Saliva, Sweat, and Exhaled Breath as Alternative Specimens for Exercise Chemistry: Analytical Advantages, Limitations, and Comparative Progress

ABSTRACT

This critical narrative review evaluates saliva, sweat, and exhaled breath as noninvasive interface matrices for exercise chemistry rather than as general substitutes for blood or urine. It synthesizes peer‐reviewed evidence on exercise‐related sampling biology, analytical platforms, quantitative performance, validation limits, and translational readiness; because protocols and platforms remain heterogeneous, the evidence is interpreted qualitatively rather than pooled meta‐analytically. Saliva is strongest for repeated neuroendocrine, autonomic, immune, oxidative, and short‐latency metabolite measurements, but its interpretation depends on flow rate, timing, oral contamination, and normalization. Sweat is most mature for sweat rate and sodium/chloride loss monitoring and is uniquely compatible with wearable platforms, whereas sweat lactate, glucose, cortisol, and cytokines require stronger physiological validation before they can be treated as systemic markers. Exhaled breath offers second‐scale access to volatile metabolism, substrate use, and airway responses, but breathomics remains instrumentally demanding and vulnerable to ventilation, diet, humidity, and ambient‐air confounding. The review concludes that useful applications should match each analyte and platform to a defined exercise question, a matrix‐specific validation model, and a practical decision context.

1. Introduction

Exercise perturbs almost every major physiological system, but the chemistry of that perturbation is often sampled too sparsely to capture its true dynamics (Jason Heikenfeld et al. 2019). Classical blood measurements remain the reference for many endocrine, metabolic, inflammatory, and hydration‐related questions, and urine is valuable for cumulative exposure or recovery‐oriented outputs, yet both are imperfect when investigators need minute‐to‐minute resolution during training, competition, or environmental stress (J. Heikenfeld et al. 2018; Malon et al. 2014). Repeated venous sampling can alter behavior, is logistically difficult in field settings, and usually narrows study designs to small cohorts or tightly controlled laboratories (Amann et al. 2014). Urine is easy to collect but inherently integrates over longer time windows and is therefore poorly aligned with rapid physiological transitions such as acute autonomic activation, the lactate turn point, or early recovery (De Lacy Costello et al. 2014). These constraints explain the renewed interest in alternative specimens such as saliva, sweat, and exhaled breath, which can often be obtained repeatedly with minimal interruption to the athlete and minimal need for medical infrastructure (Heaney et al. 2019).

The appeal of these matrices lies in more than convenience. Saliva can report hypothalamic–pituitary–adrenal activation, sympathetic drive, mucosal immunity, oxidative balance, and selected metabolomic shifts with sampling intervals far shorter than are practical for venous testing (Hayes et al. 2015; Lindsey et al. 2025; Neves et al. 2023; Ntovas et al. 2022). Sweat can be accessed continuously through patches, microfluidic collectors, or integrated wearable devices, which has made it the flagship matrix for real‐time sports biosensing (Baker and Wolfe 2020; Emaminejad et al. 2017; W. Gao et al. 2016; Jalal et al. 2024; Lin et al. 2022; Ma et al. 2024; Pirovano et al. 2020; Van Hoovels et al. 2021; Xuan et al. 2023; Yang et al. 2024; Zhu et al. 2025). Exhaled breath offers perhaps the highest temporal fidelity of all because volatile metabolites and gases can change on the order of single breaths or seconds during graded exercise, substrate transitions, altered oxygen availability, or recovery (Ager et al. 2020; Bikov et al. 2011; Heaney et al. 2022; Henderson et al. 2021; Kippelen et al. 2002; Osswald et al. 2021; Sheel et al. 1999; Stang et al. 2015; Tondo et al. 2024; Verges et al. 2005). This broader move toward peripheral and wearable biochemical monitoring is also reinforced by general analytical‐chemistry assessments of body‐biofluid access and sensor development (Ghaffari et al. 2021; J. Kim et al. 2019; Rao et al. 2026; Sempionatto et al. 2022). The analytical message is therefore tempting: If blood is burdensome and these matrices are easy, why not substitute them?

That substitution logic is precisely where the field still struggles. Alternative specimens are not merely diluted versions of blood. They are secretions or excretions shaped by organ‐specific transport, local metabolism, evaporation, bacterial contributions, ambient contamination, and matrix‐dependent breakdown. Saliva contains systemic constituents, but also metabolites derived from oral microbiota, gingival fluid, food residues, and changes in salivary flow (Malon et al. 2014; Neves et al. 2023; Ntovas et al. 2022; Pitti et al. 2019). Sweat begins as a secretory‐coil ultrafiltrate of extracellular fluid, yet its final composition is reworked by ductal ion reabsorption, sweat‐gland metabolism, regional sweat rate, skin surface contamination, and environmental exposure (Baker 2017; Baker and Wolfe 2020; F. Gao et al. 2023; Jason Heikenfeld et al. 2019). Exhaled breath reflects pulmonary gas exchange and volatile metabolism, but measured concentrations depend strongly on alveolar ventilation, humidity, dead‐space handling, ambient air composition, and the analytical platform used for acquisition (Ager et al. 2020; Amann et al. 2014; Bikov et al. 2011; De Lacy Costello et al. 2014; Kippelen et al. 2002; Osswald et al. 2021; Sheel et al. 1999; Stang et al. 2015; Verges et al. 2005). Consequently, convenience without physiological qualification can generate elegant measurements that answer the wrong question.

Exercise chemistry is therefore the right conceptual frame for comparing these matrices. Rather than asking whether saliva, sweat, or breath can replace blood in a general sense, exercise chemistry asks a narrower and more useful question: which specimen can report a defined analyte class, at a defined time scale, with sufficient analytical specificity and biological interpretability to support a decision? That decision may be mechanistic, such as linking substrate switching to breath acetone kinetics; operational, such as flagging high sweat sodium losses in athletes with repeated cramping; or recovery‐oriented, such as identifying whether salivary cortisol, alpha‐amylase, or immunoglobulin A are responding to load, sleep loss, or environmental strain (Baker 2017; Baker and Wolfe 2020; Hayes et al. 2015; M. J. Kim et al. 2020; Lindsey et al. 2025; Neves et al. 2023; Schouten et al. 1988; Stang et al. 2015).

Recent work has advanced the field in three major ways. First, systematic reviews and meta‐analyses have clarified that exercise responses in saliva are analyte‐specific and heavily time‐dependent rather than uniformly directional (Hayes et al. 2015; Lindsey et al. 2025; Neves et al. 2023). Second, sweat sensing has moved beyond proof‐of‐concept single‐analyte devices toward integrated multiplexed systems with temperature compensation, wireless readout, and sport‐oriented form factors (Emaminejad et al. 2017; F. Gao et al. 2023; W. Gao et al. 2016; Jalal et al. 2024; Lin et al. 2022; Ma et al. 2024; Pirovano et al. 2020; Van Hoovels et al. 2021; Xu et al. 2021; Xuan et al. 2023; Yang et al. 2024; Zhu et al. 2025). Third, breathomics has shifted from a few canonical volatiles toward pathway‐level interpretation using real‐time secondary electrospray ionization high‐resolution mass spectrometry, proton‐transfer‐reaction mass spectrometry, and pattern‐recognition methods such as electronic noses (Ager et al. 2020; Bikov et al. 2011; Dickinson et al. 2023; Garzinsky et al. 2021; Heaney et al. 2022, 2016; Henderson et al. 2021; Hori et al. 2020; M. J. Kim et al. 2020; King et al. 2009; Kippelen et al. 2002; Königstein et al. 2021; Mochalski et al. 2023; Osswald et al. 2021; Ota et al. 2020; Sheel et al. 1999; Stang et al. 2015; Tondo et al. 2024; Turner et al. 2011; Verges et al. 2005). Yet these advances have not eliminated core translational problems. Studies are still small, sampling protocols remain heterogeneous, cross‐matrix correlations are often assumed rather than proven, and many wearable or real‐time systems are analytically impressive but physiologically under‐validated (Baker and Wolfe 2020; F. Gao et al. 2023; Jason Heikenfeld et al. 2019; Osswald et al. 2021; Van Hoovels et al. 2021).

