Physiological mechanisms determining eccrine sweat composition

Scientists and practitioners have employed various methods to collect sweat samples for composition analysis. Whole body washdown is considered the most accurate method because sweat runoff from the entire body surface area is collected and accounted for and it does not interfere with normal evaporative sweating. Recovery of water and Na⁺ using this method has been reported to be ~ 99 to 102% and has a within-method coefficient of variation (CV) of ~ 4% (Baker et al. 2018b; Shirreffs and Maughan 1997). For these reasons, the WBW method is recommended, especially when conducting nutrient balance studies or if interested in determining total loss of micronutrients or other constituents via sweating. However, WBW is a meticulous method that requires a well-controlled laboratory setting. Therefore, in most studies, sweat is collected from one or more small regions of the body using a myriad of techniques, which have been described in detail elsewhere (Baker 2017, 2019; Taylor and Machado-Moreira 2013). Briefly, regional sweat collection techniques for sweat composition analysis include scraping or dripping methods, filter paper, absorbent patches, sweat pouches, sweat capsules, arm gloves/bags, and sweat collectors (e.g., Macroduct^®/Megaduct). While regional sweat collection is practical, each method has its limitations, and some more than others. For example, covering the skin surface with the collection system can alter the local environment, impacting skin temperature and/or skin wettedness (Fig. 1). In turn, these changes can have confounding effects on the sweating process and/or introduce skin surface contaminants into the sweat sample. On the other hand, open-air collection techniques (scraping or dripping methods) allow evaporation of water and volatile constituents of sweat (Buono 1999; Talbert 1922) (Fig. 1). Depending on the collection device (material), the exact methods used (timing, skin preparation, etc.), and the constituent of interest, the confounding effect (background noise) could range from negligible to very large. These factors will be discussed in more detail in “Contamination considerations”.

Another important consideration is inter-regional variability in the sweat constituent concentrations. It is well documented that many sweat micronutrient (e.g., Na⁺, Cl⁻, Ca²⁺, Fe²⁺, Mg²⁺, Zn²⁺, Cu²⁺) concentrations can vary by up to two- to fourfold among anatomical sites (Aruoma et al. 1988; Baker et al. 2019, 2009; Patterson et al. 2000; Vellar and Askevold 1968). Inter-regional variability data are sparse for most other sweat constituents, but the few studies available seem to follow the same general trend, with sweat amino acid concentrations reported to vary by up to twofold (Liappis and Hungerland 1972), sweat lactate by up to two- to fourfold (Collins 1962; Patterson et al. 2000) and sweat bicarbonate by up to sixfold (Patterson et al. 2000) among body regions. One study measured the metabolomic profiles of sweat samples collected from the forearm, lower back, and neck (Hooton and Li 2017) over a 24-h period. The authors noted that the concentrations of many of the metabolites were very different among sites (shown via heat maps), but the inter-regional differences were not quantified.

Regional sweat concentrations can also differ significantly from WBW measurements. For example, sweat [Na⁺] and [Cl⁻] measured from the forehead, torso, and upper arms are often greater than whole body measurements. However, sweat [Na⁺] and [Cl⁻] from other sites such as the foot, calf, and thigh are similar to that of the whole body (Baker et al. 2019, 2009, 2018b; Patterson et al. 2000). For most other constituents (Ca²⁺, Mg²⁺, Fe²⁺, Zn²⁺, Cu²⁺, lactate, urea, ammonia, and lead), regional sweat measurements have led to overestimations of whole body sweat concentrations (Baker et al. 2011; Cohn and Emmett 1978; Colombani et al. 1997; Lemon et al. 1986; Patterson et al. 2000) albeit regional body mapping has not been as thorough (i.e., included as many regions) as that of Na⁺ and Cl⁻ studies. The physiological explanation underlying the inter-regional variability is unclear, but may be related to inter-regional differences in sweating rate and/or skin characteristics (e.g., presence of sebaceous glands, apocrine glands, hair/nails, etc.) as discussed in the next section.

As summarized in Fig. 1, there are several confounding factors that play a role in determining sweat composition. For instance, if sweat is collected in an area dense with apocrine glands and/or sebaceous glands (e.g., axilla, face, forehead, scalp) their secretions could artificially elevate lipid, protein, sugar, and ammonia concentrations of sweat samples collected from the skin surface (Montagna and Parakkal 1974a, b; Porter 2001). Other substances present in sweat may be there as a result of production in the sweat gland rather than (or in addition to) having originated from primary sweat secretion into the secretory coil. Some substances potentially produced by the eccrine gland include lactate, urea, cytokines, proteolytic enzymes, and antibodies (Gordon et al. 1971; Murota et al. 2015; Sato 1977; Sato et al. 1991; Sato and Sato 1994; Yokozeki et al. 1991), but mechanisms are not fully understood.

Residual sweat in the eccrine duct left over from a previous thermal event could also confound sweat composition initially excreted on the skin surface (i.e., at the beginning of exercise or heat stress). Similarly, skin surface contamination could originate from minerals, amino acids, and other constituents deposited on the skin after evaporation of everyday insensible transcutaneous water loss (Mitchell and Hamilton 1949; Rothman et al. 1949). As shown in Supplemental Table 2, several studies have found significantly higher constituent concentrations in initial sweat samples versus subsequent collections (Boysen et al. 1984; Brune et al. 1986; Ely et al. 2011; Paulev et al. 1983). For example, Paulev et al. (1983) collected sweat from the back of endurance athletes at 10-min intervals during cycling and found that sweat [Fe²⁺] gradually decreased (by 2.2-fold) over the first 20–30 min and did not change thereafter. Similar results have been reported by other investigators for Fe²⁺, Ca²⁺, Mg²⁺, Cu²⁺, Zn²⁺, protein, and urea (Boysen et al. 1984; Brune et al. 1986; Ely et al. 2011). For these reasons, it is recommended that the skin is cleaned (e.g., alcohol wipe) and rinsed with deionized water and that initial sweat is removed from the skin surface prior to sweat collection. Allowing participants to sweat for at least 20–30 min prior to applying the sweat collection device will facilitate flushing of duct and surface contaminants and will allow time for steady-state sweating rates to be reached.

