Many diseases leave behind specific metabolites, either through altered enzymatic activity or in the case of bacteria from the pathogenic organisms themselves. Those volatile organic compounds (VOC) can be detected using different techniques such as electronic noses. Our group previously described a VOC-based method for bladder tumour recognition from urine headspace. This study investigated whether urinary tract infection (UTI) can be diagnosed from VOCs in urine headspace in real time which would be much faster than urine culture.
74 urine samples of patients with confirmed urinary tract infection (UTI) or significant bacteriuria (≥ 104 colony forming units (CFU)/ml) as established by urine culture and 55 samples from healthy controls were collected and measured. VOC measurements were carried out with Cyranose 320TM, an electronic nose measuring changes in resistance induced by VOCs competitively binding to an array of 32 polymer sensors. Statistical comparison was performed using principal component analysis and linear discriminant analysis. Further measurements were performed using ion mobility spectrometry (IMS) which allows to discriminate individual components of the gas samples. Furthermore, from 28 of the UTI samples, corresponding thioglycolate broth cultures at 105 CFU/ml as established by McFarland method underwent measurement with Cyranose.
The isolated pathogens were representative bacterial strains for UTI, predominantly E. coli but also Klebsiella, Enterobacteriaceae, Pseudomonas, and Proteus. Urine samples with confirmed UTI at different corresponding concentrations in culture were compared to healthy controls. At 104 CFU/ml, UTI was detected with Cyranose with a sensitivity of 93% and specificity of 86%. At 105, sensitivity was 95% and specificity 70%; at 106 sensitivity was 95% and specificity 86%. Those numbers were comparable regardless of the bacterial strain in question. Broth cultures were compared to uninoculated thioglycolate broth with a pooled sensitivity of 85% and specificity of 100%. Based on IMS measurements, four peaks specific to UTI were identified and a decision tree was generated which resulted in a highly significant separation between UTI and control (p < 0.002).
VOC-based detection of bacterial infections is feasible in principle. While the first results from urine samples seem promising in terms of UTI detection, the methodology needs refinement in order to increase sensitivity and specificity during rapid testing. The more elaborate IMS seems to be more reliable. The individual relevant metabolites identified by IMS should further be characterised using gas chromatography/mass spectrometry to facilitate targeted rapid testing.