The present review therefore takes a critical comparative approach. It does not argue that alternative specimens should displace blood or urine, nor does it dismiss them as overly noisy. Instead, it evaluates where saliva, sweat, and exhaled breath are analytically advantageous, where they remain fragile, and how they should be matched to specific exercise questions. In this article, “decision‐ready” means that the specimen, analyte, sampling window, platform, and validation model are aligned well enough to support a defined mechanistic, monitoring, or field‐practice question. Figure 1 summarizes this logic as a qualitative framework, not as a ranked evidence score: Saliva is closest to practical repeated sampling, sweat to continuous wearable access, and breath to high temporal fidelity, whereas all three carry different burdens in biological processing, standardization, and physiological interpretation. This article is a critical narrative review with a scoping‐oriented comparative structure, not a formal systematic review or meta‐analysis. The evidence base is bounded to peer‐reviewed and authoritative sources that directly inform exercise‐related saliva, sweat, or exhaled‐breath chemistry, including sampling biology, analytical methods, quantitative exercise responses, validation against physiological endpoints, and field or wearable deployment. Evidence was prioritized when a study reported matrix‐specific controls, exercise protocols, quantitative outcomes, or clear validation conditions. Disease‐only diagnostic studies, generic biofluid‐sensing papers without exercise relevance, and purely engineering reports without matrix interpretation were used only when they clarified analytical principles. Because sampling sites, exercise protocols, platforms, environmental conditions, and normalization procedures differ widely, cross‐study values are treated as representative examples, not pooled effect estimates.

FIGURE 1.

Conceptual framework of saliva, sweat, and exhaled breath as alternative specimens for exercise chemistry. The dimensions are qualitative synthesis axes derived from matrix biology and validation constraints discussed in Sections 2, 6; they indicate accessibility, matrix‐specific analytical scope, temporal resolution, and translational barriers rather than pooled quantitative scores.

In this review, analytical validity refers to whether a method measures the target reliably in the matrix, physiological validity refers to whether the matrix signal answers the intended exercise question, and practical readiness refers to whether collection, calibration, timing, and interpretation can survive field conditions. Cross‐matrix comparison is therefore based on four shared dimensions: accessibility, temporal resolution, matrix‐specific biological processing, and validation burden.

2. Biological Basis and Sampling Considerations of Alternative Exercise Specimens

2.1. Interface Matrices Are Biologically Active Matrices

The decisive difference between blood and the three alternative matrices considered here is that blood is a transported internal compartment, whereas saliva, sweat, and breath are interface matrices. They are produced at boundaries: oral, cutaneous, and pulmonary. Their chemistry therefore reflects both internal physiology and boundary‐specific processing. This distinction explains why identical analytes can behave differently in different matrices even when sampled at the same exercise time point.

Saliva is secreted mainly by the parotid, submandibular, sublingual, and minor salivary glands. Exercise changes autonomic tone, which alters flow rate and the relative contribution of these glands. A slower flow rate concentrates some constituents, whereas sympathetic stimulation can raise protein‐rich secretion and promote faster shifts in analytes such as alpha‐amylase. In addition, oral bacteria, epithelial turnover, gingival fluid, mouth breathing, and recent food or drink intake all contribute nontrivial variance (Malon et al. 2014; Neves et al. 2023; Ntovas et al. 2022; Pitti et al. 2019). A salivary biomarker is therefore rarely interpretable without knowing whether the sample was unstimulated or stimulated, whether the athlete had eaten, and how soon after exercise the specimen was collected.

Primary eccrine sweat is formed in the secretory coil from extracellular fluid, but the final sweat reaching the skin surface is reshaped as it passes through the duct. Sodium and chloride are reabsorbed to varying degrees, whereas several other analytes may be secreted, metabolized in the gland, or concentrated by water removal and evaporation. According to Baker and Wolfe, whole‐body sweating rates across activities and environments can range roughly from 0.2 to 3.0 L/h, and athletes with “salty sweat” often exhibit sodium or chloride concentrations at or above 70–80 mmol/L (Baker and Wolfe 2020). The same nominal sweat sodium concentration can therefore reflect different secretory states depending on flow, acclimation status, body site, and collection method.

Exhaled breath lies at the opposite end of the temporal spectrum. Its appeal comes from fast exchange between blood and the gas phase for volatile analytes, but measured values depend on ventilation pattern, dead‐space management, ambient air subtraction, airway humidity, and the partition behavior of each compound. Highly soluble compounds behave differently from poorly soluble ones, and breath concentrations can change because of altered pulmonary mechanics even when systemic production is unchanged (Ager et al. 2020; Amann et al. 2014; De Lacy Costello et al. 2014; Osswald et al. 2021; Verges et al. 2005). Exercise amplifies all of these issues by changing tidal volume, breathing frequency, perfusion, and the relative contribution of alveolar versus airway compartments.

2.2. Sampling Timing, Normalization, and Contamination

Pre‐analytical variation often dominates the small effect sizes investigators hope to measure. In saliva, timing after exercise cessation is especially important. The salivary metabolome can normalize faster than blood. Pitti et al. showed a mean 3.3‐fold increase in salivary lactate after an official soccer match, but also emphasized that salivary effects may decay much faster than blood‐based analogs and are highly sensitive to normalization strategy (Pitti et al. 2019). Similar issues apply to cortisol, testosterone, and cytokines, especially when circadian rhythm or menstrual‐cycle phase are not controlled (Cullen et al. 2015; Geisler et al. 2019; González et al. 2008; Hayes et al. 2015; Lindsey et al. 2025; Minetto et al. 2005; Neves et al. 2023).

Flow correction is another major source of disagreement. Concentrations can be normalized to volume, total protein, osmolality, secretion rate, or total observed metabolite content; each choice answers a different biochemical question. Pitti et al. showed that lactate, pyruvate, glucose, galactose, and succinate changed differently depending on whether salivary data were normalized to total protein or total metabolite output (Pitti et al. 2019). Thus, apparently inconsistent literature may sometimes reflect incompatible normalization rather than contradictory biology.

Sweat pre‐analytics are equally demanding. Regional collectors do not necessarily represent whole‐body losses; body site, exercise intensity, local gland density, and evaporation all matter (Baker 2017; Baker et al. 2016; Baker and Wolfe 2020; Barnes et al. 2019). Skin cleaning, adhesive occlusion, collector dead volume, sampling duration, and storage temperature can alter the final measurement. Harshman et al. showed that amino‐acid signatures in exercise‐induced sweat were largely stable for up to 90 min, with more than 88% of untargeted metabolomic data points remaining within acceptable limits under controlled holding conditions (Harshman et al. 2019). This does not erase upstream variability from collection or gland recruitment.

Breath sampling has its own normalization problem. Some investigators use end‐tidal or carbon‐dioxide‐triggered sampling, others collect fixed volumes into bags or onto sorbent tubes, and still others perform direct online mass spectrometric acquisition breath‐by‐breath (Ager et al. 2020; Bikov et al. 2011; Kippelen et al. 2002; Königstein et al. 2021; Osswald et al. 2021; Sheel et al. 1999; Stang et al. 2015; Verges et al. 2005). The choice affects sensitivity to dead space, condensation, and within‐breath variability. Henderson et al. improved alveolar targeting in a field deployment by triggering collection when exhaled CO₂ reached 4% (Henderson et al. 2021). Even so, unrestricted food intake, outdoor background air, and medication exposure remained influential.

2.3. Analytical Progress Has Exceeded Biological Validation

The analytical landscape is more mature than the physiological one. Saliva can be measured using established immunoassays, enzymatic assays, liquid chromatography, and metabolomic platforms including nuclear magnetic resonance and mass spectrometry (Farjallah et al. 2018; Hayes et al. 2015; Kojima et al. 2025; Lindsey et al. 2025; Neves et al. 2023; Ntovas et al. 2022; Pitti et al. 2019; Yan et al. 2023; Yi et al. 2025). Sweat sensing has progressed rapidly through potentiometric ion‐selective electrodes, amperometric enzyme electrodes, soft microfluidics, and integrated wearable electronics (Emaminejad et al. 2017; F. Gao et al. 2023; W. Gao et al. 2016; Jalal et al. 2024; Lin et al. 2022; Ma et al. 2024; Pirovano et al. 2020; Sonner et al. 2015; Van Hoovels et al. 2021; Xu et al. 2021; Xuan et al. 2023; Yang et al. 2024; Zhu et al. 2025). Breath chemistry can be interrogated by gas chromatography–mass spectrometry, thermal desorption methods, proton‐transfer‐reaction mass spectrometry, secondary electrospray ionization, and electronic‐nose pattern recognition (Ager et al. 2020; Bikov et al. 2011; Dickinson et al. 2023; Garzinsky et al. 2021; Heaney et al. 2022, 2016; Henderson et al. 2021; Hori et al. 2020; M. J. Kim et al. 2020; King et al. 2009; Kippelen et al. 2002; Königstein et al. 2021; Mochalski et al. 2023; Osswald et al. 2021; Ota et al. 2020; Sheel et al. 1999; Stang et al. 2015; Tondo et al. 2024; Turner et al. 2011; Verges et al. 2005). The field no longer lacks instruments.

