Are third-generation cephalosporins unavoidable for empirical therapy of community-acquired pneumonia in adult patients who require ICU admission? A retrospective study

Community-acquired pneumonia (CAP) is a major cause of hospitalization and death worldwide [1, 2]. CAP is the leading infectious cause for admission in the intensive care unit (ICU) and one of the most common causes of sepsis, severe sepsis and septic shock [3, 4]. Guidelines published in 2006 by the French Intensive Care Society (SRLF) and the French Infectious Diseases Society (SPILF) [5] recommended a non-antipseudomonal third-generation cephalosporin (3GC, as ceftriaxone or cefotaxime) in combination with either a macrolide or a fluoroquinolone for empirical treatment of patients with severe CAP who require ICU admission. Since the publication of those recommendations, a sharp increase in 3GCs prescription has been observed, with a 58% increase in ceftriaxone use in France between 2008 and 2013 (French National Institute for Public Health Surveillance, RAISIN) [6]. Between 2003 and 2012, the incidence of third-generation cephalosporin-resistant strains, especially extended-spectrum beta-lactamases-producing Enterobacteriaceae (ESBLEs), increased from 0.17 to 0.48 per 1000 hospitalization days [7].

Furthermore, the increased incidence of ESBLEs promotes the use of carbapenems, which in turn could increase the risk of emergence and spreading of carbapenem-resistant strains. A restricted use of 3GCs could therefore be warranted [9], as it might provide a possible strategy to decrease incidence of 3GCs’ resistance [10]. Thus, the aim of our study was to assess whether the use of 3GCs in patients with CAP who require ICU admission was justified based on the microbiologically identified causative pathogen and its susceptibility pattern.

The study was conducted in a 670-bed tertiary teaching hospital in Paris, France. We retrospectively reviewed medical records of all consecutive adult patients who required ICU admission between January 2007 and March 2014 for acute respiratory failure. Pneumonia was diagnosed with the following criteria: a new alveolar, interstitial or alveolo-interstitial opacity on chest radiography, associated with at least two of the following: cough, sputum production, temperature above 38 °C or below 35 °C, and auscultatory findings consistent with pneumonia or dyspnoea. To be considered a community-acquired pneumonia, symptoms had to develop in the community or within the first 48 h after hospital admission [11–13]. Patients were admitted to the ICU from the emergency department or directly from the pre-hospital emergency medical team. The institutional review board of our institution approved the study.

For each patient included in the study, the following parameters were recorded: age, sex, underlying comorbidities (chronic lung disease, hypertension, diabetes, chronic liver disease, heart failure, chronic renal disease, HIV infection, ongoing solid cancer or blood cancer), SAPS II score, mechanical ventilation use, intravenous vasopressors administered for septic shock and ICU mortality.

A pathogen was considered causal for CAP after a comprehensive and critical appraisal of clinical and microbiological data. The microbiological diagnostic workup for all patients included: blood cultures obtained upon admission and before antibiotic administration, expectorated sputum sample in patients able to cough or blind endotracheal aspirations in patients mechanically ventilated. A broncho-alveolar lavage through fibroscopy was performed if others means were deemed unsuccessful. Respiratory samples were analysed for Gram staining initially and after 48 h of culture. According to international guidelines, immunochromatographic tests for detection of Streptococcus pneumoniae and Legionella pneumophila serogroup 1 antigen in urine were also performed [5]. The addition of PCR virus detection on lower respiratory samples or nasal–pharyngeal swabs (e.g. specific influenza PCR testing during the flu-season or respiratory virus PCR panel), or direct detection and quantitative PCR assessment of Pneumocystis jirovecii on broncho-alveolar fluid in patients with underlying immunosuppression was left at the discretion of the attending intensivist.

As per French guidelines, at our institution the combination of ceftriaxone or cefotaxime with either a macrolide or a fluoroquinolone is used as empirical treatment of patients with severe CAP requiring ICU admission. Then, as suggested by internal protocols, empiric antimicrobial therapy would be reassessed after 48 h by attending physicians and adjusted according to the microbiological cultures, if available. The most narrow-spectrum antibiotic would be used as the definitive therapy based on antimicrobial susceptibility results based on the antibiogram diameter inhibition and switched to the oral route as soon as possible. If pneumococcal pneumonia were documented, the association of a beta-lactam and a macrolide would be maintained for the first 3 days [14] and then a beta-lactam (usually amoxicillin) would be maintained alone thereafter. If all cultures would remain negative in a patient having received any antibiotic before microbiological samples, the empiric therapy would be switched to oral pristinamycin. However, special attention would be paid to the direct examination of the sputum specimen in those patients, as the presence of typical Gram stain aspects (e.g. Gram-positive diplococci or numerous Gram-negative bacilli) might modify the choice of definitive therapy. In general, fluoroquinolones use was discouraged except in case of confirmed severe legionellosis [15].

In our retrospective study, we evaluated the efficacy and relevance of the chosen empirical antibiotic therapy especially non-antipseudomonal 3GCs in the light of the microbiologically identified pathogens and their susceptibility result. To assess whether 3GCs were really justified, we focused on CAP with microbiological documentation, which required definitive beta-lactam therapy. These infections were separated into three groups according to whether 3GCs appeared unjustified (pathogen susceptible either to amoxicillin or amoxicillin/clavulanate), appropriate (pathogen cefotaxime susceptible and resistant to amoxicillin/clavulanate) or insufficient (pathogen resistant to both amoxicillin/clavulanate and non-antipseudomonal 3GCs and requiring a broader-spectrum beta-lactam).

