of combination therapy directed against S. maltophilia was associated with the primary endpoint (Table 4). Other commonly reported risk factors for S. maltophilia HAP (i.e., immunosuppression, chronic obstructive pulmonary disease (COPD), prior antimicrobial therapy) were not statistically associated with time to in-hospital death.
In multivariate analysis, only age (HR = 1.02, 95% CI [1.01; 1.04], p = 0.001) and SOFA score at S. maltophilia pneumonia diagnosis (HR = 1.1, 95%CI [1.06; 1.15], p < 0.001) were associated with in-hospital death ( Table 4). Subsequent septic shock was not included in the multivariate analysis because it was a consequence and not present at the diagnosis of S. maltophilia. The results were similar when only considering VAP in our cohort (N = 228) (Additional file 5: Table S4). Finally, we performed a statistical analysis based on a propensity score to evaluate the effect of an empirical antibiotic therapy effective on S. maltophilia on the primary endpoint (Additional file 6: Table S5). After matching, we compared 48 patients who received appropriate empirical antibiotic therapy versus 222 who did not (Additional file 7: Table S6). This analysis confirmed the previous results (HR = 0.891, 95%CI [0.498-1.593], p = 0.697).

Mono-and polymicrobial S. maltophilia HAP
The aforementioned results remained unchanged when considering only monomicrobial S. maltophilia pneumonia (n = 117) (HR = 1.08, 95%CI [1.01; 1.15], p = 0.021 for SOFA score at S. maltophilia pneumonia diagnosis). Comparisons of characteristics and outcomes between mono-and polymicrobial S. maltophilia HAP are provided in Table 5. No differences were noted in in-hospital death irrespective of an appropriate and timely empiric antimicrobial therapy between mono-versus polymicrobial S. maltophilia HAP. A similar number of VAP occurred in both groups, 91 (77.8%) versus 136 (82.9%) (p = 0.280) for mono-and polymicrobial HAP respectively. The duration of ventilation prior to S. maltophilia HAP diagnosis and ICU length of stay were shorter in patients with monomicrobial S. maltophilia HAP.

Discussion
Herein, we report the largest cohort study of critically ill patients developing S. maltophilia HAP. Regarding the large screening, the prevalence of S. maltophilia HAP remained very low. The majority of S. maltophilia HAP was VAP and occurred in patients ventilated for more than 10 days and previously exposed to several antimicrobial therapies. The mortality rate of these patients remained high, but surprisingly, the treatment delay in adequate antimicrobial therapy targeting S. maltophilia was not found to be associated with mortality. This observation may be the result of (i) a low virulence of the pathogen, (ii) the underlying condition of the critically ill patient being more contributive to the outcome than S. maltophilia HAP itself, or (iii) a 24-to 48-h delay in the treatment of S. maltophilia HAP had no real impact. Finally, if SAP-SII and SOFA score were independently associated with mortality, no specific pneumonia or antimicrobial therapy-related factors impacted the outcome.
Our study population shares common features with previously published reports [3,4,6,7,[19][20][21][22][23]. Indeed, S. maltophilia pneumonia develops in high-risk phenotypes patients, i.e., long ICU/hospital length of stay and duration of mechanical ventilation. Despite a mortality rate of approximately 50%, it is difficult to delineate direct attributable mortality of S. maltophilia HAP from mortality linked to underlying diseases [24]. Indeed, the prolonged ICU length of stay preceding S. maltophilia isolation and the number of prior antimicrobial therapies suggest noticeable patient frailty and complicated medical history. These factors have been associated to high mortality in patients with resistant bacteria in ICU [25]. S. maltophilia HAP could also be perceived as a final Previous studies that included a small number of patients [3,4,6,[19][20][21][22][23][26][27][28][29][30][31] suggested that immunecompromised conditions, COPD, prior cardiac surgery, or prior antimicrobial therapy were risk factors for S. maltophilia HAP. Conversely, our large series of mixed ICU patients did not confirm these elements. This implies that the involvement of S. maltophilia in late onset HAP should be considered and be kept in mind by all critical care physicians. However, in our series, initial antimicrobial therapy inactive against S. maltophilia was not a risk factor for in-hospital mortality, arguing against a systematic coverage of S. maltophilia by empirical antimicrobial therapy in this setting.
Although the prolonged duration of antimicrobial treatment is a well-known risk factor for emergence of multidrug resistant (MDR) bacteria [32], it did not appear to be discriminant in our study, irrespective of the class of antimicrobial agent previously administered. These results are in accordance with previously published literature on continuation or de-escalation of beta-lactam antibiotics and emergence of MDR [31]. Indeed, different regimens were used in our population, with various durations of treatment before S. maltophilia HAP diagnosis without apparent consequences on S. maltophilia emergence and susceptibility profiles. Soubirou et al. found that the increase in use of antimicrobial class was an independent predictor of S. maltophilia emergence in VAP [33]. It is however conceivable that some patients of our cohort may be colonized with other non-fermenting Gram-negative bacilli. Actually, almost 20% of patients had COPD or chronic respiratory insufficiency and might be regularly exposed to antibiotics. Nseir et al. and Saugel et al. reported 63% and 25% respectively incidence of COPD patients with S. maltophilia pneumonia [3,4].
Despite clinical signs of HAP, only 59% of patients readily received empirical antimicrobial therapy (Additional file 4: Table S3). This highlights the variable implementation and adherence to antimicrobial bundles of care and stewardship programs [34][35][36]. Pathmanathan et al. previously reported no measurable impact of antibiotic therapy in patients without evidence of consolidation which suggests colonization [23]. In our study, colonization was excluded. Although it is currently suggested to start antibiotics early in patients with suspected VAP [37], physicians may have been expecting a definitive identification with the resistance profile of microorganisms possibly involved to restrain the use of broad-spectrum antibiotics, especially in patients previously exposed to several antibiotic regimens. Tracheobronchitis and pneumonia may also be hard to be differentiated and need time to be distinguished.
Of note, the interplay between resistance and virulence remains complex [38]. In these patients already exposed to several series of antimicrobial therapies, with an extended hospital length of stay, the likelihood of MDR bacteria involvement was very high. Moreover, due to its natural resistance to multiple antimicrobial agents, only one third of empirical antimicrobial therapies was actually effective against S. maltophilia. Conversely to other authors [39], but in accordance with studies on Pseudomonas aeruginosa VAP, the delayed administration of effective antimicrobial treatment was not statistically associated with increased mortality [40,41].
The duration of S. maltophilia HAP antimicrobial therapy is still subject to debate. The comparison of short (< 8 days) versus prolonged (8 days and greater) antibiotic course could not properly be investigated in our study because of its design. Chastre et al. demonstrated that an 8-day course of antibiotic therapy for VAP was not inferior to a longer duration, but only 0.8% of patients had an S. maltophilia VAP [42]. However, it was suggested that patients infected with difficult-to-treat pathogens, immunocompromised patients, and patients at high risk for relapse may require a longer duration of antibiotic therapy. In our study, we identified neither the duration of antimicrobial treatment nor the combination of antibiotic therapies as significant risk factors for in-hospital death. Low virulence of S. maltophilia strains may partially explain these findings. In a recent retrospective study focused on the interest of combination therapy, Shah et al. reported that combination of antibiotic therapies yielded similar clinical efficacy and resistance development compared to monotherapy [43].
Optimal antimicrobial against S. maltophilia HAP may raise some concerns. The S. maltophilia strains in our study had a preserved susceptibility to ticarcillin-clavulanate and TMP-SMX, 73 and 88% respectively as expected [44]. However, only 29% of S. maltophilia HAP were treated with TMP-SMX. These discrepancies may be related to ICU physicians' habits, the fear of TMP-SMX-related side effects, and the type of antibiotics administered prior to occurrence of S. maltophilia HAP. The use of fluoroquinolones could have been considered easier. When prescribed, TMP-SMX was combined with another antibiotic effective on S. maltophilia in 80% of cases. In addition, antimicrobial agent shortages change antimicrobial therapy armamentarium, with a cessation of manufacture of ticarcillin-clavulanate in 2015 [45]. Although ceftazidime/avibactam has poor activity on S. maltophilia [46], it may restore the susceptibility to aztreonam through the inhibition of the L2 β-lactamase in vitro [47]. The clinical efficacy of this combination has been reported in a case report of a transplant renal patient [48].
Augmented renal clearance can alter the pharmacokinetics and pharmacodynamics (PK/PD) of several antimicrobial agents, mainly ß-lactams [49]. The detailed dosages of antimicrobial agents and the measured creatinine clearance were not collected in the present study. However, TMP-SMX and fluoroquinolones were the most prescribed agents. Their plasma concentrations are not dramatically influenced by augmented renal clearance and not easily monitored in daily practice.