What it still lacks is alignment between analytical performance and physiological validation. A sensor may show excellent linear range in artificial sweat yet fail to maintain calibration on skin. A real‐time breath instrument may capture second‐scale kinetics but still not distinguish whether the change is metabolic, ventilatory, or environmental. An immunoassay for salivary cortisol may be precise analytically but biologically misleading if collection is not temporally standardized. For exercise chemistry, analytical excellence is necessary but insufficient; a useful method must also survive the matrix‐specific confounders introduced by motion, sweating, ventilation, dehydration, recovery kinetics, and field logistics.

A useful way to avoid this error is to separate analytical validity from physiological validity. Analytical validity asks whether the assay or sensor measures the target accurately, precisely, and stably in saliva, sweat, or breath. Physiological validity asks whether the observed matrix signal has a defensible relationship to the exercise process being inferred. A sweat lactate trace can therefore be analytically valid as a local sweat‐gland signal while remaining physiologically unsuitable as a direct proxy for blood lactate; conversely, sweat sodium can be physiologically useful because it measures the composition of the loss pathway itself rather than a plasma surrogate.

3. Saliva in Exercise Chemistry

3.1. Hormonal and Autonomic Markers in Saliva

Saliva is the most established alternative specimen in sport science because it can be collected repeatedly without cannulation and because several classical stress biomarkers diffuse or are secreted into oral fluid in measurable form. Cortisol and testosterone remain the dominant pair, especially when investigators seek to estimate catabolic‐anabolic balance or recovery strain. Hayes et al. concluded in their meta‐analysis that salivary testosterone, cortisol, and their ratio respond to exercise, but the magnitude and direction depend strongly on exercise mode, intensity, duration, and recovery timing (Hayes et al. 2015). This is exactly why overly simplified interpretations, such as “high cortisol equals poor recovery,” are analytically weak.

Direct experimental work reinforces the point. Backes et al. reported that salivary alpha‐amylase increased from 85 ± 10 U/mL pre‐exercise to 284 ± 30 U/mL at high intensity, whereas salivary cortisol rose from 0.30 ± 0.03 to 0.45 ± 0.05 ug/dL in exercising athletes (Backes et al. 2015). Yet the same study showed that fluid condition altered moderate‐intensity cortisol, reminding us that hydration and oral fluid status can modulate the signal. Repeated bouts of downhill running also increased salivary cortisol and alpha‐amylase over recovery, but the kinetics differed and were not reducible to a single “stress” construct (Mckune et al. 2014). In taekwondo competition, salivary cortisol and alpha‐amylase aligned with anticipatory and match‐related stress, but the context was psychophysiological competition load rather than pure metabolic strain (Capranica et al. 2017).

Alpha‐amylase is especially attractive as a field marker of sympathetic activation because it often responds faster than cortisol. A single hard RPE‐based cycling session increased salivary alpha‐amylase in healthy adults (Weiss et al. 2019). However, Leicht et al. showed in elite wheelchair athletes that alpha‐amylase reflected sympathetic activity, whereas chromogranin A did not track the response equivalently (Leicht et al. 2017). Salivary autonomic markers should therefore be considered marker‐specific rather than interchangeable proxies for the same physiology.

Salivary testosterone deserves similar caution. Acute resistance exercise can elevate testosterone transiently, but the amplitude depends on muscle mass recruited, rest intervals, volume load, and sex. Geisler et al. found that lower versus upper body resistance exercise elicited different salivary testosterone and cortisol responses, highlighting the influence of exercise structure rather than simply “intensity” in the abstract (Geisler et al. 2019). For practical load monitoring, the testosterone:cortisol ratio is often used because it compresses dual endocrine information into a single index, but it is mathematically unstable when both numerator and denominator are changing on different time scales.

3.2. Immune, Inflammatory, and Oxidative Markers

Secretory immunoglobulin A remains one of the longest‐studied salivary exercise markers because it links mucosal immunity to heavy training, illness risk, and recovery. Yet classic work by McDowell et al. showed no significant sIgA change after 15–45 min runs at 50%–80% VO2max (S. L. McDowell et al. 1991), and training intervention data also suggest that maximal‐exercise effects on sIgA and cortisol can shift with conditioning status (S. McDowell et al. 1992). These findings are not a failure of the marker; rather, they reveal that moderate acute sessions do not necessarily perturb mucosal immunity in a uniform direction. By contrast, prolonged heavy loading, energy deficiency, cold exposure, and clustered competition schedules can reduce sIgA secretion rate enough to matter operationally (Lindsey et al. 2025; Neves et al. 2023; Schouten et al. 1988).

Inflammatory saliva markers are even more complex. Interleukin‐6 rises systemically with exercise because contracting muscle is a major source, but salivary IL‐6 does not mirror plasma quantitatively. Minetto et al. described differential responses of serum and salivary interleukin‐6 to acute strenuous exercise (Minetto et al. 2005), whereas Cullen et al. found only limited agreement between salivary, venous, and capillary IL‐6 at rest and in response to exercise (Cullen et al. 2015). This does not make salivary IL‐6 useless. It means that it should be interpreted as an oral‐fluid inflammatory signal shaped by local transfer processes, not as a drop‐in replacement for plasma cytokine concentration.

Oxidative and antioxidant markers may be more promising because saliva is directly accessible for redox‐oriented assays. González et al. reported exercise effects on salivary uric acid, total antioxidant activity, oxidative stress, and nitric oxide products, supporting the sensitivity of saliva to redox balance during aerobic exercise (González et al. 2008). In applied sport settings, such measures are attractive because they combine low‐cost assays with rapid repeated sampling. Still, oral bacteria, diet, nitrate intake, and recent tooth brushing can all influence nitrite‐ and nitric‐oxide‐related readouts.

3.3. Salivary Metabolomics: Progress and Current Limitations

The strongest recent advance in salivary exercise chemistry is metabolomics. Rather than targeting one hormone or enzyme at a time, metabolomic studies allow simultaneous monitoring of dozens of low‐molecular‐weight molecules related to glycolysis, amino‐acid flux, hydration, and stress responses. Pitti et al. analyzed saliva from soccer players and identified exercise‐linked changes in 56 metabolites by NMR, while showing that interpretation changed materially depending on whether concentrations were normalized to total proteins or total metabolite output (Pitti et al. 2019). Their dataset is especially important because it demonstrated a mean 3.3‐fold increase in salivary lactate after match play in non‐normalized samples, but not all metabolites behaved similarly once normalization was applied.

Recent studies extend this idea. Salivary metabolite profiling after high‐intensity rowing has been proposed as a route to screening exercise‐induced muscle damage, whereas a 2026 acute physical stress study reported that salivary metabolomic changes occur extremely rapidly, implying that the window of biological detectability may sometimes be measured in minutes rather than hours (Yi et al. 2025). These are exciting findings, but they also illustrate why saliva can be analytically unforgiving. A biomarker that changes quickly can support dense monitoring, yet it can also be missed entirely if collection is not synchronized to the physiological event. Figure 2 sharpens this point by showing that lactate remains one of the most visibly exercise‐responsive metabolites after match play, whereas succinate, pyruvate, glucose, and galactose can change in magnitude or even rank order when the denominator shifts from total protein to total metabolite output. The figure is therefore not simply illustrative; it demonstrates that part of the apparent biology of salivary metabolomics is actually normalization architecture.

FIGURE 2.

Pathway‐oriented synthesis of real‐time breath metabolomics during graded exercise, based on the exercise‐responsive metabolic classes reported by Osswald et al. (2021). The figure organizes glyoxylate and dicarboxylate metabolism, TCA‐cycle‐related signals, tryptophan‐related features, glycolytic, lipid, and ketone‐linked responses for qualitative interpretation and does not reproduce original intensity plots.

The emerging role of salivary lactate is instructive. Salivary lactate clearly increases after high‐intensity exercise, and active recovery appears to accelerate its clearance, but the timing differs from blood and the concentration is sensitive to salivary secretion, mouth contamination, and the short persistence of the oral signal (Kojima et al. 2025; Yan et al. 2023). Salivary lactate is therefore best viewed as a fast oral‐fluid marker of glycolytic stress rather than a simple noninvasive replacement for capillary blood lactate.