Scientists have also assessed the impact of skin surface contamination by comparing sweat collected with and without an oil barrier (Supplemental Table 2). The advantage of this method, also known as the anaerobic sweat pouch technique, is that the oil interface prevents contact between the skin and the sweat. Using this approach, one study found that urocanic acid concentration was sevenfold higher in arm sweat collected with the filter paper technique versus the anaerobic method (Brusilow and Ikai 1968). Boysen et al. (1984) found 2.5-fold higher [Ca²⁺], 1.6-fold higher [protein], and 1.4-fold higher [urea] in sweat collected from the upper back using the pouch technique without oil compared with the pouch with oil. Using the scraping method resulted in even more epidermal contamination, as Boysen et al. (1984) reported 6-fold, 2.7-fold, and 2.4-fold increases in sweat [Ca²⁺], [protein], and [urea], respectively, in scraped sweat versus the anaerobic method. Yet another approach to assessing the impact of epidermal contamination on sweat composition is to separate a sample into cell-rich and cell-poor aliquots. Cell-rich sweat is defined as the sweat analyzed ‘as is’ after collection from the skin surface, while cell-poor sweat is the supernatant obtained after centrifugation of that sample. Several studies (listed in Supplemental Table 2) employed this method when assessing sweat Fe and found 2- to 24-fold higher concentrations in cell-rich versus cell-poor sweat (Adams et al. 1950; Brune et al. 1986; Foy and Kondi 1957; Hussain and Patwardhan 1959; Prasad et al. 1963).

The potential for skin surface contamination to impact sweat sample concentrations varies depending upon the constituent of interest. Several substances that are of interest for sweat diagnostics are present in or produced by keratinocytes. As shown in Fig. 1, the skin-derived substances likely to confound sweat concentrations include (but are not limited to) Fe²⁺ (Weintraub et al. 1965), Ca²⁺ (Verissimo et al. 2007), K⁺ (Verissimo et al. 2007), cytokines (Dai et al. 2013; Hanel et al. 2013), cortisol (Slominski 2005; Terao and Katayama 2016), and amino acids (Dunstan et al. 2016; Sato et al. 1989). Using nuclear microscopy, Verissimo et al. (2007) obtained quantitative profiles of various elements in skin cross sections. It was found that some minerals, such as Ca²⁺, were abundant in the dead cells of the stratum corneum and less so in the living cells of the deeper layers of the epidermis. This is important because the content of the stratum corneum is likely to have the greatest impact on sweat sample contamination (through mixing of desquamated skin cells with sweat on the skin surface). According to Verissimo et al. (2007), there is ~ 500 to 2500 ng of Ca²⁺ present in a cm² of the stratum corneum. If this amount of Ca²⁺ were incorporated into a 1 mL sample of sweat collected over a 10 cm² surface area (typical absorbent patch size), it would equate to 0.12–0.62 mmol/L of Ca²⁺ contaminating the sweat [Ca²⁺] value. To put this in perspective, the range in sweat [Ca²⁺] typically reported is 0.2 to 2.0 mmol/L (Baker 2019). Based on these estimations, skin surface contamination could artificially elevate sweat [Ca²⁺] by two- to fourfold, which is in line with findings from studies comparing cell-rich vs. cell-poor sweat in Supplemental Table 2 (Boysen et al. 1984; Ely et al. 2011). By contrast, sweat [Na⁺] and [Cl⁻] seem to be minimally impacted by surface contamination (Arn and Reimer 1950; Boysen et al. 1984; Ely et al. 2011; Freyberg and Grant 1937; Robinson and Robinson 1954).

Taken together, it is apparent that the composition of sweat collected from the skin surface is influenced by not only internal sweat secretion mechanisms but also epidermal composition itself. To complicate matters further, the degree of contamination varies depending upon the method/duration of sweat collection and sweating rate. For example, prolonged encapsulation of the skin with occlusive coverings leads to moisture accumulation, which in turn accelerates desquamation. This may explain the very high concentrations of minerals often reported in studies using the arm bag technique (Cohn and Emmett 1978; Collins et al. 1971; Ely et al. 2011; Van Heyningen and Weiner 1952). In addition, for a given amount of mineral or metabolite from epidermal sources, the impact on measured sweat mineral concentration will be influenced by sweating rate such that higher sweat volumes will dilute the effect of epidermal contamination (Costill 1977; Dunstan et al. 2016). Finally, it is also important to consider potential environmental contamination, such as condensation of water vapor onto the skin surface in a sauna (Zech et al. 2015).

Sample storage methodology is another potential source of variability in sweat constituent concentrations and, not surprisingly, one of the most important factors is proper sealing of the sample vials. When well-sealed and stored at room temperature (23–25 °C), refrigeration (4–8 °C), or frozen (− 20 °C) for up to 7 days, minimal sweat evaporation and minimal change in sweat [Na²⁺], [Cl⁻], and [K⁺] has been reported (Baker et al. 2018a; Bergeron et al. 2011; Jones et al. 2008). However, when vials were capped but not sealed with Parafilm-M^® for 3 or 5 days, 19% and 32% sample evaporation has occurred; storing the vials in a plastic bag was associated with 16 and 27% sample evaporation, respectively. This translates to a significant increase in sweat electrolyte concentrations. For example, Bergeron et al. (2011) reported a 12–66% increase in sweat [Cl⁻] after 5 days of storage without Parafilm-M^® sealing.

