Breath-based Diagnosis of Infectious Diseases: a Review of the Current Landscape

1. Introduction:

Breath has been proposed to hold diagnostic clues to pathophysiologic processes since ancient times. Ancient Greek physicians described a sweet smell in patients with diabetes and a fish-like odor emanating from patients with what we now know to be kidney disease¹.

The modern era of breath analysis started when biochemist Linus Pauling characterized the landscape of volatile organic compounds (VOCs) in human breath using methods available to him at the time, finding these VOCs to originate from numerous endogenous biochemical processes, including aldehydes and alkanes from lipid oxidation and ketones from carbohydrate and fatty acid metabolism^2–5. Gas-phase metabolites and breakdown products originating from these processes are carried through the circulatory system and rapidly excreted through the lungs.

These VOCs are diluted in a bulk matrix of nitrogen, oxygen, carbon dioxide, water vapor, and inert gases, and typically present in trace concentrations in the breath. While a few highly abundant compounds, like acetone, are present in the low parts-per-million range, most VOCs are present in the low parts-per-billion (ppb) to parts-per-trillion (ppt) range. The low concentration of these analytes is one of the challenges in reliably identifying metabolic changes in the breath indicative of disease processes, with the need for methods to concentrate these metabolites prior to analysis on most analytical instrumentation. Other challenges include the natural variability of breath compounds diurnally and with varying age and sex, the potential for exogenous VOCs from the patient’s environment in the breath sample, the chemical complexity of breath samples, which can contain hundreds to a thousand VOCs, and the lack of standardized methods for breath collection, which in aggregate may lead to findings that are not generalizable⁶.

Furthermore, the term “breath analysis” is primarily used when referring to the analysis of VOCs, which are truly volatile or semivolatile and found in the gas phase of the breath sample, but it is also sometimes used to refer to the analysis of larger molecules, including nucleic acids and proteins, suspended in aerosols and droplets in the liquid component of the breath. To analyze this fraction of the breath, the breath is generally collected in a cooled container and the condensate examined for these analytes.

With the development of effective preconcentration methods and increasingly sensitive analytical instruments capable of identifying and quantifying analytes at ultra-trace levels, various analytes in the exhaled breath have been assessed for the identification of pathophysiological processes, including infectious diseases. The ultimate objective of identifying and validating breath biomarker of various infections is to identify these processes at an earlier stage than currently possible with existing culture, antigen, and molecular methods, to facilitate early, appropriate antimicrobial prescribing, reduce unnecessary antimicrobial exposure, and improve clinical outcomes in patients with these infections.

2. Breath matrices:

When searching for breath biomarkers in disease states, the breath matrix chosen for sampling plays a vital role in determining the type of information that can be obtained. Breath matrices are classified into two main phases: gaseous/volatile breath and breath condensate⁷. Most studies to date have been based on volatile breath, which contains low molecular weight volatile and semivolatile compounds⁸. VOCs in breath can originate from different sources, and their concentrations in the breath are due to both endogenous (normal human metabolism, products of microbial metabolism in the oral cavity, lungs, or digestive system) and exogenous (environmental exposure occurring through inhalation, ingestion, or skin absorption of these VOCs) factors. Exhaled breath components originate either from alveolar breath or from dead space air. Alveolar breath comes from gas exchange in the lungs, whereas the sources of dead space air, the volume of air which is inhaled but does not take part in gas exchange, include the mouth, trachea, and bronchi.

3. Instruments for breath analysis:

3.1. Gas Chromatography (GC)

After preconcentration of breath VOCs using sorbent materials, the most common technique utilized for chemical separation, detection, and measurement of these VOCs is gas chromatography coupled with a detector, such as a mass spectrometer (MS) or flame ion detector (FID)⁹. After desorption of samples, the analytes are carried through a chromatographic column by an inert gas and separated according to the chemical properties of the column, which can have a polar, mid-polar, or non-polar coating. Polar columns separate analytes based on their polarity after their chemical interaction with the column coating, while non-polar columns separate analytes based on their boiling points.

3.2. Gas Chromatography-Flamed Ionization Detector (GC-FID)

Gas chromatography can also be paired with a flame ionization detector (FID), which has the advantages of being significantly lower in cost than a mass spectrometer but still able to quantify known analytes accurately, with several studies using this method to measure hydrocarbons and other VOCs in human breath^10,11. In FID, inert carrier gas transfers analytes from the GC column to the FID detector, where volatile organic analytes in the breath sample undergo combustion in the presence of hydrogen and zero air to produce ions and free electrons, which eventually provide the information about the targeted analytes. GC-FID can be more sensitive for quantitative analysis than GC—MS, although GC-FID cannot be used to identify biologic molecules or provide qualitative data.