The translational implication is positive but modest. Saliva is not “blood without needles.” It is an analytically rich interface matrix that becomes especially powerful when its local biology is accepted rather than ignored, and this is also why Table 1 places flow rate and correction variables beside hormones, immune markers, and metabolites instead of treating them as peripheral method notes. Read comparatively, the table shows that the largest practical salivary signals, such as alpha‐amylase surges, lactate excursions, or secretion‐rate shifts in mucosal immunity, are also among the most sensitive to hydration status, oral contamination, circadian timing, and normalization choice.

TABLE 1.

Quantitative comparison of representative salivary targets used in exercise chemistry. Values are representative, protocol‐dependent examples from the cited literature rather than pooled effect estimates.

Salivary target	Representative quantitative exercise response	Common analytical platform	Practical readiness/main use case	Main interpretive limitation	Representative refs.
Flow rate	Often falls from about 0.4–0.6 to 0.2–0.4 mL/min during dehydrating or high‐intensity exercise	Gravimetry; timed collection	Required for secretion‐rate normalization	Strongly altered by hydration, mouth breathing, stimulation	Lindsey et al. (2025); Neves et al. (2023); Ntovas et al. (2022); Pitti et al. (2019)
Total protein or osmolality	Commonly rises about 1.2–2.0‐fold after hard exercise or match play	Colorimetry; osmometers	Corrects concentration changes driven by water loss	Not interchangeable with metabolite‐output normalization	Lindsey et al. (2025); Neves et al. (2023); Ntovas et al. (2022); Pitti et al. (2019)
Cortisol	0.30 ± 0.03 to 0.45 ± 0.05 μg/dL from pre‐ to high‐intensity exercise in athletes	Immunoassay	Practical marker of HPA‐axis load	Circadian rhythm and hydration confound acute interpretation	Backes et al. (2015); Geisler et al. (2019); Hayes et al. (2015); Mckune et al. (2014)
Testosterone	Acute increases of about 10–40% after some resistance protocols, but null effects are common	Immunoassay	Useful when paired with exercise context and recovery timing	Sex, time of day, and exercise mode strongly modulate response	Geisler et al. (2019); Hayes et al. (2015)
Testosterone: cortisol ratio	Often decreases about 15–30% after strenuous or poorly recovered sessions	Derived index	Screens anabolic‐catabolic balance	Ratio can be unstable when both hormones move on different time scales	Geisler et al. (2019); Hayes et al. (2015); Lindsey et al. (2025); Neves et al. (2023)
Alpha‐amylase	85 ± 10 to 284 ± 30 U/mL from rest to high intensity	Enzymatic assay	Fast salivary marker of autonomic activation	Sensitive to oral flow, oral stimulation, and psychological arousal	Backes et al. (2015); Capranica et al. (2017); Leicht et al. (2017); Mckune et al. (2014); Weiss et al. (2019)
Chromogranin A	Less responsive than alpha‐amylase in elite wheelchair athletes	Immunoassay	Potential autonomic adjunct	Poorer sensitivity than sAA in exercise settings	Leicht et al. (2017)
Secretory IgA concentration	No significant change after 15–45 min at 50%–80% VO2max in classic acute trials	ELISA	Mucosal immunity screening	Acute moderate exercise may not perturb concentration despite changes in secretion rate	S. McDowell et al. (1992); S. L. McDowell et al. (1991); Schouten et al. (1988)
Secretory IgA secretion rate	Heavy training blocks can reduce secretion rate by roughly 10%–50% depending on context	ELISA plus flow correction	Better illness‐risk context than concentration alone	Requires accurate flow‐rate measurement	Lindsey et al. (2025); Neves et al. (2023); S. McDowell et al. (1992); S. L. McDowell et al. (1991); Schouten et al. (1988)
Interleukin‐6	Typically rises after strenuous exercise, but saliva‐plasma agreement is weak	Immunoassay	Tracks oral‐fluid inflammatory activation	Poor direct transferability to blood concentration	Cullen et al. (2015); Minetto et al. (2005)
Uric acid	Usually increases after strenuous aerobic or match stress	Enzymatic assay	Low‐cost redox and purine‐turnover marker	Diet and fasting status influence baseline	González et al. (2008)
Nitric‐oxide related products	Acute post‐exercise increases have been reported in aerobic protocols	Colorimetric nitrite/nitrate assays	Redox‐endothelial adjunct marker	Strong oral bacterial contribution	González et al. (2008)
Lactate	About 3.3‐fold increase after a soccer match in non‐normalized saliva	Portable analyzers; enzymatic assay; NMR	Fast oral‐fluid marker of glycolytic stress	Faster return to baseline than blood; contamination‐sensitive	Kojima et al. (2025); Pitti et al. (2019); Yan et al. (2023)
Succinate and pyruvate	Succinate remained elevated after protein normalization; pyruvate was normalization‐sensitive	NMR; MS	Links saliva to mitochondrial and glycolytic pathway shifts	Effects may vanish or invert under alternate normalization schemes	Farjallah et al. (2018); Pitti et al. 2019; Yi et al. (2025)
Metabolomic panels	56 metabolites changed after match play; rapid global shifts also reported in acute‐stress studies	NMR; LC–MS; untargeted MS	Captures multi‐pathway response rather than single markers	Requires rigorous timing, preprocessing, and normalization	Farjallah et al. (2018); Pitti et al. (2019); Yi et al. (2025)

4. Sweat in Exercise Chemistry

4.1. Strengths and Limitations of Sweat as a Wearable Specimen

Sweat has become the emblem of noninvasive sports analytics because it can, at least in principle, be accessed continuously during motion. That advantage is real. A well‐designed wearable patch can monitor local sweat chemistry without interrupting exercise, and for a field scientist or coach, this is far more practical than repeated blood sampling. Yet sweat is also the matrix in which engineering progress has most clearly outpaced physiological certainty.

Two physiological facts frame the discussion. First, final sweat composition is not a passive reflection of plasma. Second, sweat chemistry is inseparable from sweat rate. Baker and Wolfe emphasize that ions, metabolites, cytokines, and small molecules in final sweat are shaped by secretion, ductal reabsorption, sweat‐gland metabolism, and skin contamination, so concentrations often differ substantially from blood (Baker and Wolfe 2020). That conclusion matters because many early wearable‐sensor narratives implicitly treated sweat as a convenient blood surrogate. The field now understands that this is too simple.

For electrolyte monitoring, the case is strongest. Sodium and chloride losses can be sufficiently large to affect hydration planning, especially in prolonged exercise or hot environments. In retrospective athlete datasets, whole‐body sweating rate and sweat sodium loss vary enormously across sport and environment, with whole‐body sweating rates spanning wide ranges and local sweat sodium concentrations clustering around but not limited to approximately 20–80 mmol/L (Baker 2017; Barnes et al. 2019; Sonner et al. 2015). Athletes with consistently salty sweat are obvious candidates for personalized sodium replacement strategies. Here, sweat analysis is not trying to infer plasma sodium directly; it is measuring the actual composition of a loss pathway, which is a much more defensible use case. That framing also aligns with field‐validation studies and sports‐hydration consensus documents, which repeatedly caution that individualized fluid and sodium strategies depend on validated collection methods, regional‐to‐whole‐body interpretation, and realistic sweat‐loss estimates rather than single isolated values (Belval et al. 2019; ‘American College of Sports Medicine et al. 2007; Heaney et al. 2016; McDermott et al. 2017; Sawka et al. 2005; Taylor and Machado‐Moreira 2013).

The situation is weaker for metabolites such as lactate and glucose. Sweat lactate often lies in the 1–25 mM range, far above blood lactate, which immediately indicates that glandular processes and local production are involved (Van Hoovels et al. 2021). Glucose is present at much lower concentrations, usually in the tens to hundreds of micromolar range, and can be analytically difficult to resolve without careful calibration (W. Gao et al. 2016; Lin et al. 2022; Ma et al. 2024). Correlation studies between sweat and blood are mixed for most constituents, and a recent direct comparison study again showed that blood and sweat concentrations are frequently not interdependent in the way developers hope (Klous et al. 2021). Sweat chemistry is therefore powerful, but only when the target variable is matched to what sweat actually contains and what it actually represents.