Differences in sample storage temperature seem to have minimal impact on sweat electrolyte concentrations, as (Baker et al. 2018a) found negligible changes in sweat [Na⁺], [Cl⁻], and [K⁺] between same-day analysis and post-7-day storage (in Parafilm-M^®-sealed vials) at − 20 °C, 8 °C, and 23 °C. There is a paucity of research measuring the effect of storage conditions on other sweat constituents. One study compared amino acid concentrations in aliquots of sweat stored at room temperature (25 °C) versus 37 °C for up to 90 min, prior to longer term storage at − 80 °C (Harshman et al. 2019). The authors reported that, in general, amino acid concentrations were stable at both conditions, with 90% of measurements being within 10% of the control condition (flash freezing). However, within certain individuals there were large differences in sweat alanine, arginine, and threonine concentrations between storage conditions (Harshman et al. 2019).

Because of differences in ease of use and equipment cost, a wide range of analytical techniques have been used by scientists and practitioners to measure sweat composition (Baker 2017). However, caution should be used when comparing results across the literature because sweat [Na⁺] varies significantly among common analytical techniques; in general, ion conductivity > flame photometry ≥ indirect ion-selective electrode > direct ion-selective electrode ≥ ion chromatography (Baker et al. 2014; Boisvert and Candas 1994; Dziedzic et al. 2014; Goulet et al. 2017, 2012). One study directly compared sweat [Na⁺] analyzed with all five different techniques (Goulet et al. 2017). While all techniques had low within-method variability (CVs ≤ 2.6%), there was significant variability between techniques, with ion conductivity producing sweat [Na⁺] values most different (CV = 12.3%) from the reference ion chromatography method (Goulet et al. 2017). Sweat [Cl⁻] and [K⁺] have been assessed in only a few studies, but follow a similar pattern as [Na⁺] with respect to differences among analytical techniques (Baker et al. 2014; Boisvert and Candas 1994). Future research is needed to determine the effect of variations in storage temperature/duration and analytical technique on other sweat constituents.

Na⁺ and its conjugate anion Cl⁻ comprise the most osmotically active components of the extracellular fluid. K⁺ is the major cation in the intracellular fluid compartment. Electrolyte balance plays an important role in governing passive water movement according to osmotic gradients between the intracellular and extracellular water spaces (Mack and Nadel 1996). Because Na⁺ is the primary extracellular osmolyte, ingestion of Na⁺ helps maintain extracellular fluid volume, including plasma volume. Furthermore, an increase serum [Na⁺] and osmolality with Na⁺ ingestion stimulates renal water reabsorption for more complete rehydration (Evans et al. 2017). Individuals with salty sweat (e.g., [Cl⁻] and [Na⁺] ≥ 70 to 80 mmol/L) have an increased risk of NaCl imbalances during prolonged periods of heavy sweating (Baker 2019; Montain et al. 2006). Because of the large inter-individual variability in sweating rate and sweat [Na⁺] personalized fluid/Na⁺ replacement strategies are recommended to maintain fluid and Na⁺ balance during exercise (Belval et al. 2019; McDermott et al. 2017; Sawka et al. 2007).

Very few studies have investigated mechanisms of sweat secretion for any micronutrients other than Na⁺, Cl⁻, and K⁺. As discussed above, research concerning trace elements and vitamins in sweat is confounded by skin cell contamination. Studies that have taken measures to collect cell-poor sweat suggest that final sweat micronutrient concentrations are similar to or less than blood plasma concentrations. Sweat trace element concentrations have been reported to be much more varied than that of the blood plasma (Table 1). Because of the paucity of data, it is unknown whether this is due to varying mechanisms of secretion/reabsorption in the sweat gland or the difficulty in measuring true sweat trace element concentrations untainted by the contents of skin cells.

While Ca²⁺ ions play an important role as a second messenger in the process of sweat secretion, the mechanism of Ca²⁺ movement into the lumen of the secretory coil is not well understood. In a series of studies by Gibinski et al. (1973, 1974), radioisotopes of Ca²⁺, K⁺, and Na⁺ were administered (orally or intravenously) 4–5 h before heat exposure and their concentrations were subsequently measured in sweat. Radio-labeled K⁺ and Na⁺ appeared in sweat within minutes of administration (Gibinski et al. 1973). Radio-labeled K⁺ in sweat was similar to that of blood after ~ 1 h of sweating. By contrast, radio-labeled Ca²⁺ appearance in sweat was slow and did not exceed 5% of blood plasma concentrations (Gibinski et al. 1974). The authors suggested that the binding of Ca²⁺ may explain the difference in the dynamics of Ca²⁺ compared with K⁺ and Na⁺ appearance in sweat. Approximately 50% of plasma Ca²⁺ is bound to protein or complexed with citrate, bicarbonate, or phosphate, while the other 50% is free, ionized Ca²⁺ (Jahnen-Dechent and Ketteler 2012). Similarly, 30–45% of plasma Mg²⁺ is bound or complexed, while 55–70% is free, ionized Mg²⁺ (Jahnen-Dechent and Ketteler 2012). Charged ions such as free Ca²⁺ and Mg²⁺ are hydrophilic but small so they may be secreted readily via the paracellular route. By contrast, protein-bound fractions of Ca²⁺ and Mg²⁺ may not be as readily available for secretion and/or may occur at a slower rate. This may explain why [Ca²⁺] and [Mg²⁺] of cell-poor sweat are more similar to ionized than to total plasma [Ca²⁺] and [Mg²⁺] concentrations (Gibinski et al. 1974).

