Harnessing breath biomarkers for pneumonia diagnosis and prognosis

Abstract

Introduction:

Pneumonia is a major cause of death and disability worldwide. A host of pathogens causes pneumonia, and pneumonia presents with a remarkable heterogeneity of clinical symptoms and signs and has varied outcomes. Current approaches to pneumonia diagnosis and risk stratification lack precision such that there is no universally agreed upon biomarker or scoring system. These limitations have prompted calls for novel, noninvasive, and more precise approaches to better diagnosing pneumonia and predicting outcomes.

Areas covered:

We performed a comprehensive literature search through PubMed to identify studies reporting on breath biomarkers in pneumonia published up to 31 July 2025. This manuscript explores breath-based metabolomics as a novel approach to biomarker development in pneumonia. It describes breath collection methods, including devices available and types of breath samples for analysis. It reviews the potential role of exhaled breath analysis to expedite pneumonia diagnosis, monitor response to therapy, and predict clinical trajectory.

Expert opinion:

Breath-based metabolomics could improve the recognition and management of pneumonia. It is a noninvasive, potentially continuous method that provides a direct window into the lung for novel insights into the underlying biology of pneumonia.

1. Introduction

Pneumonia is a common disease and is potentially deadly. Community-acquired pneumonia (CAP) remains a leading cause of morbidity and mortality worldwide [1] causing more than 1.5 million annual hospitalizations [2] and inpatient mortality rates of 6.5% and 30.6% at 1 year [2]. The economic burden of CAP is substantial [3] and inpatient hospital-acquired pneumonia (HAP) and ventilator-associated pneumonia (VAP) are all further associated with high mortality, hospital length of stay, and cost of care [4].

A variety of pathogens cause pneumonia with rhinovirus (9%), influenza (6%) and Streptococcus pneumoniae (5%) the most common pathogens recognized in US adults hospitalized with CAP; however, the etiology of most CAP events is never identified [5]. Most individuals infected with these pathogens do not develop pneumonia and, amongst those who do, the causal pathogen does not predict illness severity [6].

Although common, pneumonia lacks a standardized clinical definition. The diagnosis relies on chest imaging coupled with clinical signs and symptoms (e.g. fever, cough, hypoxia) [7], while the diagnosis of other forms of pneumonia, such as VAP, can be even more challenging, given that many of these patients have preexisting changes on their chest radiograph (CXR) [4]. Severe CAP is similarly prone to subjectivity, with no universal criteria and often defined by requirement for intensive care unit (ICU) admission [8].

Numerous biomarkers and severity scoring systems have been developed in response to the inherent challenges of pneumonia diagnosis and risk stratification and, consequently, there is no universally agreed upon marker or measure. For example, the pneumonia severity index (PSI) and CURB-65 are considered overly sensitive in predicting severity [9]. Procalcitonin, perhaps the most widely used biomarker, can help detect bacterial infection and determine duration of antibiotic therapy in pneumonia [10]. However, it is less reliable in diagnosing VAP or HAP, compared to CAP, possibly due to preexisting systemic inflammation in these patients [11]. It can be elevated in a number of noninfectious conditions and can take up to 48 hours to peak after infection onset [12].

These limitations have prompted interest in novel approaches to diagnose and predict outcomes in pneumonia. Breath-based metabolomics is a potential real-time, noninvasive method focused on the identification and quantification of metabolomic markers in exhaled breath. The purpose of this review is to highlight the potential role for breath analysis to aid in the clinical diagnosis, therapeutic monitoring, and outcome prediction in pneumonia.

We performed a comprehensive PubMed search to identify studies reporting on breath biomarkers in pneumonia published up to 31 July 2025. We included observational, in vitro, and animal studies in our search. Studies not published in English were excluded. Please see Table 1 for an overview of relevant studies.

Table 1.

Characteristics of included studies on acute respiratory infection or failure.