4.2. Wearable Electrochemical and Microfluidic Systems

The hardware progress in sweat analytics has been remarkable. Gao et al. established a landmark with a fully integrated wearable sensor array that simultaneously measured glucose, lactate, sodium, potassium, and temperature, while wirelessly transmitting data to a mobile interface (W. Gao et al. 2016). The article is still foundational because it combined chemistry, signal conditioning, thermal calibration, and real‐time field wearability in one platform. For exercise chemistry, this mattered more than any one analyte: It proved that sweat analysis could become operational rather than strictly laboratory‐based. Later platform‐level reviews have confirmed that the engineering frontier now lies less in detecting a signal than in making flexible biochemical sensors robust under motion, drift, and real‐world wear conditions (Ghaffari et al. 2021; J. Kim et al. 2019; Rao et al. 2026; Sempionatto et al. 2022).

Subsequent systems have diversified by target and format. Potentiometric platforms now allow dual or multiplex measurement of sodium, potassium, chloride, and pH (Jalal et al. 2024; Pirovano et al. 2020). Pirovano et al. reported sodium and potassium ion‐selective electrodes with near‐Nernstian sensitivities and on‐body cycling data showing sweat sodium rising from 1.89 to 2.97 mM and potassium from 3.31 to 7.25 mM over 90 min on their specific platform (Pirovano et al. 2020). More recent wearable sodium systems have measured concentrations across 10–200 mM, with one study observing a fall from 101 to 67 mM over 30 min of exercise as sweating progressed (Jalal et al. 2024). These values are method‐dependent, but they show how device‐level measurements can now generate time‐resolved electrolyte traces rather than isolated end‐point samples.

Lactate sensing has become the most visible sport application because lactate is intuitively associated with fatigue and threshold concepts. Van Hoovels et al. critically reviewed the field and emphasized both the promise and the unresolved contradictions: Sweat lactate is easy to detect analytically, but its physiological meaning is still debated because glandular production, local skin effects, and lag relative to blood are not fully resolved (Van Hoovels et al. 2021). In response, newer studies have emphasized on‐body validation. Xuan et al. presented a fully integrated wearable device for continuous sweat lactate monitoring during cycling and kayaking (Xuan et al. 2023), whereas Yang et al. summarized devices and noted that modern systems often cover approximately 0–100 mM, which is operationally adequate for sport monitoring even if the underlying physiology remains contested (Yang et al. 2024). Most recently, Zhu et al. reported a battery‐free sweat lactate patch with a 0.1–30 mM linear range and a sensitivity of 9.76 uA mM⁻¹ cm⁻², explicitly targeting multiday muscle fatigue and recovery tracking (Zhu et al. 2025).

Glucose and dual‐analyte devices have progressed as well. Lin et al. developed a hydrogel patch for natural sweat glucose sensing, extending sweat analytics beyond exercise‐induced heavy perspiration (Lin et al. 2022). Ma et al. integrated potassium and glucose monitoring into a wearable platform that issued threshold‐based warnings at 7.5 mM for K + and 60 or 120 μM for glucose (Ma et al. 2024). Such systems are analytically attractive because they move from chemistry to decision‐support logic. But they also show the translational dilemma: Threshold alerts imply that the biological meaning of the measured sweat concentration is already established, which is often not yet true outside well‐characterized use cases.

4.3. Interpretation Remains the Main Challenge in Sweat Analysis

Pre‐analytical and physiological context remain the limiting steps for sweat translation. Sweat rate changes analyte concentration independently of systemic state, skin contamination can contribute amino acids or trace molecules, and regional collectors do not necessarily represent whole‐body composition (Baker 2017; Baker et al. 2016; Baker and Wolfe 2020; Barnes et al. 2019; Sonner et al. 2015). Even mild dehydration can change the sweating response without necessarily altering the acute direction of sweat electrolyte change in the way practitioners expect (Baker et al. 2024). The implication is subtle but important: Sweat chemistry is often better at characterizing the sweat process itself than at estimating internal concentration directly.

Several studies illustrate how much useful information is nonetheless available when the question is framed correctly. Harshman et al. found that exercise‐induced sweat metabolomics is sufficiently stable for controlled collection and storage workflows, supporting biomarker‐discovery pipelines (Harshman et al. 2019). Delgado‐Povedano et al. identified 135 compounds in human sweat collected after moderate exercise using GC‐TOF/MS, showing that sweat contains a far richer chemical landscape than traditional electrolyte panels suggest (Delgado‐Povedano et al. 2018). A pilot proteomic–metabolomic characterization study further argued that sweat is promising for human‐performance monitoring even though protein abundance is low and some proteins appear degraded (Harshman et al. 2018). Meanwhile, high‐intensity interval training‐induced fatigue has already been profiled by sweat metabolomics in athlete populations (Meihua et al. 2023).

The lesson is not that sweat has failed as a blood surrogate. Rather, sweat is evolving into its own analytical domain. It is particularly strong for thermoregulatory losses, individualized sodium replacement, continuous regional electrolyte trends, wearable lactate engineering, and exploratory metabolomic fingerprinting under ecologically valid exercise conditions (Baker 2017; Baker et al. 2016, 2024; Baker and Wolfe 2020; Barnes et al. 2019; Delgado‐Povedano et al. 2018; F. Gao et al. 2023; Harshman et al. 2019, 2018; Klous et al. 2021; Meihua et al. 2023; Van Hoovels et al. 2021; Xu et al. 2021; Xuan et al. 2023; Yang et al. 2024; Zhu et al. 2025). It remains weaker for claiming systemic equivalence without simultaneous validation. That split between robust signal capture and selective biological mapping is quantified in Table 2. Electrolyte and sweat‐rate applications look relatively mature because they describe the sweat process directly, whereas glucose, cortisol, cytokines, and even lactate still carry much stronger matrix‐specific caveats despite increasingly capable hardware.

TABLE 2.

Quantitative comparison of representative sweat targets and sweat‐sensing systems used in exercise chemistry. Values are representative, protocol‐dependent examples from the cited literature rather than pooled effect estimates.

Sweat target or system	Representative quantitative feature	Main analytical format	Practical readiness/main use case	Main interpretive limitation	Representative refs.
Whole‐body sweat rate	About 0.2–3.0 L/h across wide exercise and environmental conditions	Body‐mass balance; washdown; patch prediction	Essential for estimating fluid loss and analyte flux	Strongly influenced by environment, acclimation, body size, clothing	Baker (2017); Baker et al. (2016); Baker and Wolfe (2020); Barnes et al. (2019)
Sweat sodium	Salty sweat often defined as about 70–80 mmol/L or higher; wearable systems can measure 10–200 mM	Ion‐selective electrodes; lab ionometry	Hydration and sodium‐replacement planning	Regional measurements do not directly equal whole‐body loss	Baker (2017); Baker et al. (2016); Baker and Wolfe (2020); Barnes et al. (2019); Jalal et al. (2024); Pirovano et al. (2020)
Sweat chloride	Commonly in the low tens of mmol/L; dynamic load‐sensitive changes measured after calibration with 10, 50, and 100 mM standards	Potentiometric chloride sensors	Useful for electrolyte‐loss profiling and sensor validation	Ductal reabsorption and site dependence complicate absolute interpretation	Baker et al. (2024); Baker and Wolfe (2020); Xu et al. (2021)
Sweat potassium	Typical exercise ranges are low‐mM; wearable patches report about 3.31 to 7.25 mM during 90 min exercise	Ion‐selective electrodes	Supports electrolyte pattern monitoring	Less directly linked to replacement strategy than sodium	Baker and Wolfe (2020); Jalal et al. (2024); Ma et al. (2024); Pirovano et al. (2020)
Sweat lactate	Usually about 1–25 mM; battery‐free patch range 0.1–30 mM with sensitivity 9.76 uA mM^‐1 cm^‐2	Enzymatic amperometry	Continuous signal for local exercise stress and device development	Origin is partly glandular; correlation with blood is inconsistent	Klous et al. (2021); Lin et al. (2022); Van Hoovels et al. (2021); Xuan et al. (2023); Yang et al. (2024)
Sweat glucose	Generally tens to hundreds of uM; integrated systems often calibrate 0–200 uM	Enzymatic amperometry	Potential adjunct for metabolic monitoring	Very low concentration and uncertain blood transfer	W. Gao et al. (2016); Harshman et al. (2019); Ma et al. (2024)
Sweat pH	Roughly 4.5–7.0 depending on site and protocol	Potentiometric sensors	Contextualizes enzyme activity and skin microenvironment	Sensitive to local evaporation and device dead volume	Jalal et al. (2024); Ma et al. (2024)
Sweat ammonia	Often sub‐mM to low‐mM and rises with exercise stress	Enzymatic or electrochemical assays	Potential marker of amino‐acid catabolism and gland contribution	Strong local‐gland and skin contributions	Baker and Wolfe (2020); Klous et al. (2021)
Sweat urea	Commonly low‐mM to tens of mM	Colorimetry; electrochemistry	Supports nitrogen‐excretion profiling	Contamination from skin surface and uncertain blood coupling	Baker and Wolfe (2020); Klous et al. (2021)
Sweat cortisol	Usually low ng/mL‐scale	Immunosensors; electrochemistry	Attractive stress‐hormone target for wearables	Transfer from blood and skin contamination remain unresolved	Baker and Wolfe (2020); F. Gao et al. (2023); Xu et al. (2021)
Sweat cytokines	Often pg./mL‐scale and near analytical limits	Immunosensors	May extend wearables into inflammation monitoring	Biological origin and stability remain uncertain in exercise settings	Baker and Wolfe (2020); F. Gao et al. (2023); Xu et al. (2021)
Multiplex temperature‐compensated arrays	Simultaneous glucose, lactate, sodium, potassium, and temperature in one wearable system	Integrated flexible electronics	Best current example of field‐oriented analytical integration	Sensor lifetime and matrix calibration still constrain long events	W. Gao et al. (2016)
Sodium‐specific microfluidic wearables	10–200 mM measurement window; coefficient of variation below 4% in validation	Microfluidic chip plus potentiostat	High‐value use case for hydration strategy	Forehead or local data may not represent whole body	Jalal et al. (2024)
Metabolomic stability handling	More than 88% of untargeted data points remained within limits for up to 90 min	LC–MS metabolomics	Supports practical sample transport and biobanking	Stability after collection does not solve secretion heterogeneity	Harshman et al. (2019)
Global sweat metabolome	135 compounds identified after moderate exercise; HIIT fatigue studies show state‐dependent panels	GC‐TOF/MS; untargeted metabolomics	Opens discovery space beyond electrolytes	Discovery‐stage findings remain far from actionable thresholds	Delgado‐Povedano et al. (2018); Harshman et al. (2018); Meihua et al. (2023)