In the present study, 3GCs were the second most frequent empirical antibiotic used for severe CAP after macrolides (53 and 70%, respectively). This is in agreement with French guidelines [5]. Streptococcus pneumoniae and Haemophilus influenzae remain, as previously observed [18–22], the most frequent bacterial species causing CAP. These two pathogens accounted for 60% cases of documented pneumonias. These pathogens are most often highly susceptible to narrow-spectrum β-lactams such as amoxicillin or amoxicillin/clavulanate [20], except for some H. influenzae isolates. In the present studies, 10 H influenzae strains (21%) were amoxicillin/clavulanate resistant, but all were susceptible to both cefotaxime and spiramycin.

The group of unjustified third-generation cephalosporin prescription is the most important with 66% (159) of patients with documented infection; in this group, treatment with amoxicillin or amoxicillin/clavulanate would have been sufficient. Appropriate 3GC’s prescription was observed in only 18% (43 patients) of our cohort. Finally, the group in which third-generation cephalosporin use appeared insufficient represents 16% (39) of patients, meaning that in these patients a beta-lactam with a broader spectrum than 3GCs would have been required.

These results suggest that the use of third-generation cephalosporins as an empiric beta-lactam for treatment of severe community-acquired pneumonia admitted to the ICU is a perfectly tailored antibiotic therapy in only one out of five patients. Most of the time, a narrower-spectrum beta-lactam might be preferred. The amoxicillin/clavulanate combination would be a satisfactory alternative to 3GCs against two-thirds of overall recovered pathogens. Moreover, non-antipseudomonal 3GC spectrum appears insufficient in one out of six patients with CAP requiring beta-lactam.

In a second part of the study, we sought to delineate patients for whom amoxicillin/beta-lactamase inhibitor would be insufficient and who would require a broader-spectrum beta-lactam. For that, we assessed the presence or absence of risk factors associated with colonization/infection with resistant strains previously reported [21]. We found that chronic lung disease, ongoing cancer, non-ambulatory status and previous antibiotic therapy were significantly associated with amoxicillin/clavulanate-resistant strains.

These amoxicillin/clavulanate-resistant isolates (mainly Pseudomonas aeruginosa, non-fermenting Gram-negative bacilli and several Enterobacteriaceae) were most of the time sensitive to amikacin (94%), suggesting that this antibiotic could be considered in patients with risk factors for infection with an amoxicillin/clavulanate non-susceptible pathogen.

Prior use of 3CGs has been identified as a risk factor for infections caused by ESBL-producing Enterobacteriaceae [8, 23, 24]. Recent experiments report a decrease in ESBLEs emergence following decrease in cephalosporins use [10]. These data support, when possible, a decrease in the use of 3CGs to control ESBL-producing Enterobacteriaceae spreading. According to our results, amoxicillin/clavulanate could be proposed as an alternative beta-lactam for severe CAP in patients without risk factors for amoxicillin/clavulanate resistance. This situation (absence of A/C resistance risk factor) is observed in 55% of patients.

Such a strategy could contribute to limit the 3GCs use and then probably the emergence of extended-spectrum beta-lactamases (ESBLs). It is important to note that conversely to French recommendations, amoxicillin/beta-lactamase inhibitor is proposed as a possible first-line empirical beta-lactam for ICU patients with CAP in North American [11] and British [25] guidelines. Reducing the use of third-generation cephalosporins for CAP is already suggested in the emergency department [26].

Among patients with risk factors for amoxicillin/clavulanate resistance, strains encountered are very similar to those observed in patients with healthcare-associated pneumonia, and even this is still debated, to strains identified in patients with ventilator-associated pneumonia and/or hospital-acquired pneumonia [16–27]. The latter patients should receive at least 3GCs for empirical treatment and antipseudomonal beta-lactam (piperacillin/tazobactam, cefepime, ceftazidime) or carbapenem associated with an aminoglycoside for the most severe patients. Amikacin appears in the present study more frequently efficient than gentamicin.

The present study has, however, several limitations. First, whereas some of the observed data are similar to previously published results, the present study is a single-centre retrospective study with a moderate size population. Thus, these results should be interpreted cautiously and confirmed in a larger population. Second, our approach relies on an extensive infectious workup including invasive procedures, which is a strategy not supported by current guidelines in non-immunocompromised patients. Nevertheless, owing to significant increase in antibiotic resistance and to the necessity of a relevant antimicrobial stewardship, an aggressive diagnostic approach dedicated to increase the rate of microbiological identification in order to use the antibiotic therapy with the narrowest spectrum seems mandatory.

In ICU patients with documented community-acquired pneumonia who require definitively a beta-lactam, we observed that 3GCs are most of the time an unjustified or insufficient antibiotic therapy. Based on the present retrospective analysis, a narrower-spectrum beta-lactam antibiotic with amoxicillin/clavulanate combined with a macrolide could replace 3GCs as empirical antibiotics for severe CAP in patients without risk factors for amoxicillin/clavulanate resistance, i.e. chronic lung disease, ongoing cancer, non-ambulatory status and antibiotic therapy during the previous 90 days. In patients with risk factors for amoxicillin/clavulanate resistance, 3GCs, antipseudomonal beta-lactam including carbapenem associated with amikacin in the most severe patients seems a relevant empirical antibiotic therapy. This strategy could promote a decrease in 3GCs’ consumption.