Apparently exhaled breath analysis can be used to detect a number of pulmonary and systemic diseases. As a medical student I was taught  about some of the classical clinical signs of certain diseases (such as diabetic ketoacidosis) which make the patient’s breath have a characteristic smell. I never thought I was very good at detecting those smells. So if you could use a machine to smell things for you, and be objective about their analysis, maybe that would be a good idea? You could detect volatile organic compounds from the patient’s breath and make non-invasive diagnoses.

There are now several electronic noses available which can reportedly provide a ‘breath print’ of exhaled compounds. Their usefulness in clinical practice is still being debated, but this new study (Rogosch T, Herrmann N, Maier RF, Domann E, Hattesohl A, Koczulla AR, et al. Detection of Bloodstream Infections and Prediction of Bronchopulmonary Dysplasia in Preterm Neonates with an Electronic Nose. The Journal of pediatrics. 2014) is probably the first in newborn infants. Because of the difficulty of obtaining breath samples the authors took endotracheal secretions from intubated preterm babies, and then held the ‘snout’ of the ‘e-nose’ over the secretions to get a breath print.

I will have to get quite sniffy about this article however; the authors report the results of 38 samples from 28 babies of whom 8 had a blood stream infection. They never report the timing of the infection, or how it related to the timing of the sampling, so we get no idea of whether the aspirates were taken from babies already being treated for infection, or during a sepsis evaluation, or even afterwards, or days before. The authors show different breath prints from babies with infection and those without. They also analyze the first aspirate from the same babies, at least the 23 of them who were under 32 weeks gestation, and show some differences in breath prints from babies who went on to develop BPD and those who did not. They don’t say which of these babies had infections and which did not.

An interesting idea, but the way the data are presented makes it impossible to tell whether this might be clinically useful, or not.