Dear Editor,

The diagnostic approach for ventilator-associated pneumonia (VAP) needs to be improved [1]. Volatile organic compounds (VOCs), produced either by invading respiratory pathogens or the patient’s pulmonary defense system, could serve as early diagnostic markers for VAP [2]. Electronic nose (eNose) technology integratively captures complex VOC mixtures to create a ‘VOC fingerprint’ using an array of semi-selective sensors [3]. We hypothesized that an eNose would be able to discriminate patients with VAP from those without VAP based on analysis of headspace air from tracheal aspirates (TAs).

In a prospective cohort study we collected TAs every third day from 45 intensive care unit (ICU) patients who were ventilated for more than 7 days. Fourteen patients developed VAP, 14 patients had airway colonization but did not develop VAP and 17 patients developed neither VAP nor airway colonization (study methodology and patient characteristics are given in the Electronic Supplementary Material). The eNose was able to accurately discriminate patients with VAP from those without VAP in both a cross-sectional and longitudinal analysis (Fig. 1, upper panels), and the use of a ‘VOC fingerprint’ was found to improve the diagnostic accuracy of the Clinical Pulmonary Infection Score (Fig. 1, lower panels) in this small cohort of patients. Notably, discrimination by the eNose was not affected by airway colonization, and the findings were independent of the number of colony forming units in the TAs.

Fig. 1
figure 1

Discrimination of patients with ventilator-associated pneumonia (VAP) from those without VAP. Left upper panel Receiver operating characteristic curves for the eNose in the cross-sectional analysis, right upper panel slope of the eNose signal over the days preceding the diagnosis, left lower panel clinical pulmonary infection score (CPIS), right lower panel combination of the eNose signal and the CPIS. AUC Area under the curve, CI confidence interval

Our study has several limitations. First, we were not able to identify which VOCs differentiate between patients with VAP and those without VAP. Furthermore, this study was performed in a highly selected cohort of patients, and the sample size was rather small, thereby limiting generalization of our findings. Indeed, the results of our study need to be confirmed in robust and larger studies.

Of interest, the results of our study suggest that the observed changes in VOC-fingerprints are not solely the result of the presence or absence of bacteria in TAs. VOC-fingerprints can change with the bacterial ecology from colonization to infection, which is known to modulate bacterial metabolism [4]. In addition, host responses may have an effect on VOC mixtures [5].

Taken together, these results suggest that volatile biomarkers may be complementary to clinical disease markers for the diagnosis of VAP. However, the exact role of eNose technology in clinical practice with respect to VAP diagnosis is as yet far from certain. One next step could be to determine which VOCs have the highest discriminative power, such as by using gas chromatography and mass spectrometry to customize eNose sensor arrays. Studies using relevant outcome measures are also needed that compare diagnostic strategies which include and do not include VOC analysis.