Date Published: April 15, 2019
Publisher: Public Library of Science
Author(s): Mitesh Patel, Dawn Fowler, Jeremy Sizer, Christopher Walton, Yung-Fu Chang.
Care of patients with potential CDI can involve isolation and use of antibiotics, often before a definitive diagnosis is available, impacting healthcare resource and contributing to antibiotic resistance. There is anecdotal evidence that the faeces of CDI patients have a distinctive odour, while it is well-established that changes in the gut microbiota are associated with changes in the volatile organic compounds (VOC) produced. A total of twenty-four candidate volatile biomarkers were identified from a review of the literature including in vitro, animal and human studies. Using thermal desorption-gas chromatography-time-of flight mass spectrometry (TD-GC-ToFMS), VOC emission rates were determined on stored frozen stool samples from 53 CDI-positive and 53 CDI-negative patients with unexplained diarrhoea which had previously been diagnosed using enzymatic and nucleic acid amplification tests. Sample preparation was limited to placement of a subsample in an appropriate container. Compounds exhibiting a statistically significant difference (p < 0.05) in emission rate between the CDI-positive and–negative groups and a corresponding area under the receiver-operator characteristic curve (ROC) >0.7 were considered potentially indicative of CDI. Seven compounds were so identified: propan-1-ol (ROC 0.75), 3-methylbutanal (ROC 0.84), ethyl propionate (ROC 0.81), hexanoic acid (ROC 0.73), 4-methylphenol (ROC 0.81), dodecane (ROC 0.80) and indole (ROC 0.85). A number of potential volatile biomarkers of CDI can be sampled rapidly and with little prior preparation from faecal samples of patients with diarrhoea. Of these 4-methylphenol (p-cresol) is of particular interest as it has been anecdotally linked to CDI and is closely related to the biology and virulence of Clostridium difficile. This approach shows promise for the rapid, point-of-care diagnosis of CDI with good sensitivity and specificity.
Clostridium difficile infection (CDI) can produce symptoms ranging from mild to severe diarrhoea to pseudomembranous colitis. Prior exposure to antibiotics is a major predisposing factor and those unable to mount a strong immune response, particularly the elderly, are most vulnerable. Severe cases can be fatal. Although considered to be usually a nosocomial infection, there is evidence that CDI is becoming more prevalent in communal settings such as prisons and care homes . CDI is associated with increased length of hospital stay and significant mortality with a concomitant drain on healthcare resources. There is limited information on the economic impact of CDI, though significant attributable costs have been estimated for both Europe and the US, with higher costs associated with recurrent CDI [2, 3].
Data analysis was carried out using SPSS (version 22, IBM UK Ltd, Portsmouth, UK). Data were expressed for each compound as rate of release per unit volume of the sample (ng.l-1.min-1). Median, upper quartile and lower quartile emission rates for C difficile positive and negative samples are given in Table 1. Emission rates for most compounds examined were found to follow a non-normal frequency distribution and moreover a number of non-detects were recorded, therefore differences between emission rates from positive and negative samples were examined using the non-parametric Mann-Whitney U test for unpaired samples. Non-detects were treated as zeros and were included in the analysis. Percentages of non-detects for each compound are indicated in Table 1. Receiver operating characteristic (ROC) curves were calculated for each compound and the areas under the curves (AUC) are shown in Table 1 with their 95% confidence bounds.
In this study, we have investigated the potential application of faecal VOC analysis for the rapid diagnosis of CDI. In order to identify potential volatile biomarkers for CDI we adopted a targeted approach focussing on twenty-four compounds which had been previously reported as being associated with CDI, produced by Clostridium difficile in vitro, or which appeared discriminatory from an initial visual inspection of the TD-GC-MS data. For convenience these compounds can be considered as falling into several broad groups: short-chain fatty acids (SCFAs) and their derivatives, ketones, aromatics and alkanes. Although the overall constitution of VOCs in the gut is influenced by diet and host metabolic processes, there is for the majority of these compounds a plausible pathway for their production by the gut microbiota. SCFAs are produced by bacterial fermentation of dietary polysaccharides  while it has been shown that these in turn can be metabolised to their homologous branched derivatives, aldehydes and esters by cultured human stool samples . Ketones and alkanes are probably produced through lipolysis, while aromatics originate from the catabolism of aromatic amino acids . Applying a set of empirical criteria to our data, we identified seven potential biomarkers for CDI from an initial candidate list of twenty-four.
We investigated whether rapidly-accessible volatile biomarkers of CDI exist in routine stool samples from patients with unexplained diarrhoea. We employed a novel sampling method which allows the rate of release (rather than concentration or abundance) of VOCs to be quantified. This method requires no sample preparation beyond placing it in an appropriate container. A targeted approach was adopted based on published data on the biology and VOC profile of C. difficile which was successful in identifying seven potential volatile biomarkers for CDI. Of these, 4-methylphenol (p-cresol) (AUC ROC curve 0.81) is considered to have considerable promise as it relates closely to the metabolism of C. difficile and its rate of production may be elevated in hypervirulent strains. Our results show that a volume of faecal headspace gas sampled over a few minutes may contain VOCs indicative of CDI. This approach could be applied to point-of-care diagnosis, though the realisation of a practical method will depend on the development of instrumentation to detect the relevant biomarkers with sufficient rapidity and analytical specificity in the complex mixture of VOCs released by faecal samples.