Author (Year)	Comparison	Sample Size	Technique	Outcome Measure	Ref.
Borras (2021)	Influenza vaccination vs none RVP positive vs RVP negative	n = 12 (subjects served as own cases) n = 7 RVP+, n = 5 RVP−	LC-MS	EBC metabolites	[54]
McCartney (2021)	Influenza A H1N1 vs rhinovirus vs healthy control	In vitro study	GC-MS	VOC profile	[38]
McCartney (2022)	COVID-19 delta vs COVID-19 omicron vs healthy control	n = 18 Delta, n = 28 Omicron, n = 96 healthy control	GC-MS	VOC profile	[39]
Ryan (2021)	COVID-19 positive vs COVID-19 negative	n = 16 NPS+, n = 15 NPS−/clinically+, n = 9 NPS−/other diagnosis	RT-PCR	EBC RT-PCR	[55]
Sharma (2023)	COVID-19 positive vs COVID-19 negative	n = 94 COVID-19 positive, n = 77 COVID-19 negative	GC-MS	VOC profile	[84]
Aliberti (2016)	CAP severity	N = 74 CAP	Chemiluminescent immunoassay	EBC cytokine levels	[95]
Davis (2020)	Clinical severity among IMV patients, measured by clinical pulmonary infection score, chest radiograph score, and development of pneumonia, sepsis, or death	n = 49 (13 developed pneumonia, 9 developed sepsis)	Multiplex based immunoassay	EBC cytokine levels	[94]
Ibrahim (2022)	Acute asthma vs acute COPD vs acute heart failure vs CAP vs control	n = 277 cardiorespiratory exacerbations (asthma n = 33, COPD n = 29, heart failure n = 22, CAP n = 27), n = 277 healthy controls	GC-MS	VOC profile	[81]
Majewska (2004)	CAP vs no CAP	n = 43 CAP, n = 20 healthy controls	Spectrofluoro metrical assays	EBC H2O2 and TBARs	[64]
Nayeri (2002)	CAP vs no CAP	n = 10 CAP, n = 10 non-respiratory infection, n = 11 healthy controls	ELISA immunoassay	EBC hepatocyte growth factor	[85]
Stolarek (2006)	CAP nonsmoker vs CAP smoker	n = 24 CAP nonsmoker, n = 19 CAP smoker	Spectrofluoro metrical assay	EBC H2O2	[68]
Ahmed (2023)	VAP vs no VAP	In vitro cultures n = 45 BAL+VAP+, n = 44 BAL−VAP−	GC-MS	VOC profile	[79]
Bakali (2024)	VAP only	n = 12 BAL+VAP+	GC-MS	Carbon disulfide	[49]
Felton (2023)	VAP only	N = 96	GC-MS	Benzene, cyclohexanone, pentanol, and undecanol	[50]
Filipiak (2024)	VAP vs no VAP	n = 32 BAL+VAP+, n = 6 BAL−VAP−	GC-MS	VOC profile	[78]
Filipiak (2024)	Candida albicans vs Staphylococcus aureus detected in VAP	In vitro cultures n = 6 C. albicans, n = 3 S. aureus, n = 4 both	GC-MS	VOC profile	[77]
Fowler (2015)	IMV lower respiratory tract pathogen-positive vs IMV pathogen-negative	n = 54 enrolled with 20 remaining pathogen negative on serial samples	GC-MS	VOC profile	[47]
Gao (2016)	Acinetobacter baumannii VAP vs colonization	In vitro cultures n = 20 A. baumannii VAP, n = 20 colonized, n = 20 IMV controls	GC-MS	VOC profile	[80]
Jin (2020)	VAP vs no VAP	n = 32 VAP, n = 32 no VAP	Enzyme immunoassay and calorimetry	EBC 8-isoprostane and NO levels	[63]
May (2015)	VAP vs no VAP	n = 20 BAL+VAP+, n = 31 BAL−VAP−	RT-PCR	EBC bacterial DNA	[62]
Schnabel (2015)	VAP vs no VAP	n = 32 BAL+VAP+ vs n = 68 BAL−VAP−	GC-MS	VOC profile	[51]
Schnabel (2015)	VAP vs no VAP	n = 33 BAL+VAP+, n = 39 BAL−VAP−	eNose	VOC profile	[52]
Van Oort (2017)	VAP vs. IMV colonized vs controls	n = 12 probable pneumonia vs n = 21 possible pneumonia vs n = 13 colonized vs n = 47 control	GC-MS	VOC profile	[48]
Van Oort (2022)	VAP vs no VAP	N = 52 BAL+VAP+, n = 56 BAL−VAP−	GC-MS	VOC profile	[53]
Baldwin (1986)	ARDS vs non-ARDS	n = 16 ARDS, n = 27 non-ARDS	Scopoletin/horseradish peroxidase assay	EBC H2O2	[65]
Kietzmann (1993)	ARDS vs IMV with ARDS risk factors vs IMV with pulmonary edema vs postoperative IMV controls	n = 7 ARDS, n = 16 ARDS risk factors, n = 3 pulmonary edema, n = 10 controls	Chemiluminescence assay	EBC H2O2	[66]
Sharma (2022)	ARDS vs non-ARDS	n = 5 swine with ARDS	GC-MS	VOC profile	[83]
Sznajder (1989)	ARDS vs non-ARDS	n = 55 ARDS, n = 13 non-ARDS	Spectrophoto metric assay	EBC H2O2	[69]
Zhou (2019)	ARDS vs non-ARDS	n = 21 ARDS, n = 27 non-ARDS	GC-MS	VOC profile	[82]