Limitations
Our study differs from previous studies, where patients with and without S. maltophilia infection were compared. We did not consider S. maltophilia colonizations but only HAP, unlike previous studies [4]. In the case of a polymicrobial sample, it is uncertain which bacteria were responsible for the HAP. One may argue that our study suffers from inaccurate diagnoses differentiating between VAP and ventilator-associated tracheobronchitis due to its retrospective design using ICD codes. We acknowledge that the diagnosis method (ETA versus BAL or PSB) may influence the detection of S. maltophilia. This was a pragmatic study that describes different ICU practices, and to date, there is no formal evidence of improved outcomes depending on the diagnosis method used [50]. Despite strict inclusion criteria, and search for consensus in case of debatable case, it is possible that the physician's judgment and diagnosis reported in the medical record were inaccurate. However, 80% of HAP were VAP in our study and excluding nonventilated patients did not alter the observed results.

Conclusions
S. maltophilia HAP had a very low incidence in critically ill patients but was associated with high mortality rate in this large multicenter study. Its onset is hard to predict because of lack of specific risk factors but occurs mainly in long-stay ICU patients. The present study did not provide evidence of a significant effect of delay, duration, or combination of antimicrobial therapy on mortality. Efforts in developing novel and effective approaches for prevention are warranted.

Take-home message
In critically ill patients with Stenotrophomonas maltophilia hospital-acquired pneumonia, delay to appropriate antimicrobial treatment, combination antimicrobial therapy, or prolonged antimicrobial therapy may not be associated with increased survival.
Authors' contributions PG, HB, and AB contributed to the study design, screening for eligibility and inclusion of patients, data analysis, writing of the first draft and of the final version, and review of the manuscript. NM contributed to the screening for eligibility and inclusion of patients, data analysis, writing of the first draft and of the final version, and review of the manuscript. DH contributed to the writing of the first draft and of the final version, data analysis, and review of the manuscript. SL and EN contributed to the screening for eligibility and inclusion of patients, writing of the last draft, and review of the manuscript. MM, JG, CC, GL, PT, CD, AK, TG, JP, and J-ML contributed to the screening for eligibility and inclusion of patients and review of the manuscript. All authors read and approved the final manuscript.

Funding
This study has been solely funded by the Department of Anaesthesiology and Critical Care Medicine of the University Hospital of Nancy, France. The Unit of Methodology, Data and Statistics kindly provided free of charge RedCap support.