5. Exhaled Breath in Exercise Chemistry

5.1. Advantages and Analytical Challenges of Exhaled Breath

Exhaled breath provides an analytical opportunity no liquid specimen can fully match: second‐scale access to volatile metabolism during exercise. This is especially attractive when physiological transitions are rapid, such as the onset of exercise, threshold crossing during graded workloads, substrate switching under dietary manipulation, or early recovery. Yet the same features that make breath attractive also make it difficult. Measured concentrations are shaped by pulmonary perfusion, airway exchange, humidity, dead space, and ventilation pattern in addition to systemic production (Ager et al. 2020; Amann et al. 2014; De Lacy Costello et al. 2014; Osswald et al. 2021; Verges et al. 2005).

The field's modern phase has been driven by more sophisticated instrumentation. Osswald et al. used SESI‐HRMS during graded cycle ergometry and identified 333 exhaled metabolites that changed significantly from the start of the ramp to VO2max, with pathway‐level decreases in glyoxylate and dicarboxylate, TCA, and tryptophan‐related metabolites (Osswald et al. 2021). This study is important because it repositions breath analysis from a handful of gases toward continuous metabolite‐network tracking. Instead of asking only whether breath acetone rises, the study asks how the breath metabolome reorganizes as workload climbs to exhaustion.

Classical single‐analyte work remains valuable, however. Acetone is strongly linked to lipid oxidation and ketone‐body metabolism, whereas isoprene has been associated with cholesterol biosynthesis and rapid hemodynamic or muscle‐related effects. King et al. showed that acetone and isoprene display distinct concentration profiles during exercise on an ergometer (King et al. 2009). Later work refined the picture by showing that acetone behavior depends on fitness level, nutrition, and inhaled background concentrations (Ager et al. 2020; Hori et al. 2020; M. J. Kim et al. 2020; Königstein et al. 2021; Ota et al. 2020). The biological implication is useful: Breath can report metabolism quickly, but only analyte‐specific models can explain the response.

5.2. Breath Acetone and Isoprene as Key Targets

Among volatile compounds, acetone is the most compelling exercise marker because it often increases when lipid oxidation is emphasized. Konigstein et al. found that maximal breath‐acetone ratios during submaximal cycling were significantly higher in high‐fit than lower fit individuals, paralleling differences in fat oxidation and beta‐hydroxybutyrate responses (Königstein et al. 2021). Nutritional manipulations push the signal further. After sprint exercise, reduced carbohydrate intake increased breath acetone to 0.90 ppm at 4 h compared with 0.66 ppm in a normal‐carbohydrate condition (Ota et al. 2020). Even inhalation of molecular hydrogen has been reported to augment exercise‐associated breath acetone, again indicating sensitivity to oxidative and substrate‐handling processes (Hori et al. 2020). For exercise chemistry, these findings suggest that breath acetone is not merely a fasting marker; it can act as a dynamic readout of substrate selection during and after exercise.

Isoprene behaves differently. In the maximal‐exercise pilot study by Heaney et al., isoprene tended to decrease at 10 min after exercise and return toward baseline by 60 min, with the 10‐ and 60‐min time points differing significantly (p = 0.041) (Heaney et al. 2022). The pattern indicates fast kinetics but also high inter‐individual variability. Work by King and colleagues supports the idea that isoprene has pronounced within‐exercise dynamics during ergometer work (King et al. 2009), and a recent dedicated review further highlights that isoprene physiology is multi‐determinant and still incompletely resolved (Mochalski et al. 2023). These characteristics make isoprene a fascinating mechanistic analyte but a fragile practical biomarker unless the sampling protocol is tightly standardized.

Breath chemistry can also connect physiology to behavior in near real time. Breath acetone measurement has already been used to model exercise‐induced energy and substrate expenditure (M. J. Kim et al. 2020). Such models remain early, but they show where the field may eventually go: Breath chemistry could become part of continuous systems for estimating metabolic state, not just cataloging compounds.

5.3. Breathomics, Nitric Oxide, and Pattern Recognition

Breath chemistry is not limited to acetone and isoprene. Henderson et al. monitored participants during 3 days of prolonged exercise in a field campaign using PTR‐ToF‐MS and found that acetone increased after the first day of walking and decreased toward baseline before walking on day two, whereas short‐chain fatty acids such as acetic, butanoic, and propionic acids emerged as potential indicators of gut‐microbiota‐related responses to exercise and medication use (Henderson et al. 2021). This study matters because it shows that breathomics can survive outside highly controlled laboratories and can detect cumulative exercise effects rather than only acute laboratory ramps.

Pattern‐recognition approaches offer another route. Bikov et al. showed more than a decade ago that exercise changes exhaled volatile patterns measurable by an electronic nose (Bikov et al. 2011). More recently, Tondo et al. demonstrated that e‐nose breathprints could discriminate normoxia, hypoxia, and hyperoxia with cross‐validated accuracies of 63%, 65%, and 71% between pairwise conditions ^[31]. Although not specific to athletic performance per se, such classification suggests that breath patterning can detect altered oxygenation states relevant to training, altitude exposure, and respiratory load.

Exhaled nitric oxide provides a third axis of information. Unlike VOCs linked to substrate metabolism, nitric oxide is usually interpreted within airway physiology, vascular regulation, and inflammation. Sheel et al. reviewed the exercise response of exhaled NO and argued that exercise generally increases exhaled NO production, though the compartmental source is debated (Sheel et al. 1999). More targeted work showed that high‐intensity exercise acutely reduced exhaled NO partial pressure from 1.47 to 1.11 mPa at 2800 m and from 1.54 to 1.04 mPa at 180 m, with normalization within 20 min (Stang et al. 2015). Endurance‐athlete studies in normoxia versus hypoxia, and work in athletes with exercise‐induced hypoxemia, have further demonstrated that exhaled NO kinetics can reflect airway and pulmonary vascular strain rather than simply whole‐body metabolism (Kippelen et al. 2002; Verges et al. 2005). This airway‐oriented perspective is practically important because exercise‐induced bronchoconstriction can measurably impair performance and therefore gives breath analysis a clinically relevant role beyond metabolite tracking alone (Dickinson et al. 2023; Price et al. 2014).