3.3. Gas Chromatography-Mass Spectrometry (GC-MS)

GC-MS is widely used in breath research, especially for identification and interpretation of VOCs present in human breath. Sample preparation is an integral part of this technique where breath compounds are preconcentrated in a suitable device prior to analysis by GC-MS. Solid-phase microextraction (SPME) is one the most popular techniques for headspace and solvent-free extraction, using needle-like fibers with a sorbent coating¹². However, SPME is limited to the extraction of hydrophilic compounds and is sometimes biased towards hydrophobic compounds in the breath. To overcome these limitations, sample preparation through sorbent trap devices has become a more widely used preconcentration method for VOCs.^13,14 GC-MS is based on the measurement of the mass-to-charge ratio (m/z) after analytes are fragmented by electron or chemical ionization, and the daughter ions are monitored for interpretation.^15,16,17 Large reference libraries, selection of pre-determined run methods, and high sensitivity (in the ppt level when used in combination with preconcentration methods) for detection and quantification of breath compounds are the major advantages of this technique. However, the cost of GC-MS instruments and the need for regular operation and maintenance by highly skilled personnel have limited its widespread use in regular clinical settings.

3.4. Gas Chromatography Time-of-Flight Mass Spectrometry (GC-TOF-MS)

For quantitative analysis of breath compounds, GC-TOF-MS has gained considerable popularity. In GC-TOF, ions are pulsed down a field-free flight tube, with smaller ions traveling faster than larger ions, with separation of analytes as they are registered on the detector. The primary advantage of GC-TOF is the ability to separate analytes quickly and also its high-resolution capability. Unlike scanning instruments, TOF-MS can acquire the chromatogram at a microsecond scale depending on the acquisition potential. GC-TOF-MS has been applied successfully to the detection of trace VOCs in the breath.^18,19,20

3.5. Proton Transfer Reaction (PTR) and Selected Ion Flow Tube (SIFT) Mass Spectrometry

Proton Transfer Reaction Mass Spectrometry (PTR-MS) and Ion Flow Tube Mass Spectrometry are potential platforms for on-line quantification of VOCs without any preconcentration. Both instruments rely on chemical ionization of the analytes. PTR-MS utilizes a soft ionization technique, where a specific reactant (generally H₃O⁺) interacts through proton transfer with all analytes having an affinity for protons higher than water molecules^21,22. After ionization, the molecules are accelerated through a drift tube after the application of a magnetic or electric field and passed to MS for analysis. In SIFT-MS, ions are passed through a flow tube by a carrier gas in the presence of a small voltage to avoid possible diffusion to the walls of the tube^23,24. Both instruments consist of a ion-generation zone, a reaction zone, and a detection zone²⁵. The primary advantages of PTR-MS and SIFT-MS are the feasibility of fast analysis and high sensitivity, with the potential to measure VOCs from ppb to ppt levels²⁶. However, both instruments are expensive and require regular maintenance by expert personnel.

3.6. Ion Mobility Spectrometry/Differential Mobility Spectrometry (IMS/DMS)

IMS/DMS are fast and sensitive analytical tools for the detection of a wide range of gas-phase molecules. These techniques were first used for the detection of chemical warfare agents in the atmosphere²⁷. IMS/DMS can detect trace analytes in the ppb to ppt range. In IMS, ions are separated at ambient pressure in the presence of an external electric field^28,29. The ions are transported to an electrode by a carrier gas and the drift time for ion swarm towards the electrode is dependent on the mobility of ions.³⁰ An electrical field is applied to the ions, and the ions reach the electrode at different times. IMS and DMS have the potential to provide online, rapid analysis of breath analytes. However, preconcentration is generally necessary, along with prior knowledge of target analytes to determine analyte levels.

3.7. Electronic nose sensor

The electronic nose (eNose) is a relatively simple device that can potentially be applied to breath analysis³¹. The eNose contains a sensor array that undergoes a physical change in response to a chemical input, and these changes are converted into electric signals, with a particular response to a chemical analyte or mixture of chemicals^32,33. The signal produced by the eNose is dependent on the nature of the odor, the interaction of the chemical or chemical mixture with the sensors, and the type of sensors in the array.³⁴ Unlike laboratory-based benchtop instruments, eNoses are generally portable and do not require expensive components or skilled operators, with integrated neural network analyses being applied to discriminate the eNose signal from disease vs. non-disease states. Although electronic noses are relatively inexpensive and rapid, they are limited in their sensitivity and subject to ‘blind spots’, with alternative selective sensors sometimes being required for particular analytes.³⁵ The specific analytes that differentiate disease vs. non-disease states cannot be identified using eNoses, so alternative detection techniques need to be used to identify and validate disease-specific biomarkers.