Abbreviations: CAP, community-acquired pneumonia; IMV, invasive mechanical ventilation; VAP, ventilator-associated pneumonia; BAL, bronchoalveolar lavage; RVP, respiratory viral panel; NPS, nasopharyngeal swab; TB, tuberculosis; NO, nitric oxide; H2O2, hydrogen peroxide; TBARs, thiobarbituric reactive substances; ARDS, acute respiratory distress syndrome.

2. Methodologies in exhaled breath research

Exhaled breath analysis is not new and several tests have become standard in specific clinical settings. For example, continuous exhaled carbon dioxide measurement to assess the adequacy of patient ventilation is routine in many hospital environments. Exhaled breath nitric oxide is used as a biomarker of eosinophilic airway inflammation and it can help characterize asthma [13]. Not surprisingly, both exhaled carbon dioxide and nitric oxide have been explored as potential biomarkers in CAP as well [14–20]. Further investigation is clearly needed.

In addition to oxygen, carbon dioxide, nitric oxide, and other gases, many molecules diffuse across the alveolar membrane and capillary blood in the direction of higher to lower vapor pressure during respiration [21]. As a result, each alveolar breath contains hundreds of different metabolites, with the potential to provide insight into biochemical processes in the human body. Exhaled breath contains small inorganic compounds (e.g. nitric oxide, oxygen, and carbon dioxide), volatile organic compounds (VOCs; e.g. aldehydes, hydrocarbons, alcohols, ketones, and esters), and nonvolatile compounds (e.g. proteins, lipids, oxidants, and nucleotides) [21,22].

Breath samples can be collected in spontaneously breathing or mechanically ventilated patients. Individuals can breathe normally or tidally for a few minutes into a mouthpiece or face mask [23–27]. Breath sampling from mechanically ventilated patients occurs either in-line with the ventilator circuit or, for exhaled breath vapor (EBV), directly from the exhaust port [27–29]. In-line samples offer the advantage of being collected closer to the patient, prior to being potentially lost in the ventilator circuit or trapped in the expiratory filters. However, unlike samples from the exhaust port, in-line samples cannot be collected continuously because the ventilator circuit must be interrupted to place and remove the collection device [28].

There is no uniform standard for breath collection, though several techniques are available. VOCs and small compounds are most easily collected in the gas phase as EBV while the larger, nonvolatile compounds are captured as exhaled breath condensate (EBC) or aerosols. For EBV, metabolites can be preconcentrated and sampled using various methods such as solid-phase microextraction (SPME), needle-trap device (NTD), or adsorption tubes prior to analysis with gas chromatography mass spectrometry (GC/MS) or proton-transfer-reaction mass spectrometry (PTR-MS) [30–33].