5.4. Challenges in Standardization Under Real‐World Conditions

Breath chemistry is perhaps the most vulnerable of the three matrices to environmental and protocol drift. Ambient VOC background, mouthpiece materials, humidity, inspired gas composition, and the decision to collect fixed‐volume versus end‐tidal or real‐time samples all materially affect the data (Amann et al. 2014; De Lacy Costello et al. 2014; Heaney et al. 2022; Henderson et al. 2021; Osswald et al. 2021; Tondo et al. 2024). Ager et al. explicitly demonstrated that inhaled acetone concentrations can influence exhaled acetone at rest and during exercise (Ager et al. 2020). That paper is an important caution: If the athlete is exercising in a chemically complex environment, failure to measure the inspired fraction may compromise the entire analysis.

Breathomics therefore has a high ceiling and a high entry cost. It is analytically strongest when the biological question is rapid and volatile, the instrumentation can control or measure the breathing process, and the environment is characterized. It is weaker when collection is delayed, ambient air is uncontrolled, or investigators want a universal biomarker across different breathing patterns and workloads. For those reasons, exhaled breath currently excels in mechanistic metabolic studies, high‐resolution threshold work, field‐proofed VOC profiling, and airway‐oriented assessments, whereas routine team‐sport deployment remains early‐stage (Ager et al. 2020; Bikov et al. 2011; Heaney et al. 2022; Henderson et al. 2021; Hori et al. 2020; M. J. Kim et al. 2020; King et al. 2009; Kippelen et al. 2002; Königstein et al. 2021; Mochalski et al. 2023; Osswald et al. 2021; Ota et al. 2020; Sheel et al. 1999; Stang et al. 2015; Tondo et al. 2024; Verges et al. 2005). Figure 2 represents the strongest version of that argument because it shows that exercise breathomics has moved beyond single‐marker phenomenology into coordinated pathway‐level behavior, including graded changes in glyoxylate and dicarboxylate, TCA‐related, tryptophan‐related, and other metabolite classes during CPET.

VOC classification studies add a complementary message: maximal‐exercise breath analysis can be treated less as pathway reconstruction and more as a classification problem, where multivariate separation and candidate‐feature selection discriminate exercise‐responsive VOC states rather than assigning every peak to a single mechanistic origin. Taken together, Figure 2 and Table 3 show that breath chemistry now spans mechanistic pathway mapping, canonical volatile monitoring, and pattern‐recognition classification. Table 3 extends that comparison quantitatively and makes the central limitation hard to miss: Translational maturity remains analyte‐specific, with acetone, isoprene, and FeNO already interpretable under defined conditions, whereas broader breathomic signatures still depend heavily on platform choice, annotation confidence, and environmental control.

TABLE 3.

Quantitative comparison of representative exhaled‐breath targets used in exercise chemistry. Values are representative, protocol‐dependent examples from the cited literature rather than pooled effect estimates.

Breath target or feature	Representative quantitative response	Main analytical format	Practical readiness/main use case	Main interpretive limitation	Representative refs.
Acetone	Often about 0.3–2.0 ppm at rest; rises with prolonged exercise, lipid oxidation, and low‐carbohydrate recovery; 0.66 versus 0.90 ppm at 4 h after sprint exercise under normal versus reduced carbohydrate intake	PTR‐MS; GC–MS; real‐time MS	Strong candidate for substrate‐utilization monitoring	Influenced by diet, fasting state, and inspired background acetone	Ager et al. (2020); Henderson et al. (2021); Hori et al. (2020); M. J. Kim et al. (2020); Königstein et al. (2021); Ota et al. (2020)
Isoprene	Dynamic with workload transitions; 10 versus 60 min post‐VO2max difference reached p = 0.041 in a pilot study	GC–MS; PTR‐MS	Fast kinetic marker of exercise response	High inter‐individual variability and uncertain mechanistic origin	Heaney et al. (2022); King et al. (2009); Mochalski et al. (2023)
FeNO or exhaled NO	Partial pressure decreased from 1.47 to 1.11 mPa at 2800 m and from 1.54 to 1.04 mPa at 180 m, recovering within 20 min	Chemiluminescence NO analyzers	Useful for airway or pulmonary response profiling around exercise	More reflective of airway physiology than systemic metabolism	Kippelen et al. (2002); Sheel et al. (1999); Stang et al. (2015); Verges et al. (2005)
Real‐time breath metabolome	33 metabolites changed significantly during graded CPET	SESI‐HRMS	Pathway‐level dynamic metabolomics in vivo	Requires specialized instrumentation and controlled breathing workflow	Osswald et al. (2021)
Glyoxylate and dicarboxylate pathway markers	Significant decreases across graded exercise	SESI‐HRMS	Expands breath analysis beyond canonical VOCs	Compound identification and source attribution remain challenging	Osswald et al. (2021)
TCA‐related breath metabolites	Significant decreases toward VO2max	SESI‐HRMS	Connects breath to oxidative metabolism	Biological interpretation still relies on cross‐platform confirmation	Osswald et al. (2021)
Tryptophan‐related metabolites	Significant decreases during incremental exercise	SESI‐HRMS	Suggests immune‐metabolic coupling in breath	Low abundance and pathway assignment require high‐confidence annotation	Osswald et al. (2021)
Short‐chain fatty acids	Acetic, butanoic, and propionic acids increased after prolonged exercise day 1 and declined toward baseline before day 2	PTR‐ToF‐MS	Captures cumulative exercise and gut‐microbiota‐related signals	Sensitive to diet, medication, and ambient chemistry	Henderson et al. (2021)
Breathprint by e‐nose	63%–71% cross‐validated discrimination among normoxia, hypoxia, and hyperoxia	Electronic nose	Practical pattern‐recognition approach for respiratory‐state classification	Limited chemical specificity	Bikov et al. (2011); Tondo et al. (2024)
Fitness stratification by breath acetone	Maximum BrAce ratios were higher in high‐fit individuals during submaximal cycling (p = 0.03)	PTR‐TOF‐MS	Links breath chemistry to aerobic phenotype	Requires fasting and tightly controlled workload	Königstein et al. (2021)
Hydrogen‐assisted breath acetone response	1% H2 inhalation significantly augmented exercise breath acetone (p < 0.01)	Real‐time breath acetone monitoring	Demonstrates sensitivity to metabolic perturbation	Intervention‐specific and not generalizable without replication	Hori et al. (2020)
Exercise energy prediction	Breath acetone has been used to model exercise‐induced energy and substrate expenditure	Sensor plus modeling	Moves breath chemistry toward actionable metabolic estimation	Model transferability across populations is still limited	M. J. Kim et al. (2020)
Background‐air sensitivity	Inhaled acetone measurably alters exhaled acetone at rest and during exercise	PTR‐MS	Forces environmental control into protocol design	Ambient‐air correction is mandatory for real‐world deployment	Ager et al. (2020)
Post‐maximal recovery kinetics	Isoprene tends to dip early after exhaustive exercise and recover by 60 min	TD‐GC–MS	Useful for recovery‐phase VOC tracking	Delayed offline sampling sacrifices second‐scale kinetics	Heaney et al. (2022)
Field‐campaign deployment	Repeated outdoor walking experiments captured stable exercise‐related breath trends	PTR‐ToF‐MS	Demonstrates ecological feasibility of breathomics	Logistics and environmental control remain demanding	Henderson et al. (2021)

6. Cross‐Matrix Comparison and Translational Outlook

6.1. Different Specimens Are Suitable for Different Research Questions

The most important conclusion from the current literature is that saliva, sweat, and exhaled breath are not competing to become a universal noninvasive replacement for blood. They are solving different analytical problems. Saliva is strongest when the target is a rapidly sampled neuroendocrine, mucosal‐immune, oxidative, or mixed metabolomic signal, and when field practicality is important but second‐to‐second resolution is not (Hayes et al. 2015; Kojima et al. 2025; Lindsey et al. 2025; Neves et al. 2023; Ntovas et al. 2022; Pitti et al. 2019; Yan et al. 2023; Yi et al. 2025). Sweat is strongest when the question concerns continuous surface‐accessible chemistry, especially electrolyte loss, regional dynamic trends, or wearable‐device development (Baker 2017; Baker et al. 2016, 2024; Baker and Wolfe 2020; Barnes et al. 2019; Delgado‐Povedano et al. 2018; Emaminejad et al. 2017; F. Gao et al. 2023; W. Gao et al. 2016; Harshman et al. 2019, 2018; Jalal et al. 2024; Klous et al. 2021; Lin et al. 2022; Ma et al. 2024; Meihua et al. 2023; Pirovano et al. 2020; Van Hoovels et al. 2021; Xu et al. 2021; Xuan et al. 2023; Yang et al. 2024; Zhu et al. 2025). Exhaled breath is strongest when the target chemistry is volatile, highly dynamic, and tightly coupled to ventilation or substrate oxidation (Ager et al. 2020; Bikov et al. 2011; Heaney et al. 2022; Henderson et al. 2021; Hori et al. 2020; M. J. Kim et al. 2020; King et al. 2009; Kippelen et al. 2002; Königstein et al. 2021; Mochalski et al. 2023; Osswald et al. 2021; Ota et al. 2020; Sheel et al. 1999; Stang et al. 2015; Tondo et al. 2024; Verges et al. 2005).