3.8. Nanomaterial gas sensors

Nanomaterial sensors have increasingly been investigated for breath VOC analysis, with the potential for inexpensive, point of care assessments.³⁶ The nanomaterials most commonly used in these gaseous nanosensors include single-walled carbon nanotubes (CNTs), nanowires of various materials, nanoporous chemioptical materials, and monolayer capped metal nanoparticles (MCNPs)^37,38. Nanoscale coating of sensor material provides a large surface area-to-volume ratio, facilitating trapping of VOCs in exhaled breath^39,40. While designing nanomaterial gas sensors for clinical applications, it is necessary to tailor the sensor’s dynamic range to VOC concentrations in the breath. The sensitivity of these sensors depends on their materials, as the optimal design of these nanomaterial sensors also requires prior knowledge of analyte targets⁴¹. A significant limitation of these nanomaterial sensors is that it is not possible to design a nanomaterial that captures the full range of volatile analytes that are biomarkers of disease states.

4. Diagnostic uses of clinical breath tests for infections

The 2019 CDC Antibiotic Resistance (AR) Threats Report estimates 2.8 million infections due to AR pathogens with 35,900 attributable deaths per year.⁴² One primary driver for the selection of these AR organisms is the excessive use of empiric antimicrobial therapy, which is in turn driven by the relatively low sensitivity of cultures and other microbiologic techniques in identifying certain infectious diseases accurately and rapidly. Considerable progress has been made over the past decade in the identification of VOC biomarkers associated with specific diseases, laying the groundwork for faster and earlier diagnosis of infections, with an increasing focus on determining optimal diagnostic cut-off values of analyte concentrations for this purpose.

In this section, we review some of the most relevant studies that have characterized breath analytes for the identification of specific infections.

4.1. Viral infections

The ongoing COVID-19 pandemic has illustrated the need for widespread non-invasive, rapid, and accurate testing for screening and diagnosis to contain the spread of highly contagious viral infections. Unlike bacteria, fungi, or parasites, viruses do not have their own metabolism, instead using the host cell’s metabolic machinery to replicate and spread, making it somewhat challenging to identify unique viral metabolites in the breath.

4.1.1. SARS-CoV-2

While rapid SARS-CoV-2 viral RT-PCR and antigen assays have come into widespread use over the past year, these assays still have diagnostic limitations and availability bottlenecks. To shorten testing time and make mass screening more feasible, breath based VOCs analysis has been proposed as an alternative sampling and testing technique⁴³. A recent study was performed to characterize metabolites in exhaled breath in adults undergoing invasive mechanical ventilation with severe COVID-19, compared to those with non-COVID-19 acute respiratory distress syndrome (ARDS)⁴⁴. The authors concluded that breath VOCs including methylpent-2-enal, 2,4-octadiene, 1-chloroheptane, and nonanal may help distinguish patients with COVID-19 from those with ARDS. This study, however, was limited by a small sample size (40 patients) and did not have an external validation cohort. Another study found that aldehydes (ethanal, octanal), ketones (acetone, butanone), and methanol were able to distinguish patients with COVID-19 from those with other conditions, including asthma, COPD, bacterial pneumonia, and cardiac disease⁴⁵. One recent pilot study in examined the exhaled breath of 26 children with suspected SARS-CoV-2 infection, finding six VOCs (octanal, nonanal, heptanal, decane, tridecane, and 2-pentylfuran) to be significantly elevated in infected patients compared to noninfected individuals. The sensitivity and specificity of this study were reported to be 100% and 66.6% respectively, although the sample size was small.⁴⁶ Jendrny et al. performed another pilot study to characterize COVID-19 patients through scent identification by dogs. They aimed to track the specific scent of SARS-CoV-2 infection and trained 18 dogs to discriminate saliva or tracheobronchial secretions samples from patients with SARS-CoV-2 patients versus samples from uninfected controls for 1 week. The dogs could identify the samples from patients with SARS-CoV-2 with an average diagnostic sensitivity of 82.63% and specificity of 96.35%, but with significant variability in performance between animals.⁴⁷ Ryan et al. proposed exhaled breath condensate (EBC) as an alternative test sample for SARS-CoV-2 RT-PCR analysis to help reduce the false negative testing rate of nasopharyngeal swab (NPS) samples. Although they reported enhanced sensitivity using the EBC sampling method (it detected the virus in 93.3% of negative NPS samples when using four gene targets (S/E/N/ORF1ab)), the sample size was small and the statistical significance of this difference could not be properly assessed.⁴⁸ Wintjens et al. used a commercially available electronic nose (Aeonose, The Aeonose Company, Zutphen, Netherlands) for preoperative screening of 219 asymptomatic patients, and reported a sensitivity of 86% and a NPV of 96% for asymptomatic infection, although this work requires further validation.⁴⁹

