The appropriateness of empirical antimicrobial therapy for pneumonia is a critical issue in ICU. Current guidelines suggest to use local epidemiological data and individual patient’s risk factors to guide probabilistic antimicrobial therapy [11]. This may lead to the overuse of broad-spectrum antimicrobial therapy, particularly carbapenems. In the present study, we hypothesized that a systematic detection of ESBL-E digestive colonization may help to limit the use of carbapenems despite the high incidence of MDR pathogens risk factors in ICU. It is presumed that bacterial flora changes during ICU stay with colonization of the upper airway by digestive flora [24]. The performance of rectal swab to predict EBSL-E presence or absence in respiratory samples was then investigated in early (≤5 days) and late (>5 days) period after ICU admission. Moreover, this approach could be pragmatic taking into account the evolution of inflammatory patterns along time with an initial intense inflammatory response that may result in organ dysfunction and early death, followed by a later phase characterized by a post-aggressive immunosuppression [25]. Medical and surgical ICU patients were investigated, insuring good external validity. The main results were: (1) Medical and surgical patients had similar prevalence of EBSL-E in rectal swab and in respiratory samples, despite very different clinical characteristics; (2) the early and late groups showed very different prevalence of EBSL-E in rectal swabs and respiratory samples; (3) when rectal swabs were negative, the risk of having ESBL-E in respiratory samples was very low for both early and late groups.
The prevalence of positive rectal swab was similar to the previously reported data, reaching 15.0% in rectal swabs performed before day 5 [10] and 36.8% in rectal swabs performed after day 5. This prevalence was higher than the one found by Bruyère et al. (6.8%) in a study conducted in France between 2006 and 2013 [21].
Early after ICU admission (≤5 days), when the rectal swab was negative for ESBL-E, respiratory samples were also negative for ESBL-E in 99.2% of cases. This may help reduce the prescription of carbapenems when pneumonia is suspected. When the rectal swab was positive for ESBL-E, only 14.5% of respiratory samples were positive for ESBL-E. As the patient may develop a pro-inflammatory response secondarily to pneumonia at the early phase of sepsis, the risk would be too high not to use carbapenems when the rectal swab is positive for ESBL-E and the clinical condition is severe. These patients might present a particular condition such as long-term hospitalization or iterative use of antimicrobial drugs selecting ESBL-E in their digestive flora. It would be interesting to identify risk factors of having a respiratory sample positive for ESBL-E when rectal swab is positive for ESBL-E in a patient cohort with pneumonia.
Late rectal swabs, performed after 5 days of ICU admission, showed a higher incidence of EBSL-E (36.8%) than early rectal swabs (15.0%). Among these samples, interestingly the NPV remained very good (93.4%) despite the presence of several MDR pathogens risk factors since these patients are hospitalized in ICU for a median of 13 days with a likely previous antimicrobial therapy. This may also help reduce the prescription of carbapenems when pneumonia is suspected. When the rectal swab was positive for ESBL-E, 34.4% of respiratory samples were also positive for ESBL-E, a relatively high incidence. Since these samples were related to the late phase of ICU stay, a sensible solution would be to wait for the microbiological results before antimicrobial therapy initiation when the clinical condition is not life threatening. As a consequence, overuse of carbapenems may be avoided in the late phase, with minimal individual risk and better control of MDR pathogens selection risk.
Relying on our results and on the American thoracic society guidelines [11], we suggest a decision tree for empirical antimicrobial therapy in patients with respiratory tract specimen positive culture and suspicion of pneumonia (Fig. 3).
One can argue that the very high NPV found in early (99.2% [98.7–99.6]) and late (93.4% [91.9–95.0]) groups were due to the low prevalence of ESBL-E in respiratory samples. However, in this study, the overall prevalence of ESBL-E in early and late respiratory samples was 8.1% (203/2498), which was higher than the prevalence described in the French ICU nosocomial infection surveillance network in 2013 (6.4%) and 2014 (4.9%) [26, 27]. Consequently, our results can be generalized to ICUs with roughly the same ESBL-E prevalence in respiratory tract specimens’ cultures.
The link between rectal swabs and respiratory samples positive for EBSL-E is based on a debatable hypothesis, suggesting a contamination from digestive flora to the respiratory tract [24]. The evolution of ESBL-E positive samples proportion versus duration of ICU stay revealed surprising results. From day 0 to day 20, the incidence of ESBL-E positive rectal swabs remained less than 35%, whereas the incidence of ESBL-E positive respiratory samples was increasing along time in a linear trend. This supports the idea of an increased respiratory colonization by EBSL-E from digestive flora. One can then hypothesize that the delay between the first ESBL-E positive rectal swab and the respiratory sample might guide the decision to use carbapenems. If the first ESBL-E positive rectal swab is early during the hospitalization, there might be a high risk to observe an EBSL-E positive respiratory sample. On the contrary, if rectal swabs were negative for several days with a recent ESBL-E positive rectal swab, there might be a low risk of ESBL-E in the respiratory sample.
However, there are some limitations to our study. As high as 27% of the culture-positive samples have been excluded due to the absence of rectal swab. Moreover, the weight of patients who had multiple respiratory sampling may have influenced the results. The evidence of a clinical benefit of ESBL-E digestive colonization surveillance to predict the presence or the absence of ESBL-E in respiratory samples needs to be confirmed in a clinical study involving pneumonia and not only respiratory samples. As we chose to focus on respiratory samples and not episodes of pneumonia, we did not study antimicrobial therapy regimens. We chose to focus on ESBL-E presence or absence in respiratory samples and in previous rectal swabs, regardless of the Enterobacteriaceae species. Indeed, the association of ESBL-E presence or absence in rectal swab and respiratory sample is relevant for the choice of empirical antimicrobial therapy, whatever the Enterobacteriaceae species. In addition, ESBL-E are transmitted through plasmids from an Enterobacteriaceae strain to another, making the concordance interpretation between respiratory and rectal ESBL-E difficult.