Most (~ 70%) of plasma Zn²⁺ (albumin) and nearly all (≥ ~ 95%) of plasma Cu²⁺ (ceruloplasmin) and plasma Fe²⁺ (transferrin) are bound to carrier proteins. It is unclear how these trace elements are secreted by clear cells of the eccrine gland. In other cells throughout the body divalent metal transporter 1, divalent cation transporter 1, or other carrier-mediated systems play an important role in moving Fe²⁺, Zn²⁺, and Cu²⁺ across the membrane (Anderson and Frazer 2017; Jahnen-Dechent and Ketteler 2012; McArdle 1992; Roohani et al. 2013). With regard to vitamin secretion in sweat, ascorbic acid and thiamine are large polar molecules (Table 1) and, therefore, may be secreted through a paracellular route albeit exact mechanisms are unknown. Few studies have measured water-soluble vitamin concentrations in sweat (Mickelsen and Keys 1943; Thapar et al. 1976) and in some cases interpretation is difficult due to limited methodologies (scraping) used (Tang et al. 2016). The authors are unaware of any studies reporting sweat concentrations of fat-soluble vitamins. More research is needed to understand mechanisms of trace element and vitamin movement across cell membranes of the eccrine gland secretory coil and duct.

Supplemental Table 3 summarizes the literature comparing the effect of different sweat stimulation methodologies on sweat composition. The results have been mixed and difficult to interpret, due in part to a lack of statistical analysis in some of the early papers. Separate studies have found higher (Ikai et al. 1969), lower (Fukumoto et al. 1988; Kozlowski and Saltin 1964), or similar (Verde et al. 1982) sweat [Na⁺] and [Cl⁻] with exercise versus passive heat stress. Likewise, sweat [Na⁺] and [Cl⁻] have been reported to be higher (Collins 1962; Sato et al. 1970; Shwachman and Antonowicz 1962) or similar (di Sant'Agnese and Powell 1962; Shwachman and Antonowicz 1962) with pharmacological stimulation versus passive heating. As shown in Supplemental Table 3, results were also equivocal for sweat [K⁺], [Ca²⁺], and [Mg²⁺] and no data to the authors’ knowledge are available for Zn²⁺, Cu²⁺, or vitamins. In most of these studies, sweating rate was not standardized or not reported. Even when sweating rate was matched between exercise and passive heating the impact on sweat [Na⁺] and [Cl⁻] was inconsistent (Kozlowski and Saltin 1964; Verde et al. 1982). Some studies collected sweat from different body regions (di Sant'Agnese and Powell 1962) or via different techniques (Shwachman and Antonowicz 1962) between stimulation methods, which further confound the interpretation of results.

Passive and active mechanisms of sweat stimulation have also been compared in studies measuring maximum ion reabsorption rates. Recently, Gerrett et al. (2018) compared the effects of exercise and passive heating on sweat gland ion reabsorption rates on the forearm, chest, and lower back by determining the sweating rate threshold for increasing galvanic skin conductance (Amano et al. 2016). While sweat ion concentrations were not measured, this study found that maximum ion reabsorption rates were higher during moderate-intensity exercise (60% VO_2max for 30 min) than passive heating (immersion of lower legs in 43 °C water bath for 30 min). These results suggest that factors associated with physical activity may influence the rate of Na⁺ and Cl⁻ reabsorption in the eccrine duct (Gerrett et al. 2018). More work is needed to clarify whether exercise mediates enhanced ion reabsorption rates via thermal (body core temperature or mean skin temperature) and/or nonthermal (sympathetic or hormonal) mechanisms.

Discrepancies between methods of sweat stimulation can confound interpretation of studies investigating the impact of various host or external factors on sweat electrolyte concentrations. For example, one study found lower local sweating rates and a higher sweat [Na⁺] on tattooed skin than contralateral non-tattooed skin when stimulated by pilocarpine iontophoresis (Luetkemeier et al. 2017). However, in another study, skin tattoos did not alter local sweating rate or sweat [Na⁺] of exercise-induced sweating (Rogers et al. 2019). As expected, sweating rates were much larger in the exercise study (mean of 1.19 mg/cm²/min) than the pilocarpine iontophoresis study (means of 0.18 and 0.35 mg/cm²/min), which reiterates the mechanistic and practical differences between active and passive pharmacological sweat stimulation.

Taken together, the ecological validity of sweat solute concentrations of pharmacological sweat is questionable. As such, stimulation methodology should be carefully considered when designing research and interpreting sweat composition results. The methods used to induce sweating should fit the purpose of sweat testing and intended application of the data. For example, if the intent is to use the results to inform athletes on their individual sweat electrolyte losses, then the sweat testing should be conducted during exercise and in conditions (i.e., thermal environment) specific/relevant to their sport (Baker 2017).

As shown in Supplemental Table 3, most of the literature suggests that sweat metabolite concentrations are higher with active versus passive sweating. For instance, concentrations of lactate (Astrand 1963; Mitsubayashi et al. 1994), ammonia (Mitsubayashi et al. 1994), amino acids (Liappis and Hungerland 1972), creatinine and urea nitrogen (Fukumoto et al. 1988) have been higher in sweat when stimulated via exercise as compared with passive heating. Most amino acid (Souza et al. 2018), alcohol, carbohydrate, and fatty acid (Delgado-Povedano et al. 2018) concentrations have been higher with exercise than pharmacologically induced sweating. However, there are exceptions, as one study found higher concentrations of lipid mediators in pharmacological sweat than exercise-induced sweat (Agrawal et al. 2018) and another study reported no differences in lactate concentration between exercise and passive heating (Fellmann et al. 1983). Sweat pH does not seem to vary between active and passive sweating, as long as the local environment (ventilation and humidity) is standardized between stimulation methods (Talbert 1919, 1922). Few studies have compared the two types of passive sweating, but available data suggest that lactate (Collins 1962) and amino acid (Souza et al. 2018) concentrations may be higher with passive heating than pharmacological sweat.