A variety of devices have been developed for EBC collection including the ‘K-tube’ by our group [34,35], ECoScreen^® (Erich Jaeger GmbH, Hoechberg, Germany), TurboDECCS (MEDIVAC, Parma, Italy), and RTube^™ (Respiratory Research, Inc., Austin, TX, U.S.A.) [25,35,36]. These devices are optimized to cool the breath sample into liquid or solid form, facilitating the detection of larger biomarkers such as proteins and lipids that are present in fine aerosols. This technique can technically be applied to VOC capture by subjecting the EBC to SPME or other headspace extraction techniques, although this is less efficient than direct gas-phase collection. Once obtained, EBC samples undergo liquid chromatography mass spectrometry (LC/MS) or other analysis for metabolite measurement [34].

Some specialized devices are further advanced and appear close to commercialization for specific respiratory infections. EBCare (exhaled breath condensate analysis and respiratory evaluation), a smart mask device, was recently developed for real-time in situ EBC biomarker analysis. The device integrates a tandem passive cooling strategy, microfluidic system inspired by the capillary phenomenon observed in plants, and a wireless electrochemical biosensor array for portable monitoring and analysis. It has been used to study respiratory airway inflammation, including in COVID-19 [37]. Similarly, TreSenso Tech developed a portable breath analyzer designed to detect VOCs in breath samples for early diagnosis of tuberculosis, with plans to expand to other respiratory problems (https://tresensotech.com/#technology). These devices rely on advanced technologies that promise to transform breath testing from experimental to clinically applied, precision-based science.

3. Applications of exhaled breath analysis in pneumonia

3.1. Community-acquired pneumonia

Community-acquired pneumonia (CAP) occurs when a bacterial, viral, or fungal pathogen invades the lower respiratory tract, resulting in infection and host inflammatory response. Standard diagnosis relies on clinical signs and symptoms as well as chest imaging findings, most commonly chest radiography. However, the lack of highly sensitive and specific clinical or diagnostic testing for CAP results in overdiagnosis and missed cases [38].

Breath analysis, particularly with EBV, may facilitate the early recognition of respiratory infections causing pneumonia, due in part to host-response to infection as well as VOC signal from bacterial/fungal pathogens in the pulmonary space [39]. For example, VOC profiles in a cell culture model of human airway epithelial cells distinguished influenza from rhinovirus infections and predicted timing of infection [40]. Breath samples analyzed from adults with and without confirmed COVID-19 infection demonstrated a COVID-19 volatile signature that differed by COVID-19 variant [41]. Importantly, a general COVID-19 model that did not account for variant demonstrated moderate accuracy (0.73 ± 0.06), which improved to 0.82 ± 0.12 for Delta only and 0.84 ± 0.06 for Omicron only [41]. These findings likely reflect underlying differences in the host response to different variants, which should be considered in the design of future breath-based tests for COVID-19 and other respiratory pathogens.

EBC could also help identify infected patients. In a study of hospitalized adults with confirmed upper respiratory viral infection, EBC discriminated influenza infection from other viruses [42]. EBC also improved detection of SARS-CoV-2 viral RNA via reverse transcription PCR (RT-PCR) in patients with negative RT-PCR nasopharyngeal swabs despite being clinically positive for COVID-19 [43].

Concentrations of hydrogen peroxide (H₂O₂) and thiobarbituric reactive substances (TBARs), both reactive oxygen species, have been explored as markers in CAP [44]. These compounds, which reflect lipid peroxidation and oxidative stress, contribute to the host defense against microorganisms and are elevated in pneumonia [44]. They are also elevated in other pulmonary and systemic inflammatory conditions including acute hypoxemic respiratory failure, acute respiratory distress syndrome (ARDS), and sepsis, even in patients without pneumonia [45–49].

3.2. Ventilator-associated pneumonia

Ventilator-associated pneumonia (VAP) is one of the most frequent infections in patients requiring invasive mechanical ventilation, with significant associated morbidity, mortality, and increased healthcare costs [50]. It is defined as a nosocomial bacterial pneumonia that develops in patients exposed to invasive mechanical ventilation for at least 48 hours. VAP can be classified as early-onset (within 48–72 hours of intubation) or late-onset (after 72 hours). Early onset causes are typically caused by antibiotic-sensitive bacteria (e.g. methicillin-sensitive Staphylococcus aureus, Haemophilus influenzae, and Streptococcus pneumoniae), while late onset VAP is often caused by antibiotic-resistant pathogens (e.g. methicillin-resistant Staphylococcus aureus, Pseudomonas aeruginosa, Acinetobacter, and Enterobacter) [51].

