Our main findings can be summarized as follows: in our study, the BLT-guided strategy led to: (1) a higher rate of appropriate antimicrobial therapy being administered due to early escalations of empirical beta-lactam prescriptions and (2) a higher rate of optimal antimicrobial therapy being administered due to early de-escalations of beta-lactam prescriptions that were too broad-spectrum.
One of the most challenging problems in clinical practice is that despite the use of updated guidelines, about 15% of ICU patients receive inappropriate EAT [3, 9,10,11]. We observed a similar rate of patients receiving inappropriate EAT in this study. This is of particular concern because inappropriate EAT leads to higher mortality in the ICU . However, an across-the-board prescription of empirical carbapenems in the most severely ill patients is not acceptable, as the link between carbapenem consumption and the emergence of carbapenemase-producing Enterobacteriaceae or multi-drug resistant non-fermenting GNB is by now well-demonstrated [21,22,23,24,25], even with a short exposure . Thus, the World Health Organization  and national plans for controlling antibiotic consumption  have encouraged the development of rapid diagnostic tests evaluating bacterial resistance, which should ideally enable both early escalation in the case of inappropriate therapy and early de-escalation in the case of empirical therapy that is too broad-spectrum. To our knowledge, no test meeting these requirements has yet been validated for GNB in a clinical setting. Taking into account the very good diagnostic performance of the BLT [14,15,16,17, 27], we performed this pilot study and demonstrated that the BLT is a promising approach to reach this dual objective. Indeed, the BLT-guided adaptation of antimicrobial therapy independently increased its appropriateness and optimality to 98% and 95%, respectively, within a median of 30 hours before the availability of antibiotic susceptibility test results. Theoretically, this could have additive beneficial effect. First, by decreasing the duration of inappropriate beta-lactam therapy, the use of the BLT may be associated with improvement in infection cure. Second, by decreasing the patient’s exposure to broad-spectrum beta-lactams, the use of the BLT may be associated with better preservation of intestinal microbiota and lower acquisition of multi-resistant bacteria carriage , which have been suggested to contribute to better outcome . Due to the incidence of these events, demonstration of a BLT-related improvement in outcome would need additional studies including larger samples, especially as mortality may not be influenced by inappropriate EAT in the less severely ill patients  or in some low-risk sites of infection  in which definite antimicrobial therapy is appropriate.
As for other diagnostic tests in the field of infectious diseases, clinical uptake of the BLT should be based on multifaceted evaluation . Although several studies have previously reported the very good intrinsic diagnostic performance of the BLT [14,15,16,17, 27], we reported its added clinical value for the first time, even when empirical beta-lactam antibiotic was chosen according to guidelines. The cost of the procedure also appears as strength since the device is inexpensive (3.5 €) and takes little time to perform (2 minutes for the assay and 15 minutes for the incubation prior to analyze the result). When compared with other biological tests, the BLT appears cost-effective. Indeed, its cost is quite similar to that of the disk-diffusion AST (3 €), and it is much cheaper than molecular detection of antibiotic resistance, such as GeneXpert® methicillin-resistant Staphylococcus aureus detection (18 €). To further illustrate the size of the required investment, in our institution the cost of BLT is amortized after the first and third doses of cefotaxime when de-escalating from carbapenems and piperacillin-tazobactam, respectively (Additional file 5: Figure S1).
Our study may have several limitations. First, the number of patients included in each group may appear relatively small. However, according to the sample size calculation performed prior to the beginning of the study, the numbers were appropriate to demonstrate the superiority of the BLT-guided adaptation strategy in improving the appropriateness and optimality of the antimicrobial therapy once the results of microbiological sample culture are available. We did not observe differences in morbidity and mortality between patients treated according to conventional and BLT-guided strategies, but our study is likely to be underpowered to detect small differences in the incidence of safety parameters on the one hand, and improvement in the outcome of patients with early beta-lactam escalation on the other hand. Second, although the sensitivity of the BLT is above 95% for ESBL detection, lower sensitivity (around 60%) has been reported for detection of AmpC-overproducing Enterobacteriaceae [14, 15]. Indeed, we reported one false negative for a cephalosporinase-overproducing GNB, which led to delayed escalation to Cefepime, similar to what would have occurred without the use of the BLT. Nevertheless, our study demonstrated that this event is much less incidental in clinical practice than the BLT-guided early escalations of inappropriate EAT, suggesting a BLT benefit/risk balance in favour of its use. Finally, although our case-control study is the first to report the potential utility of a rapid phenotypic test detecting GNB resistance to 3GC in clinical settings, definitive evidence will require a large, randomized, controlled trial.
Looking ahead, if the use of the BLT for Enterobacteriaceae colonies saves time in assessing the adequacy of antimicrobial therapy compared to the AST, recent developments could lead to another 24 hours being saved. Indeed, several protocols enable the efficient detection of Enterobacteriaceae that produce ESBL directly from the microbiological samples [17, 27, 31]. As such, it could be hypothesized that the direct use of the BLT on a microbiological sample that is positive for GNB on direct examination could shorten the duration of inappropriate EAT to a couple of hours.