Although promising, it is not clear whether these signatures are specific for SARS-CoV-2 or whether they represent generalizable metabolic changes from respiratory infections. This warrants further study.

4.1.2. Influenza A

Traxler et al. obtained breath from swine infected with influenza A virus and observed that breath acetaldehyde, propanal, n-propyl acetate, methyl methacrylate, styrene and 1,1-dipropoxypropane were increased during infection and declined after infection⁵⁰. They used needle trap microextraction coupled with GC-MS to differentiate infected and non-infected swine. Based on culture headspace extraction, the same group reported an increased level of n-propyl acetate in cells with viral infection when compared with bacterial infection⁵¹. Purcaro et al. performed culture headspace extraction from microliter plates seeded with human laryngeal cells (both for Influenza A and RSV) and found compounds (RSV: 2-methyl-pentane, methyl sulfone, 2,4-dimethyl-heptane, 4-methyl-octane; Influenza A: acetone, n-hexane and other unidentified molecules) that could effectively differentiate infected and uninfected cells, and also observed fluctuations related to the degree of infection⁵². Another study used both headspace culture extraction and human breath analysis and reported that 2,8-dimethylundecane was decreased, and some alkanes present in exhaled breath were either increased or decreased after influenza vaccination⁵³. The liquid phase of breath has also been investigated as a potential sample for influenza RNA testing — to determine the size of particles carrying respiratory viral RNA during coughing and breathing, Gralton et al. monitored the exhaled breath of 53 patients with symptomatic respiratory viral infections and observed that they produce a range of particles carrying viral RNA when coughing and breathing⁵⁴.

4.2. Parasitic infections

While parasitology is one of the most neglected areas of infectious disease research, efforts have been made to identify breath profiles for some protozoans and helminths. The development of these assays might aid in the diagnosis of these neglected infections in areas with limited resources where they are most prevalent.

4.2.1. Malaria

A field study was performed in children with and without uncomplicated Plasmodium falciparum infection to assess the efficiency of a breath test to monitor natural human malaria by using thermal desorption-gas chromatography/mass spectrometry. These authors identified six VOCs — methyl undecane, dimethyl decane, trimethyl hexane, nonanal, isoprene, and tridecane —as potential markers of infection. They also observed that infection was correlated with the monoterpenes α-pinene and 3-carene in breath⁵⁵. Another similar study by Berna et al. concluded that concentrations of 9 compounds (carbon dioxide, isoprene, acetone, benzene, cyclohexanone, and 4 thioethers) varied significantly in individuals during the course of infection,⁵⁶ with diurnal cyclical elevation of thioethers in the breath with P. falciparum but not with P. vivax infection. Further studies could focus on species-specific diagnosis, helping guide antimalarial treatment, and this technique could also be an effective tool to assess infection clearance.⁵⁷

4.2.2. Cutaneous leishmaniasis

Welearegay et al. used ligand-capped ultrapure metal (gold, copper, platinum) nanoparticle sensors for the detection of breath compounds of 28 Tunisian patients with cutaneous leishmaniasis and 32 healthy controls, followed by GC-MS analysis. They reported 96.4% sensitivity and 100% specificity for leishmaniasis with 9 potential biomarkers of infection: 2,2,4-trimethyl pentane, 4-methyl-2-ethyl-1-pentanol, methylvinyl ketone, nonane, 2,3,5-trimethyl hexane, hydroxy-2,4,6-trimethyl-5-(3-methyl-2 butenyl)cyclohexyl methylacetate, octane, 3-ethyl-3-methylheptane, and 2-methyl-6-methylene-octa-1,7-dien-3-ol. This study not only expands on the diagnosis of this neglected entity but also provides an encouraging example of the use of chemical sensors for accurate and early non-invasive detection of pathogens. ⁵⁸