With respect to underlying mechanisms, it is interesting to note that the higher sweat lactate concentrations were generally associated with lower sweating rates (Astrand 1963; Collins 1962). This is consistent with the notion that sweat lactate concentrations become diluted with higher volumes of sweat (Buono et al. 2010). However, many of the other studies did not standardize sweating rate and/or anatomical location of sweat collection between methods. It is, therefore, difficult to determine the reason for higher concentrations of other metabolites with active versus passive sweating.

Only one study has compared the effects of different stimulation techniques on sweat cytokine concentrations. Didierjean et al. (1990) measured IL-1α and IL-1β in sweat collected from the hand after stimulation via passive heating, sauna bathing, or spontaneous sweating. Both IL-1 cytokines were significantly higher with passive heating (Supplemental Table 3), which was also associated with maximum sweat induction (Didierjean et al. 1990). Because epidermal contamination was avoided (via use of the anaerobic technique with a petroleum barrier), the authors concluded that at least some of the IL-1 may have originated from the sweat glands themselves (Didierjean et al. 1990). However, it is unclear whether the differences in sweat IL-1 concentrations were due to the method of stimulation or factors (e.g., increased gland stress) associated with profuse sweating.

It is well established that heat acclimation results in a significant enhancement in the rate of sweat secretion by eccrine glands, resulting in improved tolerance to passive and active heat stress (Allan and Wilson 1971; Kirby and Convertino 1986; Pandolf et al. 1988). Partial acclimation occurs with passive heat stress or exercise training, but exposure to both exercise and heat stress is required to achieve full heat acclimation (Saat et al. 2005; Tipton et al. 2008). In addition, heat acclimation is usually associated with a decrease in sweat [Na⁺] and [Cl⁻]. The decrease in sweat [Na⁺] and [Cl⁻] begins after 2–3 consecutive days of heat exposure and continues over time (Buono et al. 2018; Karlsen et al. 2015), resulting in an up to 30–60% decrease after 7–10 days (Allan and Wilson 1971; Buono et al. 2007, 2018; Chinevere et al. 2008; Johnson et al. 1944; Karlsen et al. 2015; Kirby and Convertino 1986; Nielsen et al. 1997; Robinson et al. 1950). Seasonal variation in sweat [Na⁺] has also been reported to be ~ 30 to 60% decrease from winter to summer (Bates and Miller 2008; Inoue et al. 1995). The linear relation between sweat flow rate and sweat [Na⁺] (discussed above) is maintained with heat acclimation. However, there is a downward shift in the regression line such that at any given sweating rate on the forearm, heat acclimation results in significantly lower forearm sweat [Na⁺] (Buono et al. 2007). The decrease in sweat [Na⁺] and [Cl⁻] despite an increase in sweating rate can be explained by the contrasting effects of acute changes in sweat flow rate versus the longer term adaptations in the sweat gland that occur with heat acclimation. The physiological mechanism underlying the decreased sweat [Na⁺] and [Cl⁻] with heat acclimation is related to alterations in the hormonal control of Na⁺ reabsorption by aldosterone, possibly increased sensitivity of the eccrine gland to circulating aldosterone concentrations (Kirby and Convertino 1986). As described above, aldosterone influences Na⁺ reabsorption by increasing the activity of Na⁺–K⁺-ATPase on the basolateral membrane in the eccrine sweat duct (Ladell and Shephard 1961; Sato and Dobson 1973).

It is important to note that a salt deficit is required to stimulate enhancement in NaCl reabsorption. Studies have found no change or a marginal increase in sweat [Na⁺] and [Cl⁻] when salt intake is sufficient to replace sweat electrolyte losses incurred during the heat acclimation protocol, (Armstrong et al. 1985a; Eichner 2008; McCance 1938; Weiner and Van Heyningen 1952). This finding is in line with the notion that NaCl conservation by the sweat glands is mediated by circulating aldosterone. Moreover, Yoshida et al. (2006) found that individual variations in sweat [Na⁺] during exercise were correlated with resting plasma aldosterone concentration, but not to plasma aldosterone during exercise. Therefore, the genomic action of aldosterone (influenced by chronic NaCl balance as a result of heat acclimation and diet) may have a stronger impact on inter-individual variations in sweat [Na⁺] than the rapid non-genomic action of aldosterone (influenced by acute exercise and dehydration) (Yoshida et al. 2006). This concept is supported by two recent studies that induced plasma aldosterone changes via dietary Na⁺ restriction for 3–5 days (Braconnier et al. 2019; McCubbin et al. 2019). McCubbin et al. (2019) reported a significant negative correlation between resting pre-exercise plasma aldosterone and sweat [Na⁺] measured during a subsequent bout of 2-h exercise in endurance athletes. Braconnier et al. (2019) also found that [Na⁺] of passive sweat (stimulated via pilocarpine iontophoresis) was negatively correlated with plasma aldosterone concentration in healthy normotensive participants.

The changes in sweat [Na⁺] and [Cl⁻] with altered dietary salt intake have been extensively reviewed in recent publications (Baker 2019; McCubbin and Costa 2018) and, therefore, will not be discussed in detail here. In brief, most (Allsopp et al. 1998; Armstrong et al. 1985a; Braconnier et al. 2019; Costa et al. 1969; Hargreaves et al. 1989; Komives et al. 1966; McCance 1938; McCubbin et al. 2019; Sigal and Dobson 1968; Weiner and Van Heyningen 1952) but not all (Costill et al. 1975; Koenders et al. 2017; Konikoff et al. 1986; Robinson et al. 1956) studies have shown that dietary Na⁺ restriction is associated with a decrease in sweat [Na⁺] and [Cl⁻]. The mixed results may be explained in part by methodological differences among studies, including the duration and degree of dietary manipulation. Changes in sweat [Na⁺] and [Cl⁻] seem to be less likely when salt consumption is altered by smaller (and perhaps more realistic) amounts (Costill et al. 1975; Koenders et al. 2017) or for a short period of time (less than 3 days, including just before/during exercise) (Hamouti et al. 2012; Koenders et al. 2017; Konikoff et al. 1986; Robinson et al. 1956, 1955). This result is perhaps not surprising based on the time course of sweat gland responsiveness to changes in aldosterone and the notion that genomic effects of aldosterone on sweat [Na⁺] are stronger than nongenomic actions.