This framing helps clarify why attempts to compare the matrices on a single convenience scale are not very informative. Saliva and sweat are easier to store and batch‐process than breath, but breath offers unmatched temporal resolution. Sweat is the most natural partner for wearables, but salivary assays remain analytically more mature for many hormones and proteins. Saliva and breath are less affected by skin contamination than sweat, but saliva is more vulnerable to local microbial metabolism than investigators sometimes acknowledge. In other words, “noninvasive” is not a single attribute. It is a bundle of trade‐offs among sample access, analytical stability, biological specificity, and engineering burden.

The literature also suggests that the matrices should not always be treated as substitutes. In some settings, they are better viewed as complementary channels. For example, salivary cortisol or alpha‐amylase can characterize psychophysiological load, sweat sodium can inform fluid‐electrolyte strategy, and breath acetone can report substrate utilization during the same training block. A multimatrix design may therefore offer more interpretable exercise chemistry than any single alternative specimen alone, provided the study is powered and the timing architecture is explicit (W. Gao et al. 2016; Heaney et al. 2019; Jason Heikenfeld et al. 2019; Osswald et al. 2021; Pitti et al. 2019). The numerical ranges in Tables 1, 2, 3 should be read as protocol‐bound examples. They compare what each matrix can plausibly answer, not universal thresholds or interchangeable effect sizes across heterogeneous studies.

6.2. Matrix‐Specific Validation Is Still Needed

A recurring problem across the field is that analytical novelty is often rewarded before physiological validation. This is especially visible in wearable sweat sensing, but similar patterns occur in salivary biosensors and compact breath‐analysis systems (Baker and Wolfe 2020; Ma et al. 2024; Malon et al. 2014; Osswald et al. 2021). Future studies should prioritize at least four validation layers.

First, sampling biology must be characterized. For saliva, this means flow rate, stimulation status, circadian timing, oral health, recent intake, and rapid post‐exercise collection windows. For sweat it means body site, sweat rate, skin cleaning, evaporation control, and whether the measurement is intended to represent regional chemistry or whole‐body loss. For breath it means dead‐space handling, alveolar targeting, ventilation logging, ambient air measurement, and humidity‐aware processing (Ager et al. 2020; Baker 2017; Baker and Wolfe 2020; Jason Heikenfeld et al. 2019; Henderson et al. 2021; Osswald et al. 2021; Pitti et al. 2019).

Second, analyte meaning must be matrix‐native. Sweat lactate should not be treated as a blood‐lactate clone unless simultaneous studies prove the intended relationship under the relevant exercise conditions (Cullen et al. 2015; Minetto et al. 2005). Salivary IL‐6 should not be used as though it were plasma IL‐6. Breath acetone should not be interpreted without nutritional context (Van Hoovels et al. 2021; Yan et al. 2023). Progress will accelerate when studies stop asking whether alternative matrices are “as good as blood” in the abstract and instead ask whether they answer the intended sport‐science question better than the incumbent method (M. J. Kim et al. 2020; Klous et al. 2021; Königstein et al. 2021).

Third, the field needs realistic reference standards. Many studies still validate against benchtop chemistry in artificial matrices or under sedentary conditions. Exercise chemistry requires validation during movement, heat strain, dehydration, repeated bouts, and real recovery windows. That is particularly important for wearables, where motion, sweat flux, adhesive failure, and sensor drift can dominate performance after promising benchtop calibration (Emaminejad et al. 2017; F. Gao et al. 2023; W. Gao et al. 2016; Jalal et al. 2024; Lin et al. 2022; Ma et al. 2024; Pirovano et al. 2020; Van Hoovels et al. 2021; Xu et al. 2021; Xuan et al. 2023; Yang et al. 2024; Zhu et al. 2025).

Fourth, there is a pressing need for interoperable multimodal datasets. Breathomics, metabolomics, electrochemical wearables, heart‐rate‐derived load, temperature, and performance metrics are increasingly collected in parallel, but often not aligned temporally or analytically. Exercise chemistry will benefit most when these streams are fused into mechanistic models rather than reported as isolated biomarker snapshots. Recent discussions of exercise metabolomics and immunometabolism make the same point from a systems‐biology perspective: Biomarker behavior is multilevel, and single‐analyte interpretation becomes less reliable as training and recovery stressors accumulate (Baker et al. 2020; Nieman and Pence 2020; Padilha et al. 2022).

6.3. Future Research Directions

For analytical chemists, the next phase of work should emphasize method comparability, calibration transfer, internal‐standard strategies, and stability‐aware workflows rather than only new targets. Salivary metabolomics needs more explicit normalization frameworks and better reporting of pre‐analytical conditions. Sweat analytics needs flux‐aware or rate‐aware models that distinguish concentration changes from secretion‐rate changes. Breath analysis needs hybrid platforms that integrate environmental sensing, ventilatory metadata, and real‐time compound identification to reduce interpretive ambiguity (Baker and Wolfe 2020; Heaney et al. 2019; Jason Heikenfeld et al. 2019; Osswald et al. 2021; Pitti et al. 2019; Tondo et al. 2024).

There is also clear room for chemistry‐led innovation at the matrix interface. Soft materials that reduce contamination or evaporation, passive enrichment strategies for low‐abundance analytes, real‐time internal calibration, and hybrid sensing platforms that combine targeted electrochemistry with untargeted confirmation could materially improve field‐ready robustness. Importantly, such innovations should be evaluated not only for sensitivity and limit of detection but for the extent to which they improve physiological specificity under real exercise conditions.

Finally, the field should resist the tendency to promise universal point‐of‐care decision making too early. Many current applications are already valuable without that claim. Salivary biomarker panels can enrich recovery studies and tactical‐readiness screening. Sweat patches can support individualized sodium replacement and generate region‐specific time‐series data during training. Breathomics can reveal substrate and airway dynamics with a granularity that blood sampling cannot match. These are meaningful contributions even before full clinical‐grade or competition‐grade automation is achieved, and Figure 3 makes the sequence explicit: Matrix‐aware sampling and validation have to come before multimatrix fusion; otherwise, later stages merely scale pre‐analytical noise instead of actionable physiology.

FIGURE 3.

Translational roadmap for exercise chemistry using alternative specimens. The roadmap summarizes the review's proposed sequence from matrix‐aware sampling to validated analytics, multimatrix data fusion, and decision‐ready sport implementation; it is a conceptual synthesis and does not report primary data.

7. Conclusion

Saliva, sweat, and exhaled breath are analytically credible only when used as matrix‐specific tools rather than interchangeable blood substitutes. The most defensible near‐term applications are (i) salivary cortisol, alpha‐amylase, secretory IgA, and selected metabolite panels for repeated load, recovery, mucosal‐immune, and short‐latency metabolic profiling when timing, flow rate, and oral contamination are controlled; (ii) sweat rate and sweat sodium/chloride monitoring for individualized fluid‐electrolyte planning when local measurements are calibrated against whole‐body loss and environmental conditions; and (iii) breath acetone, isoprene, and FeNO or related volatile/airway markers for substrate‐use, recovery, and respiratory‐response questions when diet, ventilation, humidity, and ambient air are recorded. Sweat lactate, sweat glucose, sweat cortisol, cytokines, and broad breathomic fingerprints remain promising but less decision‐ready because local secretion biology, annotation confidence, calibration drift, and cross‐platform transfer still limit physiological interpretation. Future work should therefore report sampling timing, normalization, matrix‐specific controls, field calibration, and simultaneous comparator measures before moving from analytical detection to practical decision making. When the analyte, specimen, time scale, and validation model are matched to the physiological problem, alternative specimens can answer exercise‐chemistry questions that blood or urine cannot practically capture.

Conflicts of Interest

The author declares no conflicts of interest.

Data Availability Statement

Data sharing not applicable to this article as no datasets were generated or analysed during the current study.