4.2.3. Echinococcosis

Another promising study examined breath metabolites in 32 patients with cystic echinococcosis (CE) in Tunisia and 16 patients with alveolar echinococcosis (AE) in Poland, with 51 control patients enrolled at both study sites, finding a distinct profile for each subtype of this infection. Using GC-Q/TOF and ultrapure metal nanoparticle chemoresistive gas sensors for faster sample analysis, the authors reported 75% sensitivity and 86.7% specificity for CE, 92.9% sensitivity and 88.9% specificity for AE, and 91.7% sensitivity and 92.7% specificity for distinguishing between CE and AE. They found two compounds, 1-tridecene and (E)-13-docosenoic acid, to be specific for CE and seven compounds (hexadecane, heptadecane, eicosane, 11-(pentan-3-yl)henicosane, tetratriacontane, 2-methyloctacosane, hentriacontane) in AE. The authors highlighted the potential of this test, especially using nanoparticle chemoresistive gas sensors, for population screening in remote areas.⁵⁹

4.3. Bacterial infections

It remains highly challenging to identify the specific microbial etiology of bacterial pneumonia, despite its high incidence worldwide, and empiric antibiotic therapy remains the most common strategy in the clinical management of suspected bacterial pneumonia. Current diagnostic methods for bacterial pneumonia are highly limited in their sensitivity, in particular, and many patients with suspected bacterial pneumonia are treated with antibiotics without any diagnostic workup.

A study by Rosón et al. assessed 533 patients with community-acquired pneumonia (CAP) and found that only 39% were able to produce quality sputum samples, with sensitivity of only 57% for identification of Streptococcus pneumoniae.⁶⁰ Another study by Ewig et al. found sputum Gram stain and culture only useful in 24% of suspected CAP cases.⁶¹ Molecular analysis of sputum with techniques such as the rapid multiplex BioFire FilmArray Pneumonia Panel (FA-PP) demonstrate little incremental diagnostic sensitivity over culture, with a positive percentage agreement of 94.4% with standard respiratory cultures.⁶² Finally, Garg et al. described the spatial variations of bacteria within the lung in cystic fibrosis patients, which raises even more questions about the reliability of respiratory samples.⁶³ Because of these limitations in the sensitivity of existing diagnostic testing for CAP, breath-based testing for bacterial pneumonia is a potentially attractive alternative, with the relative simplicity of sample collection and ability to access microbial VOCs from within the airways and lung parenchyma.

4.3.1. Tuberculosis (TB)

A systematic review by Saktiawati et al. examining 14 studies that used an electronic nose to diagnose pulmonary tuberculosis found a pooled sensitivity and specificity of 93%, but they also concluded that further development is needed to provide real-time results. It is also necessary to conduct studies in patients with culture-negative TB, where a breath test might have most utility, particularly children.⁶⁴ A study by Kolk et al. in South Africa analyzed breath samples from 171 patients using GC-MS and found a sensitivity of 62% and specificity of 84% after a secondary validation. They found seven compounds that discriminated breath samples of TB and non-TB patients and they were different from those produced in vitro by M. tuberculosis. This led to the conclusion that these markers were potentially generated by the host response to TB rather than product of bacterial metabolisms.⁶⁵ Another study by Phillips et al. also used GC-MS to test 226 high-risk symptomatic patients and reported 84% sensitivity, 64.7% specificity and 85% accuracy for diagnosing active pulmonary TB overall. They reported a profile of active pulmonary tuberculosis that included the following ten compounds: oxetane, 3-(1-methylethyl)-; dodecane, 4-methyl-; cyclohexane, hexyl-; bis-(3,5,5-trimethylhexyl) phthalate; benzene, 1,3,5-trimethyl-; decane, 3,7-dimethyl-, tridecane; 1-nonene 4,6,8-trimethyl-; heptane 5-ethyl-2-methyl; and 1-hexene, 4-methyl-.⁶⁶ Both studies suggest that breath testing for TB might have a greater benefit when used for screening rather than diagnosis, although the field of TB VOC biomarkers in general suffers from a lack of reproducible biomarkers between individual studies.