Relatively few studies have tested the effect of physical training on sweat composition. This may be due in part to the difficulty in separating the effects of training from heat acclimation, since regular physical exercise elicits partial heat acclimation. The available studies suggest that aerobic training is associated with an increased sweat flow rate related to an increased cholinergic sensitivity and decreased threshold for sweat onset (Araki et al. 1981; Buono et al. 1991; Buono and Sjoholm 1988; Greenleaf et al. 1972; Inoue et al. 1999). However, the effect on sweat [Na⁺] and [Cl⁻] is less clear because no longitudinal studies are available. To the authors’ knowledge, mostly cross-sectional studies comparing groups with different aerobic capacities have been conducted to date. For example, Araki et al. (1981) measured sweat [Cl⁻] from the upper back of trained and untrained women during 2 h cycling at fixed absolute workloads (79 and 160 Watts) in a hot, humid, still-air environment. Sweat [Cl⁻] was significantly lower in trained participants at both work rates in both winter and summer test sessions. However, the authors also reported hidromeiosis in the trained group, which could explain in part their lower sweat [Cl⁻] (Araki et al. 1981). In another study, Henkin et al. (2010) measured sweat [Na⁺] and [Cl⁻] from the scapula of swimmers (VO_2max  = 54.2 ± 5.7 ml/kg/min), runners (60.5 ± 5.8 ml/kg/min), and non-athletes (45.2 ± 2.9 ml/kg/min) during 30 min cycling in the heat at a fixed relative intensity of 65–75% maximal heart rate. Despite the significantly higher aerobic capacity and 50% higher sweating rate of the swimmers compared with the non-athletes there were no differences in sweat [Na⁺] or [Cl⁻]. By contrast, sweat [Na⁺] and [Cl⁻] were significantly lower in the runners than swimmers and non-athletes (Henkin et al. 2010). This may suggest potential NaCl conservation by the sweat glands with training, but the confounding effect of partial heat acclimation cannot be ruled out in this study since it was conducted in Brazil in the later winter where outdoor temperature reached 24 °C (Henkin et al. 2010).

Hamouti et al. (2011) measured sweat [Na⁺] from the lower back of trained (VO_2peak = 4.0 ± 0.8 L/min) and untrained (VO_2peak = 3.4 ± 0.7 L/min) participants during three bouts (40, 60, and 80% VO_2peak) of cycling in the heat. As expected, sweat [Na⁺] increased with an increase in workload for both trained and untrained participants. Sweat [Na⁺] tended to be higher in trained versus untrained participants at the two higher workloads. However, local sweating rate was also higher in the trained participants (since absolute workloads were higher in trained vs. untrained). When sweat [Na⁺] was normalized for local sweating rate (which accounts for lack of standardizing absolute workload), there were no differences between groups (Hamouti et al. 2011), thus suggesting that higher sweat [Na⁺] in trained individuals was a function of higher sweat flow rate rather than higher sweat [Na⁺] per se. Taken together these cross-sectional studies suggest that elevated aerobic fitness is not associated with enhanced Na⁺ reabsorption in the sweat gland. Similar conclusions have been drawn from a sweat gland training study involving 2 h of sweating induced by twice daily intradermal injection of acetyl-β-methylcholine for 10–18 days (Johnson et al. 1969).

Recently, Amano et al. (2017) measured maximum sweat ion reabsorption rates (Amano et al. 2016) during exercise in distance runners (VO_2max = 59.1 ± 1.4 ml/kg/min), sprinters (VO_2max = 43.3 ± 1.2 ml/kg/min), and untrained (VO_2max = 38.0 ± 2.2 ml/kg/min) participants. They found enhanced maximum sweat ion reabsorption rates on the back region of distance runners and sprinters compared with untrained individuals, but no differences between distance runners and sprinters (Amano et al. 2017). Furthermore, there were no differences in sweat ion reabsorption rates among groups on the forearm and actual sweat ion concentrations were not reported (Amano et al. 2017). Considering the mixed results among studies and paucity of longitudinal data, more research is needed to clarify the effects of physical training on sweat composition. Future research should assess sweat electrolyte concentrations with repeated physical training in a longitudinal study design, as parallel design studies comparing well-trained to moderately- or poorly trained participants cannot control for potential confounding factors that may impact sweat composition.

Another question of interest is how do sweat electrolyte concentrations change throughout the course of a single bout of physical activity? Prolonged heavy sweating and humid environments associated with elevated skin wettedness can lead to hidromeiosis, a condition that causes a gradual decline in sweating rate (Collins and Weiner 1962). The mechanism of suppressed sweat flow rate with hidromeiosis is thought to be due to mechanical occlusion of sweat ducts as a result of swelling of the keratinized layer around the duct (Brown and Sargent 1965). It is unclear whether sweat suppression impacts sweat [Na⁺], [Cl⁻], or [K⁺], as most hidromeiosis studies have not included sweat electrolyte measurements. Interestingly, with heat acclimation the sweat glands become resistant to hidromeiosis such that higher sweating rates can be maintained (Ogawa et al. 1982; Taylor 2014). It is also important to note that studies have reported no decline in sweating rate and no change in sweat [Na⁺] or [K⁺] throughout 3–7 h of exercise of low intensity (e.g., walking) and light sweating (0.3–0.6 L/h) (Ely et al. 2011; Montain et al. 2007).