In VAP, pathogen identification is crucial for accurate diagnosis and treatment. The diagnosis currently requires culture technology, which can take two to three days to become positive. Metabolites in breath profiles, which are available instantaneously and at the bedside, have identified lower respiratory tract pathogens in patients at risk for VAP [52,53]. In a longitudinal study of mechanically ventilated patients with sterile brain injury, breath volatile profiles distinguished pathogen-positive versus pathogen-negative patients while blood inflammatory biomarkers (procalcitonin, IL-10-, IL-6, IL-10/IL-6 ratio, white blood cell count) did not [52]. VOC panels have also been explored to discriminate between patients with and without VAP [53–58]. These studies were limited by small sample sizes and have not been externally validated.

Like EBV, EBC has been explored as a biomarker for VAP in mechanically ventilated patients. Molecular analysis of bacterial DNA in EBC correlated with pathogens isolated from bronchoalveolar lavage fluid samples, suggesting that EBC could provide a timelier adjunct to current culture technology and expedite diagnosis [59]. Nitric oxide and 8-isoprostane levels have also been found to be elevated in the EBC of patients with VAP [60].

4. Identifying specific VOC markers through in vitro approaches

In vitro studies suggest that different bacteria produce a variety of different metabolites, which could facilitate the identification of specific pathogens [61]. A systematic review of 31 articles identified candidate biomarkers associated with the six most common and pathogenic bacteria in sepsis [61]. Only a small number of these metabolites were exclusive to one of the bacterial pathogens studied, and it is much more common to examine the relative ratios of VOCs to specifically identify the microorganisms. Differences in culture media, timing of bacterial growth phases, and genomic variation between bacterial strains likely contributed to inconsistent findings between studies and have limited global identification of specific biomarkers. However, very compelling proof-of-concept studies suggest this line of inquiry has tremendous potential.

Unfortunately, pathogen-related biomarkers isolated in vitro do not always directly translate clinically, possibly due to being absent or below the limits of detection in vivo. For example, a study investigating the synergistic role of Candida albicans in increasing the virulence of Staphylococcus aureus pneumonia reported inconsistent volatile biomarkers in the in vitro versus in vivo settings [62]. In vitro analyses identified breath-based metabolic signatures for the most common pathogenic bacteria in VAP (i.e. A. baumannii, E. coli, K. pneumoniae, and P. aeruginosa). These findings could not be replicated in vivo, although the small sample size available for each pathogen may have contributed to these results [63]. Another, larger study identified 19 VOCs in the in vitro culture of four common pathogens (S. aureus, P. aeruginosa, K. pneumoniae, and E. coli), with 14 of these replicated in the exhaled breath of infected patients [64].

Discrepancies between in vitro versus in vivo experiments could also reflect host-pathogen interactions. To date, only one study reported high diagnostic accuracy, both in vitro and in vivo, in the identification of a specific pathogen. In the in vitro setting, a set of nine VOCs (2,5-dimethyl-pyrazine, 1-undecene, isopentyl 3-methylbutanoate, decanal, 1,3-naphthalenediol, longifolene, tetradecane, iminodibenzyl, and 3-methyl-indene) were cultivated from the headspace of A. baumannii [65]. A set of eight VOCs (1-undecene, nonanal, decanal, 2,6,10-trimethyl-dodecane, 5-methyl-5-propyl-nonane, longifolene, tetradecane, and 2-butyl-1octanol) discriminated among ventilated patients with A. baumannii VAP, ventilated patients colonized with A. baumannii, and ventilated controls [65]. The different in vitro versus in vivo profiles (only four compounds were the same) again suggests that biomarkers derived in vitro often do not translate to the in vivo setting. These apparent discrepancies likely reflect dynamic changes in the in vivo setting that could provide insight into timing of infection, host response, and anticipated illness trajectory.

