More research is required to understand factors influencing antibiotic prescribing in complex conditions like suspected ventilator-associated pneumonia.

We read with interest the commentary by Yoshimura et al. in relation to our recently published trial investigating whether a novel, rapid biomarker-based rule-out of ventilator-associated pneumonia (VAP) could improve antibiotic stewardship (1).

Antimicrobial resistance is a growing global crisis and strategies that aim to improve antibiotic stewardship are urgently needed (2). Critically ill patients are at high-risk of developing healthcare-associated infections and receive a high burden of antibiotics (3,4). The diagnosis of VAP is challenging as clinical and radiological features are non-specific, with 30–60% of patients with suspected VAP ultimately having infection confirmed (5-8). Respiratory sampling techniques can have a significant impact on the VAP rate observed, with a diagnostic approach based on endotracheal aspirate seeming to provide far higher false positive rates than an approach based on invasive bronchoscoping sampling (9). Moreover, at least in some important studies, a bronchoscopic approach has been associated with less antibiotic use, less antimicrobial resistance, and no signal of harm (10,11).

Our work aimed to address the overuse of antibiotics for VAP through the development of novel diagnostics. Our trial followed previous single-centre derivation and multi-centre validation studies (12,13). Yoshimura et al. highlight three potential areas that might explain the similar antibiotic use observed in the two groups we studied (one having a biomarker-based recommendation on antibiotic use, the other routine care). These centred on the quality of bronchoalveolar lavage (BAL), the low proportion of patients with low biomarker levels in the intervention group, and the low compliance with a recommendation to stop antibiotics when biomarkers confidently suggested VAP was absent. These will be considered in turn.

The assay failures in our trial, pointed out by Yoshimura et al., actually related to use of the diagnostic assay. This was due to technical issues as a research assay was adopted into hospital laboratories. Although there is considerable variation in the use of bronchoscopy for the diagnosis of VAP in the UK (14), bronchoscopy is still a common procedure in intensive care units. In our study BAL was performed to a standard operating procedure, often by experienced bronchoscopists. Consistent performance of the biomarker-test across our studies suggests to us that BAL was performed to a consistently high standard.

Yoshimura and colleagues are correct in saying that, in our intervention group, there were fewer “low biomarker” results than anticipated. The trial was powered based on a modelled change in the frequency distribution of antibiotic-free days in the 7 days following BAL (15), and not on the diagnostic performance of the biomarker or the expected number of patients with a rule-out test. The trial did retain adequate statistical power to detect a difference in antibiotic use, had there been adequate compliance with biomarker-guided recommendations.

While we believe the technical issues discussed above did not significantly influence the result of our trial, we completely agree with the authors that complex clinical and behavioural determinants were responsible for non-compliance with biomarker-guided recommendations, and ultimately for the similar, high use of antibiotics in each group. Our trial incorporated a process evaluation that captured some of this complexity. The study points to an important disconnect between clinicians’ stated intentions and their prescribing practice which, in our experience, was sufficient to obviate the potential benefits of a good diagnostic test. Our study suggests that considerable work needs to be done to understand these complex behavioural issues and prescribing influences, if diagnostic tests are to have a greater chance of influencing outcomes in conditions like suspected VAP.

Yoshimura and colleagues are conducting an important trial seeking to optimise antibiotic use through Gram stain of respiratory samples. Data suggest that the use of endotracheal aspirates (rather than BAL) may actually result in more antibiotic use (9). While Gram stain may result in a narrower spectrum of antibiotics, it will be very interesting to see whether the total number of days of antibiotics can be reduced. If there is room to add a process evaluation within the time-frame of their trial, our experience suggests it may yield valuable insights. We wish the authors every success with their trial, we very much hope it can deliver benefits to patients with VAP, and we look forward to the results.

The article’s supplementary files as