A pilot study by Mellors et al. reported two biomarkers (4-(1,1-dimethylpropyl)phenol and 4-ethyl-2,2,6,6-tetramethylheptane) that were not previously associated with M. tuberculosis both in vitro and in the exhaled breath of 9 macaques, and they also described 38 breath compounds that discriminated between infected and uninfected macaques with an AUC of 98%.⁶⁷ The same group recently published another study with 31 pediatric subjects where they identified 4 compounds (decane, 4-methyloctane, and two unidentified analytes (A and B)) that differentiated 10 patients with confirmed TB (either by culture or Xpert MTB/RIF) from 10 patients without microbiological evidence of TB with a sensitivity of 80% and a specificity of 100%. Patients with a high clinical suspicion of TB without positive microbiological confirmation of TB were excluded from these calculations; however, these 4 compounds were also found in their breath.⁶⁸

McNerny et al. validated an immunosensor and bio-optical breath analyzer device to detect the M. tuberculosis antigen Ag85B in cough droplets on a field study. The proposed breath analyzer was able to detect this antigen in 23 of 31 (74%) patients with a specificity of 79%. While a larger study would be needed for validation, this approach shows promise for diagnosis of culture-negative TB, which remains a significant unmet need⁶⁹.

4.3.2. Pseudomonas aeruginosa

Most studies in Pseudomonas aeruginosa infection have focused on the identification of respiratory tract colonization with this species. Purcaro et al. collected the exhaled breath of mice in Tedlar sampling bags, with GC-TOF-MS analysis. They reported nine metabolites (4 alkylated hydrocarbons, isoborneol, p-cymene, 2-hexanone, an alkylated alcohol and an unknown compound) that could differentiate between infected and uninfected mice and ten metabolites (2-hydroxyethyl acetate, 1,3-dimethylcyclopentanol, 2-methyl-2-butenal, (E)-, 4-cyclopentene-1,3-dione, aldehyde 1, alkylated hydrocarbon 5 and 6, cyclohexanol, and two methylated fatty acids) that could differentiate between strains of P. aeruginosa. They used UV-killed and live P. aeruginosa isolates and found compounds such as 2-hexanone, among others, to be increased in mice infected with killed bacteria, and different compounds including isoborneol to be increased in mice infected with live bacteria, the hypothesis being that the latter were the result of bacterial metabolism or the interaction of pathogen and murine host.⁷⁰ Suarez-Cuartin et al. used electronic noses to differentiate patients with bronchiectasis with and without P. aeruginosa respiratory tract colonization with 72% accuracy, and they were able to differentiate colonization by P. aeruginosa from other colonizers (such as Haemophilus influenzae) with 89% accuracy.⁷¹ Robroeks et al. characterized 14 VOCs via GC-TOF-MS in the breath of pediatric patients with cystic fibrosis that identified with 100% accuracy the presence of airway colonization by Pseudomonas aeruginosa.⁷² Another study reported increased amounts of hydrogen cyanide (HCN) in the breath of children with cystic fibrosis and P. aeruginosa colonization. High levels of breath HCN have been associated with P. aeruginosa in vivo and in vitro. ^73,74

Rabis et al. used multi-capillary column-ion mobility spectrometry (MCC-IMS) to examine the breath of 57 adult patients and found 21 signals that could differentiate patients colonized or infected with Pseudomonas aeruginosa from controls.⁷⁵

4.3.3. Staphylococcus aureus

A study in pediatric patients with cystic fibrosis was able to identify distinct VOC profiles (9 compounds, some increased in infected patients (1,4-pentadiene, ethanol, acetone, 2-butanone, undecane, 2-methyl-naphthalene) and some decreased (3-hydroxy-2-butanone, hexanal, isopropyl myristate) using GC-MS that differentiated patients with and without S. aureus airway infection with 100% sensitivity and 80% specificity. This study was limited by a small sample size, with only 18 patients in total, so further evaluation in a larger cohort is necessary.⁷⁶ Zhu et al. studied S. aureus and P. aeruginosa pneumonia murine models for a 120-hour period using secondary electrospray ionization-mass spectrometry (SESI-MS) with partial least squares-discriminant analysis (PLS-DA) and described the infection “breathprint” pattern at six time points, rather than determining specific volatile compounds (similar to the electronic nose approach). They found changes in these breathprints predictive of time to clearance of infection between subjects, which they suggested might be used to predict the course of infections in humans and guide antimicrobial prescribing practices. They also concluded that examining breath signatures overall, rather than focusing on specific metabolites, might provide a fuller picture of the interaction between bacterial and host metabolites and mitigate the effect of inter-subject variation.⁷⁷