The effects of heat acclimation, physical training, and diet on sweat trace mineral concentrations have been studied to a lesser extent than Na⁺ and Cl⁻. Some initial studies seemed to indicate a conservation of sweat trace mineral concentrations with heat acclimation (Chinevere et al. 2008; Hoshi et al. 2002; Klesges et al. 1996) and even throughout the course of a single bout of exercise (DeRuisseau et al. 2002; Montain et al. 2007; Waller and Haymes 1996). However, Ely et al. (2013) determined that the decline in sweat mineral concentrations in previous studies was likely an artifact of epidermal contamination when using the arm bag technique and/or not pre-washing/cleaning the skin at the site of collection (Ely et al. 2011). It may be that progressive flushing of mineral residue lying on the skin surface with repeated profuse sweating may have contributed to the decrease in sweat mineral concentrations in previous studies (Ely et al. 2013, 2011). There is little if any information available to suggest that physical training impacts sweat trace mineral or vitamin concentrations. Several studies have investigated the impact of acute supplementation or chronic dietary intake of trace minerals and vitamins on sweat composition. Most studies, particularly those in healthy individuals with no known mineral or vitamin deficiencies, have reported no association between dietary intake and sweat concentrations for Zn²⁺, Fe²⁺, Ca²⁺, Cu²⁺ or ascorbic acid (DeRuisseau et al. 2002; Jacob et al. 1981; Lugg and Ellis 1954; Mitchell and Hamilton 1949; Vellar 1968b; Wheeler et al. 1973).

The effects of heat acclimation or physical training on cytokine, immunoglobulin, or cortisol in human sweat have not been researched. A few studies have investigated sweat lactate, amino acid, and urea albeit with equivocal results regarding the impact of heat acclimation and physical training on these metabolites. Some (Fellmann et al. 1983; Lamont 1987; Pilardeau et al. 1988), but not all (Green et al. 2004) studies suggest that a higher fitness level is associated with lower sweat lactate concentrations. However, these results may be confounded by differences in sweating rate between groups. Lower sweat gland metabolism with lower sweat flow rates would result in less lactate production (Pilardeau et al. 1988). On the other hand, there could be a dilution effect of increased sweating rates (with physical training and/or increased absolute exercise intensity) on sweat lactate concentrations (Lamont 1987). Similarly, Liappis et al. (1979) found that total amino acid concentrations in sweat were significantly lower in trained (physically active for an average of 9 h per week) than untrained (average of 1 h of sports per week) men during 15 min of cycling at 150 W while covered with a plastic blanket (12,797 vs. 24,855 µmol/L, respectively). The authors speculated that regular stimulation of sweating via physical training caused an adaptation in the sweat glands to limit excretion of essential amino acids (Liappis et al. 1979). However, sweating rate was not standardized between groups, as sample volume collected from the forearms ranged from 1 to 6 mL in 15 min (Liappis et al. 1979). Therefore, it is possible that the more fit individuals had a higher sweat volume that could have diluted the amino acid concentrations.

Limited data on heat acclimation and sweat metabolite concentrations are available. Weiner and Heyningen (1949) found that lactate concentrations in arm bag sweat decreased with 20 -day acclimation in one participant. However, in a follow-up study, the investigators found no changes in whole body lactate concentration with heat acclimation, despite an increase in sweating rate (Weiner and Van Heyningen 1952). The literature on heat acclimation and sweat urea has reported mixed results between studies (Robinson and Robinson 1954) and significant inter-individual variability (McCance 1938). To date, no study has investigated the effects of heat acclimation or physical fitness on sweat ammonia concentrations. Physical strain and muscular activity are known to increase blood ammonia due to metabolic degradation and purine nucleotide cycle activity (Mitsubayashi et al. 1994; Schulz and Heck 2003). Endurance-trained individuals produce less ammonia during submaximal exercise (Holloszy and Coyle 1984). In addition, since ammonia is thought to passively diffuse from the plasma into sweat (Sato et al. 1989), it may be logical to hypothesize that sweat ammonia concentrations should be lower in trained versus untrained participants. Still, this idea is speculative, and confounded by the impact of sweating rate and possible ammonia contamination from apocrine and sebaceous glands. More work is needed to understand how sweat ammonia, lactate, and other metabolites are impacted by changes in physical training and heat acclimation if they are to be used as a biomarker for fitness, training, or performance purposes.

Supplemental Table 4 shows a summary of published studies comparing constituent concentrations in sweat versus blood. Although there is a long list of studies, few of them were well-controlled, adequately powered, and reported correlation and statistical results. Three of six studies suggested there was a significant positive correlation between sweat and blood [Na⁺] or [Cl⁻], but only for a portion of the data set (Hew-Butler et al. 2010; Robinson et al. 1956; Talbert and Haugen 1927). The other three studies found no correlation between sweat and blood [Na⁺] or [Cl⁻] (Johnson et al. 1944; McCubbin et al. 2019; Mickelsen and Keys 1943). Thus, based on the available literature to date, there is no clear evidence of a correlation between sweat and blood for [Na⁺] and [Cl⁻]. This is perhaps not surprising because, even though primary sweat is isotonic with the blood, the rate of reabsorption of Na⁺ and Cl⁻ in the eccrine duct is independent of plasma electrolyte concentrations. For example, acute changes in sweating rate alone can result in up to threefold changes in final sweat [Na⁺] (Buono et al. 2008). Additionally, in a well-controlled study in 15 male endurance athletes, McCubbin et al. (2019) found a significant correlation between pre-exercise plasma aldosterone and sweat [Na⁺] but no correlation between plasma [Na⁺] and sweat [Na⁺] (McCubbin et al. 2019). This and other studies (discussed above) suggest that factors impacting resting plasma aldosterone play an important role in determining final sweat [Na⁺] (Braconnier et al. 2019; Yoshida et al. 2006).