5. Making clinical diagnosis more precise

Exhaled breath could help differentiate pneumonia from other acute cardiorespiratory diseases, thereby expediting diagnosis and appropriate therapies. For example, a study of ventilator-dependent adults admitted within 72 hours to the intensive care unit (ICU) identified significantly elevated exhaled and nasal NO levels in mechanically ventilated patients with pneumonia as compared with mechanically ventilated patients without pneumonia [15]. In another study of adults hospitalized with acute cardiorespiratory breathlessness, a multibiomarker score (101 breath biomarkers) differentiated exacerbation subtypes (acute heart failure, acute asthma, acute chronic obstructive pulmonary disease, and CAP) with moderate accuracy overall (AUC 0.72) [66]. The diagnostic accuracy for CAP was 0.79; this improved to 0.91 after adjusting for smoking status, time to sample acquisition, modified early warning score (mEWS-2), and recent antibiotics or steroids prior to admission [66]. These findings were not internally replicated due to the small sample sizes within subgroups.

The potential for exhaled breath to expedite diagnosis is well-recognized in other acute diseases. For example, in ARDS, a known but poorly understood complication of severe pneumonia, exhaled breath monitoring distinguished ARDS and non-ARDS with high accuracy (AUC 0.87) [67]. Longitudinal breath analysis also identified volatile metabolic changes consistent with ARDS an average of three hours earlier than clinical adjudication in an animal model [68]. Although not specific to pneumonia, these findings underscore the utility of breath profiling as a noninvasive tool to support diagnostic decision-making in the clinical setting.

6. Understanding clinical trajectory

In addition to diagnostic information, breath metabolites may provide noninvasive, continuous monitoring of the host response to respiratory infections and clinical interventions. In a longitudinal analysis of patients with severe COVID-19, five patients with divergent clinical outcomes were monitored for up to 10 days [69]. Although merely hypothesis generating given the small sample size, the patients demonstrated a range of severity with corresponding differences in metabolite profiles [69]. Similarly, a time series analysis of breath profiles characterized the clinical trajectory of nine ARDS patients and nine non-ARDS patients, with breath signatures able to characterize clinical worsening as well as ARDS resolution [67].

Preliminary studies in CAP suggest that measures of oxidative stress, inflammation, and immune response could be useful in monitoring clinical trajectory and treatment response. For example, exhaled NO levels in nine nonsmoking patients with CAP decreased after treatment and did not differ significantly after recovery from those of controls [16]. Amongst 43 adults admitted with CAP and serially monitored for 10 days, H₂O₂ and TBAR concentrations decreased with successful treatment [44]. Hepatocyte growth factor (HGF), a protein produced by mesenchymal cells in many organs and important in tissue repair and immune regulation, has been observed to increase in the acute phase of pneumonia and to subsequently decrease in serum but not in EBC after recovery [70]. The findings, although based on a sample size of only 10 patients with pneumonia, might be interpreted to mean that EBC provides an easily accessible window into ongoing lung repair and healing that is not replicated systemically.

7. Phenotyping patients with pneumonia

Exhaled breath could ultimately help phenotype pneumonia and thereby begin to elucidate the biological mechanisms underlying susceptibility to infection and differential clinical outcomes. Prior studies have applied exhaled breath analysis to characterize airway inflammation [71,72]. Recognizing the potential value of exhaled breath in molecular phenotyping, several studies in obstructive lung disease have associated breath profiles with airway inflammatory cell type and steroid responsiveness [73–77]. VOC profiles also predicted asthma exacerbations [78]. The findings suggest that breath analysis might offer a unique, quantitative parameter of lung inflammation and provide additional insights into inflammatory response pathways contributing to divergent outcomes in pneumonia. Such investigations would likely require large sample sizes to adjust for underlying comorbid conditions and associated inflammatory pathways.