4.3.4. Streptococcus pneumoniae

Van Oort et al. performed intratracheal inoculation of male adult rats with S. pneumoniae or P. aeruginosa in conjunction with controls treated with saline. They then performed mechanical ventilation 24 hours later to extract breath VOCs and analyzed these compounds using GC-MS and SIFT-MS. In their GC-MS results, they found eight compounds that could discriminate between infected and noninfected animals (4-methyloctane, octane, 2–5-dimethyl, tetrachlorethylene, an unidentified naphthalene compound, two unidentified cyclic compounds, an unidentified branched aldehyde, and another unidentified compound) with an ROC area under the curve (AUC) of 0.93; fourteen compounds could differentiate between S. pneumoniae and controls (octane, 4-methyl-, octane, 2,5-dimethyl, hexadecane, hexane, 2-,4-dimethyl-, 2-propanol, 1-methyloxy-, nonane, 2-methyl-, heptane, 2-,4-dimethyl, two unidentified cyclic compounds, an unidentified naphthalene compound, and four other unidentified compounds) with a ROC AUC of 0.93; three compounds could distinguish P. aeruginosa from controls (2-propenoic acid, 2-ethylhexyl ester, unidentified branched aldehyde, and an unidentified cyclic compound) with a ROC AUC of 0.98, and the ROC AUC for S. pneumoniae vs P. aeruginosa was 0.99. GC-MS proved to be more accurate for discrimination than SIFT-MS. These results support the potential for establishing species-specific bacterial profiles in bacterial pneumonia.⁷⁸

4.3.5. Helicobacter pylori

After alcohol breathalyzer tests, the urea breath test for H. pylori infection is probably the most commonly used breath test, although this assay requires administration of exogenous ¹³C-labelled urea. H. pylori in patients infected with this bacteria hydrolyzes the ¹³C-labelled urea into CO₂ and ammonia, with the labelled CO₂ being exhaled through the lungs and detected in the breath. This test not only helps diagnose H. pylori infection but can also assess the response to treatment.⁷⁹ A study by Maity et al. developed a point-of-care and cost-effective residual gas analyzer mass spectrometry-based sensor coupled with a high vacuum chamber to detect H. pylori from human breath in real-time. They reported a 100% sensitivity and 93% specificity with PPV and NPV of 95% and 100%, respectively, when compared to biopsy H. pylori assessment in 35 patients.⁸⁰. A study by Ulanowska et al. used SPME-GC/MS to identify compounds such as 2-butanone, isobutane, and ethyl acetate that differentiated patients with H. pylori from healthy controls.¹³

4.3.6. Acinetobacter baumannii

One of the biggest challenges in identifying bacterial pneumonia is differentiating airway colonization from active infection. Gao et al. studied 20 patients with A. baumannii ventilator-associated pneumonia (VAP), 20 patients with A. baumannii airway colonization, and 20 ventilated patients with no A. baumannii in their respiratory tract, and identified eight compounds (1-undecene, nonanal, decanal, 2,6,10-trimethyldodecane, 5-methyl-5-propylnonane, longifolene, tetradecane and 2-butyl-1-octanol) that discriminated between active infection and colonization.⁸¹ Only four of these compounds (1-undecene, decanal, longifolene, tetradecane) matched their previous in vitro culture results, which supports the findings by Zhu et al. that reported only a 25–34% of shared peaks between in vivo and in vitro samples, suggesting a significant metabolic shift between bacterial growth in media and bacterial growth in the lung milieu.⁸²

4.3.7. Ventilator-associated pneumonia (VAP)

A commercially available electronic nose was assessed for the diagnosis of VAP.^83,84 Using computerized tomography scans as part of the reference standard, these studies reported a relatively low level of accuracy (~60%) for discriminating patients with and without VAP. A pilot study by Filipiak et al. found some compounds to overlap between the headspace gas of cultured swabs and the breath of patients infected with these bacteria. The presence of dimethyl sulfate (DMS) was associated with non-species-specific emerging infection. This study was also able to identify markers associated with pathogen-derived metabolites in 22 ventilated patients. The presence of multiple co-infections and the small sample size did not allow for full characterization of a profile for each bacterial species in vivo, but they did report some specific compounds such as 3-methyl-1-butene for H. influenzae, 1-undecene for P. aeruginosa, propene and butane for S. aureus (also showing decreased levels after treatment), and acetonitrile and 2-pentanone for E. coli.⁸⁵

Fowler et al. examined a wide range of breath VOCs present in lower respiratory tract infections and concluded that the host inflammatory response might influence breath VOCs.⁸⁶ Another study identified 1-propanol as a potential marker to track general bacterial growth in patients with pneumonia⁸⁷. Some of these identified VOCs are potentially products of bacterial metabolic pathways and may be viable markers for screening of patients with suspected VAP.⁸⁸

4.3.8. Acute appendicitis

Andrews et al. studied breath samples from 53 patients in a surgical unit with suspected acute appendicitis using ion-molecule reaction-mass spectrometry (IMR-MS) and found significant differences in the concentration of acetone, isopropanol, propanol, butyric acid and other unidentified compounds with m/z ratio of 56, 61 and 87 in patients with proven acute appendicitis compared to patients without appendicitis.⁸⁹