Few studies have compared concentrations of other micronutrients in sweat and blood; nonetheless, most have reported no correlations for Fe (Paulev et al. 1983; Vellar 1968a) or ascorbic acid (Mickelsen and Keys 1943). This may be due in part to the methodological issues, such as inconsistency in accounting for skin contamination. Interestingly, the only significant correlation reported was between cell-free sweat [Fe²⁺] and serum [Fe²⁺] albeit the correlation coefficient was small (r = 0.37) (Vellar 1968a). These results taken together with the lack of association between dietary micronutrient intake and corresponding sweat micronutrient concentrations suggest little support for using sweat as a surrogate for blood.

Some have suggested that sweat electrolyte concentrations can be used as a biomarker to detect dehydration or predict sweating rate (Alizadeh et al. 2018; Gao et al. 2016; Rose et al. 2015; Zhou et al. 2016); however, studies measuring sweat electrolyte concentrations alongside changes in hydration status have found mixed results. Sweat [Na⁺], [Cl⁻], and [K⁺] have been shown to increase (Lichton 1957; Morgan et al. 2004), decrease (Armstrong et al. 1985b; Costill et al. 1976), or not change (Amatruda and Welt 1953; Morgan et al. 2004; Robinson et al. 1956; Walsh et al. 1994) in response to dehydration during exercise and/or heat stress. Furthermore, studies have demonstrated significant changes in sweat [Na⁺], [Cl⁻], and/or [K⁺] in response to variations in exercise intensity (Baker et al. 2019), environment (Dziedzic et al. 2014), or diet (McCubbin et al. 2019) despite no significant differences in hydration status during exercise. Moreover, as discussed above, the correlation between sweating rate and sweat [Na⁺] for between-participant comparisons is poor (Baker et al. 2018b; Patterson et al. 2000).

The correlation between blood and sweat metabolite concentrations are of recent interest due to novel biotechnological advances enabling local metabolite detection from micro-sampling devices (Hauke et al. 2018; Heikenfeld et al. 2018). Supplemental Table 4 summarizes studies that have tested the utility of sweat as a biomarker for changes in blood metabolite and cytokine concentrations. The results for urea, ammonia, bicarbonate, amino acids, and glucose have been mixed at best. Moyer et al. (2012) found a significant correlation between sweat and blood glucose concentration in diabetic patients in whom the range in blood glucose was quite broad (~ 60 to 360 mg/dl). Another study found a significant correlation in diabetic patients, but not healthy participants (Nyein et al. 2019). These results suggest that sweat glucose may be sensitive to large changes in blood glucose, but perhaps not to relatively small changes (e.g., in non-diabetics). In fact, there is currently little evidence of a correlation between sweat and blood for glucose or other metabolites such as lactate, amino acids, and urea in healthy individuals at rest or during exercise (Supplemental Table 4). This is due in part to significant methodological issues, such as sub-optimal analytical technique (Silvers et al. 1928), very low number of participants, or unclear/incomplete analysis and reporting of results (see Supplemental Table 4). One study (Alvear-Ordenes et al. 2005) did find a significant correlation between sweat and blood in rugby players, but metabolite concentrations in sweat only accounted for 7% (ammonia) or 45% (urea) of the variation in blood concentrations. More well-controlled correlation studies are needed.

Several studies listed in Supplemental Table 4 did not report correlation results but observed the change in sweat metabolite concentration after manipulation of blood concentrations via substrate ingestion (glucose, amino acids, ammonium chloride, urea, or bicarbonate) or having participants perform exercise of various intensities (lactate, ammonia). In the ingestion studies, an increase in blood metabolite concentrations was sometimes (Boysen et al. 1984; Czarnowski et al. 1992; Komives et al. 1966) but not always (Ament et al. 1997; Hier et al. 1946; Patterson et al. 2002) associated with a significant concomitant rise in sweat concentrations. In one study, significant increases in sweat glucose concentration were observed in two participants in response to infusion and ingestion of a glucose bolus, which caused an increase in blood glucose from 60 to 360 mg/dl (Boysen et al. 1984). Again, it is still unclear if smaller changes in blood glucose would elicit measurable concomitant changes in sweat glucose. An increase in blood lactate and ammonia concentrations during exercise has been associated with a decrease (Ament et al. 1997) or no change (Alvear-Ordenes et al. 2005; Green et al. 2000; Weiner and Van Heyningen 1952) in corresponding sweat lactate and ammonia concentrations. These and other studies (Buono et al. 2010) illustrate the lack of support for using sweat lactate or ammonia as biomarker for exercise intensity.

On the other hand, a few studies have consistently reported a strong correlation (Gamella et al. 2014; Hauke et al. 2018; Phillips and McAloon 1980) and close to 1:1 ratio (Buono 1999) for ethanol concentrations in sweat compared with blood. This may be explained by the straightforward mechanism of ethanol secretion (passive diffusion, as discussed above) and the low likelihood of surface contamination or eccrine gland production. Therefore, sweat ethanol could possibly be used as surrogate for blood ethanol concentration albeit breathalyzers are already available as a non-invasive practical alternative to blood. Finally, there is some interest in sweat cytokines and cortisol as biomarkers for health and wellness (inflammation, stress, etc.). Two studies by the same group (Cizza et al. 2008; Marques-Deak et al. 2006) have found significant correlations between 24-h insensible sweat and plasma cytokine concentrations. More work is needed, however, to corroborate these findings and determine the utility of sweat cytokines for monitoring healthy individuals or patient populations. No studies have compared cortisol concentrations in sweat and blood.