Protein biomarkers isolated from EBC on presentation could potentially discriminate between pneumonia subtypes to predict prognosis. Cytokines obtained daily for five days in EBC from 49 mechanically ventilated patients correlated with disease severity, pneumonia, sepsis, and death [79]. In a study of adults hospitalized with CAP, those with severe CAP on presentation had lower levels of anti-inflammatory cytokines in the EBC and higher levels of pro-inflammatory cytokines in serum obtained within 24 hours of admission [80]. Notably, the authors did not obtain bronchoalveolar lavage samples for comparison. These findings suggest that CAP patients with severe outcomes suffer from dysregulated pro- and anti-inflammatory immune pathways not fully captured in the systemic circulation. Once further validated, EBC cytokine measurement could provide valuable insights into the pathobiology of severe pneumonia. These insights could complement current approaches, primarily focused on ARDS and sepsis, which have identified hyper- and hypo-inflammatory subtypes associated with clinical severity and response to treatment [81,82].

8. Understanding lung immunity and risk of pneumonia

In addition to revealing what is happening in the infected lung and patient, we anticipate that breath analyses may usefully report an individual’s pneumonia risk. Pneumonia is an acute event that depends on underlying risk factors [6,83], similar to a heart attack or stroke. The discovery that simply measuring someone’s serum cholesterol and blood pressure could help gauge their risk for future infarctions [84] had wide-ranging effects on biological understanding, medical care, and public health practice that led to a two-thirds reduction in deaths due to heart attack or stroke [85,86]. Gauging someone’s risk for future pneumonia by evaluating the state of defenses against pneumonia, such as the quantities and qualities of diverse types of macrophages, epithelial cells, and resident memory lymphocytes in the lung [87], has great promise for moving the needle against this disease.

Although yet to be tested, breath analyses should be informative about pneumonia risk. People who would be predicted to have elevated risk for pneumonia due to co-morbidities [88], advancing age [89], or exposures [90] exhale demonstrably different metabolites compared to those without such risk factors. Things that decrease pneumonia risk, such as influenza vaccination, can be sufficient to shift the metabolite profile exhaled by healthy people [42,91]. Different cell-types produce different VOC profiles, including T lymphocytes [92] whose numbers and biology in the lung are key determinants of pneumonia prevention [87]. Lung epithelial cells help defend against pneumonia [87], and VOC profiles can distinguish among lung epithelial cells and report the responses of these cells to pneumonia risk-elevating exposures like pollutants [90]. In our opinion, it should be a community priority to determine whether and how metabolites in exhaled breath may be used to report on the lung biology and immune status that combine to determine an individual’s risk for pneumonia.

9. Safety concerns during breath collection

Collection of exhaled breath samples is considered safe, with low risk to the patient and operator. Unlike bronchoscopy, it does not cause changes in ventilation, oxygenation, or hemodynamics; it also does not worsen bronchospasm or lung hyperinflation in patients experiencing exacerbations of obstructive lung disease [93,94]. The pattern of breathing is normal, making it safer than spirometry, although some spontaneously breathing patients have been noted to hyperventilate at the beginning of sample collection [27,93]. In patients on mechanical ventilation, transient worsening of hypoxemia can occur when breaking the in-line circuit during collection device exchanges. When using the exhaust port, device exchange can cause obstruction resulting in apnea, barotrauma, volutrauma, or pneumothorax [27]. Individuals obtaining these samples should receive appropriate training to minimize associated risks.

10. Limitations and future directions

Despite the tremendous potential of exhaled breath, the pace of breath biomarker research in pneumonia is relatively slow. This is likely due in part to technological limitations and lack of standardization in breath collection methods, making reproducibility a challenge [93,95–97]. Environmental conditions, patient characteristics (e.g. age, sex, pregnancy), diet, and smoking are some known potential confounders [93,95]. EBC is mainly formed from water vapor, which reduces the concentration of measurable lung proteins, thereby adding an additional layer of complexity to measurement [93,95]. Standardization of collection procedures and technological developments to lower the levels of detection will advance the field of breathomics research and its integration into clinical practice [35,93,98,99]. This will require inter-institutional collaboration and engagement among stakeholders to agree upon standardized protocols for breath collection, analysis, and dissemination of results.