4.4. Fungal infections

Fungal infections are a major cause of significant morbidity and mortality, particularly in immunocompromised patients. The diagnosis of these infections is often delayed due to the nonspecific symptoms of invasive fungal disease, limited sensitivity of respiratory tract and blood cultures and fungal antigen tests, and the general debilitation of immunocompromised patients, which makes definitive invasive biopsy procedures challenging.⁹⁰ Because of the difficulty of diagnosing these infections, mortality due to invasive fungal disease remains very high despite the availability of a wide range of highly potent antifungal drugs.⁹¹ More sensitive, rapid, and noninvasive tests would have a major impact on the management of patients with suspected invasive fungal disease.

4.4.1. Invasive aspergillosis

Aspergillus species are ubiquitously distributed in the environment and have virulence factors that make invasive aspergillosis (IA) the most common invasive mold infection in immunocompromised patients and a common cause of morbidity in patients with chronic lung disease. Aspergillus fumigatus has been described to produce 2-pentylfuran when it is cultured on blood agar and nutrient agar media. Based on this observation, one group analyzed breath from healthy subjects, patients receiving chemotherapy, and adults colonized or infected with A. fumigatus with underlying comorbidities such as bronchiectasis, cystic fibrosis, or immunosuppression. The authors reported detection of 2-pentylfuran was able to detect A. fumigatus in the airways with moderate sensitivity (77%) and specificity (78%)⁹². However, further studies have not found 2-pentylfuran to discriminate breath samples from patients with and without A. fumigatus infection. Koo et al. characterized the breath volatile metabolic profile of Aspergillus fumigatus and other Aspergillus species in vitro, and in the breath of 64 patients with suspected invasive aspergillosis using thermal desorption GC-MS. They identified a signature of sesquiterpene compounds including α-trans-bergamotene, β-trans-bergamotene, a β-vatirenene-like sesquiterpene, and trans-geranylacetone that could identify patients with IA with 94% sensitivity and 93% specificity.⁹³ A study using an electronic nose suggested a distinct VOC profile in breath of patients with prolonged chemotherapy-induced neutropenia with IA vs. controls.⁹⁴ This group also found that A. fumigatus respiratory tract colonization could be determined with 89% accuracy in patients with cystic fibrosis using this electronic nose.⁹⁵

4.4.2. Oral candidiasis

Hertel et al. collected breath samples from 10 patients with oral candidiasis and 10 controls and found no correlation with the compounds that have been described in vitro in cultures of Candida spp. These authors did not find a specific volatile profile in oral candidiasis, but found that compounds like methyl acetate, 2-methyl-2-butanol, hexanal and longifolene declined with antifungal therapy and that other compounds such as 1-heptene, acetophenone, 3-methyl-1-butanol, decane, and chlorbenzene increased after treatment in some patients.⁹⁶

5. Conclusions:

Breath analysis has the potential to improve the identification of pathophysiological processes, especially infectious diseases, as microbes have many unique metabolic pathways that are distinct from human metabolism. While there are many technical challenges in the rigorous identification and validation of breath biomarkers that distinguish patients with and without these infections, breath remains a particularly attractive matrix because of the noninvasive nature of sample collection and the ability to assess microbial processes that are particularly challenging to identify with standard culture, antigen, and molecular amplification-based approaches, especially infections that are predominantly based in the lung. With the recent development of increasingly sensitive analytical instruments and point-of-care technologies capable of identifying and quantifying volatile analytes at ultra-trace levels and with further delineation of biomarker signatures that accurately identify patients with specific infectious disease syndromes, breath-based assays for these infections are highly likely to be adopted in clinical settings over the next decade. These breath-based assays would offer noninvasive, rapid, real-time identification of specific infections earlier than possible with current methods, in turn facilitating early, appropriate antimicrobial prescribing, reducing unnecessary antimicrobial exposure, and ultimately improving clinical outcomes in patients with these infections.

Synopsis.

Various analytical methods can be applied to concentrate, separate, and examine trace volatile organic metabolites in the breath, with the potential for non—invasive, rapid, real-time identification of various disease processes, including an array of microbial infections. While biomarker discovery and validation in microbial infections can be technically challenging, it is an approach that has shown great promise, especially for infections that are particularly difficult to identify with standard culture and molecular amplification-based approaches. In this review, we discuss the current state of breath analysis for the diagnosis of infectious diseases