Of the many therapeutic decisions, physicians have to face in daily ICU practice choosing initial antibiotic therapy in the patient with suspected severe nosocomial sepsis is one of the more challenging. To favourably impact the outcome, antibiotic therapy covering the offending pathogen has to be initiated without delay [1, 2], which implies administration within 24 h of clinical deterioration and within 1 h of septic shock. This therapeutic choice is little supported by the microbiology lab, as microbiological identification and susceptibility testing usually require 48 h. In patients at risk for infection with multidrug resistant (MDR) pathogens, the clinician has to resort to broad-spectrum antimicrobials, which are themselves linked with the emergence of multidrug resistance. In this respect, appropriate empirical antibiotic therapy should have a balanced antimicrobial spectrum that includes the susceptibility of the infectious pathogen, but does not add unnecessary selection pressure. As the prevalence and complexity of MDR patterns steadily rise, finding this balance is increasingly difficult. Therefore the empirical use of broad-spectrum antibiotic drugs in patients at risk for MDR pathogens is advocated [3, 4], meeting the need to restrict antimicrobial selection pressure by promoting subsequent de-escalation guided by culture results.

As an alternative to empirical combination antibiotic therapy, a more focused initial antibiotic selection guided by surveillance cultures (SC) has been reported. This strategy rests upon the observation that invasive disease by typically nosocomial (and potentially MDR) pathogens is preceded by colonization of such easy to sample anatomical sites as the gut and the oropharynx, the lower airways and catheter insertion sites [5]. In order to be of clinical use in directing initial antibiotic therapy, SC must be able to detect this colonization rapidly and with high sensitivity, as false negative results would place the patient at risk for inappropriate therapy. Moreover, a focused antibiotic choice, with limitation of unnecessary broad-spectrum drugs, requires a low number of false positive surveillance results.

In this journal, two publications contribute to the test characteristics of SC as predictors of MDR infectious etiology. Papadomichelakis et al. [6] found SC to have good positive predictive values (67–94%) and negative predictive values (73–100%) for ventilator-associated pneumonia (VAP) caused by Acinetobacter, Pseudomonas or Klebsiella species. For bacteremia caused by these microorganisms, positive predictive values were low (43–54%), while negative predictive values remained high (88–100%). Consequently, surveillance-guided initial antibiotic therapy was appropriate in 91 and 86% of patients with VAP and bacteremia, respectively. However, in patients without detected colonization the rate of appropriate therapy was unacceptably less than one to two. These data are in favour of surveillance-assisted empiric therapy in a setting with a high prevalence of MDR, albeit that the local MDR problem was insufficiently appreciated in the absence of individual surveillance results. The historical cohort study by Jung et al. [7] examined concordance between microbial etiology of VAP and preceding tracheal aspirates. Surveillance results (available in 80% of patients) were concordant with VAP etiology in 72% but discordant in 28% of the cases. The rate of empiric appropriate therapy was 85%, which was significantly higher than what would have been achieved by following the American Thoracic Society guidelines or the locally used, strictly empirical scheme.

The data by Papadomichelakis et al. and Jung et al. add to the number of studies showing good concordance between SC and infectious etiology resulting in benefits in terms of appropriateness of empiric therapy [6–11]. However, previous reports have found more variable predictive values, depending upon surveillance methodology and underlying prevalence of MDR infection [5, 12, 13]. A prerequisite for good positive and negative predictive values seems to be a sampling frequency of at least twice weekly. Although the Jung et al. paper used a once-weekly surveillance, a larger previous report using the same approach provided disappointing results (Table 1). In contrast to sampling frequency, the relative contribution of sampling distinct colonization sites is more unclear. Extending SC to more sites increases sensitivity [6, 10, 14], a high negative predictive value of SC (>90% in [15–18]) increases its antibiotic-saving potential, as it reliably defines a patient category in which extensive broad-spectrum antimicrobial coverage is unnecessary. On the other hand, if the number of false positive surveillance results increases, this gain can be rapidly lost. As yet, the optimal trade-off is to be defined. However, surveillance-assisted empiric therapy has been associated with atleast a theoretical reduction in antibiotic consumption, by comparing it with hypothetical, strictly empirical schemes [10, 11, 14]. Finally, the added value of surveillance-assisted empiric therapy is highly dependent on the local microbial ecology and the patient’s risk profile for MDR infection. As more reports solidify the clinical usefulness of SC, cost remains probably the most important factor prohibiting a general use of systematic SC [10, 11]. Obviously, ICUs with a high prevalence of MDR will benefit the most, as will be patient populations with a high risk for MDR infection, such as patients with a complex history, a prolonged hospital stay and numerous previous antibiotics. To reduce the cost one can consider restricting surveillance to this ‘difficult’ patient category.

Table 1 Overview of studies reporting concordance between surveillance culture results and etiology of nosocomial infections
Full size table

Let us return to daily practice: are systematic SC an option to find the balance between high rates of initial appropriate antibiotic therapy on one hand, and a rationed use of broad-spectrum drugs at the other, and does their contribution justify their cost? Four considerations have to be made: (1) what is the local prevalence and pattern of MDR? (2) How appropriate is current empirical therapy and, if suboptimal, can it easily be increased through adaptation of an empirical scheme? (3) What are the resources available for microbiological work-up and what is the willingness to invest? (4) How urgent is the need to restrict the use of broad-spectrum antibiotics? If one would decide to adopt a systematic surveillance program, careful evaluation of appropriateness and consumption of empiric therapy before and after intervention would significantly add to the discussion about the value of surveillance.