Small sample sizes limit the impact and interpretability of studies to date. For example, exhaled breath profiles in interstitial lung disease (ILD), including sarcoidosis, differ from those in pneumonia [100]. However, there are some shared biomarkers, suggesting overlap in infectious and inflammatory pathways and potentially reducing diagnostic accuracy in patients with multiple conditions. Larger studies with more diverse patient populations are needed to evaluate and account for potential common infectious and inflammatory pathways. Ideally, once the methods for sampling and analysis of breath samples are standardized and validated, breath studies should include large sample sizes across multiple centers.

11. Conclusions

Pneumonia is a common and potentially deadly condition that can be difficult to diagnose. Exhaled breath analysis is a promising field of research that could potentially improve early diagnosis and risk stratification. The collection procedure is simple and noninvasive and can be performed serially for sequential and longitudinal sampling without more invasive testing. Breath analysis is an exciting area of research, although far from ready for clinical implementation given lack of consistency in study methodology. Once further validated and standardized, breath biomarkers could improve our understanding of the biological mechanisms underlying differential clinical outcomes in pneumonia and thereby facilitate targeted therapies.

12. Expert opinion

Breath-based metabolomics offer the potential to improve the recognition and management of pneumonia. Breath analysis is a noninvasive, potentially continuous method of measurement that provides a direct window into the lung. Current studies that have investigated both serum and breath biomarkers report different metabolomic profiles, but this is to be expected given the sources. Given the importance of the host immune response, breath analysis might also help us understand a person’s risk of developing pneumonia. Different immune cell types produce different breath profiles. Knowing the immunologic landscape of a person’s lung has the potential to identify those most at risk and target them for interventions, much in the way that we approach prevention in cardiovascular disease.

In addition to advancing our understanding of the mechanisms underlying the pathobiology of pneumonia, preliminary studies suggest that breath profiles might help differentiate pneumonia from other acute respiratory illnesses and correlate with clinical improvement. In this way, breath testing could help target appropriate therapies, inform decisions regarding duration of antibiotics, and more efficiently allocate healthcare resources.

Moreover, breath sample collection is very safe. Unlike bronchoscopy, it is noninvasive. It can be performed continuously and does not carry associated risks of altering ventilatory or oxygenation parameters or worsening bronchospasm or obstruction. These are distinct advantages in the management of critically ill patients.

Despite its promise, exhaled breath analysis remains experimental, and much work remains prior to real-time clinical application. In our opinion, large observational studies in cohorts of patients at increased risk of pneumonia in the subsequent years are needed. Small sample size studies have limited the scientific and clinical impact of studies to date. Small sample size prohibits sufficient adjustment for underlying comorbidities, particularly concomitant lung or inflammatory conditions, that likely contribute to an individual’s breath profile. Ideally, large, multi-site studies are needed to account for underlying comorbid conditions and enhance the accuracy of breath signatures.

The lack of methodologic standardization in breath sample collection and analysis poses an additional barrier to clinical implementation. Along with small sample sizes, this limitation presents challenges in the comparison or validation of results across studies. It also impedes the performance of large, multi-site studies necessary to advance the field. Before breath research can translate to the bedside, investigators must collaborate and agree upon standardized methodology. Once in place, uniform methods will enhance the interpretability of findings, facilitate the replication and validation of study results, and encourage scientific partnerships across institutions.

Collectively, research in breath analysis suggests that exhaled breath has potential to afford novel insights into the underlying biology of pneumonia and the host response. Future studies should consider integrating exhaled breath analysis with other available risk scores and biomarkers. Large, multicenter projects that incorporate breath biomarkers with other available diagnostic and prognostic markers (e.g. labs, clinical prediction rules) can evaluate the comparative and additive performance of these approaches. Ultimately, these studies should precede the adoption of exhaled breath analysis at the bedside.

Article highlights.

• A major cause of morbidity and mortality globally, pneumonia is highly heterogeneous in etiology, clinical presentation, and severity.

• Breath-based metabolomics is a potentially continuous, noninvasive approach that could improve pneumonia diagnosis and risk stratification.

• While there are widely adopted protocols, there is not yet a universal standardized approach for breath collection, and current data is based on modest sample sizes.

• Once further validated and standardized, exhaled breath analysis, which provides a direct real-time window into the lung, has great potential to elucidate the pathobiology of pneumonia.