Sweating Rate and Sweat Sodium Concentration in Athletes: A Review of Methodology and Intra/Interindividual Variability

Athletes lose water and electrolytes as a consequence of thermoregulatory sweating during exercise and it is well known that the rate and composition of sweat loss can vary considerably within and among individuals. Many scientists and practitioners conduct sweat tests to determine sweat water and electrolyte losses of athletes during practice and competition. The information gleaned from sweat testing is often used to guide personalized fluid and electrolyte replacement recommendations for athletes; however, unstandardized methodological practices and challenging field conditions can produce inconsistent/inaccurate results. The primary objective of this paper is to provide a review of the literature regarding the effect of laboratory and field sweat-testing methodological variations on sweating rate (SR) and sweat composition (primarily sodium concentration [Na+]). The simplest and most accurate method to assess whole-body SR is via changes in body mass during exercise; however, potential confounding factors to consider are non-sweat sources of mass change and trapped sweat in clothing. In addition, variability in sweat [Na+] can result from differences in the type of collection system used (whole body or localized), the timing/duration of sweat collection, skin cleaning procedure, sample storage/handling, and analytical technique. Another aim of this paper is to briefly review factors that may impact intra/interindividual variability in SR and sweat [Na+] during exercise, including exercise intensity, environmental conditions, heat acclimation, aerobic capacity, body size/composition, wearing of protective equipment, sex, maturation, aging, diet, and/or hydration status. In summary, sweat testing can be a useful tool to estimate athletes’ SR and sweat Na+ loss to help guide fluid/electrolyte replacement strategies, provided that data are collected, analyzed, and interpreted appropriately.


Introduction
During exercise, water and electrolytes are lost as a consequence of thermoregulatory sweating. In some situations, especially when exercise is prolonged, high-intensity, and/ or in a hot environment, sweat losses can be sufficient to cause excessive water/electrolyte imbalances and impair performance [1][2][3][4][5]. It is well-established that sweating rate (SR) and sweat electrolyte concentrations can vary considerably as a result of many within-and between-athlete factors (i.e. natural or expected sources of variability); therefore, personalized fluid replacement strategies are recommended [2,6,7]. In accordance with these guidelines, many scientists and practitioners have conducted sweat tests with athletes [8][9][10][11][12][13][14][15][16][17]; however, many different methodologies have been used, which could be another source of (undesirable) variability in SR and sweat electrolyte concentrations. For example, sweat testing can be conducted using whole-body techniques or localized to a specific anatomical site. Furthermore, methods to measure sweat electrolyte concentration can vary in the type of collection system used, the timing/duration of sweat collection, skin cleaning procedure, sample storage/handling, and analytical technique. The use of invalid or inconsistent methods related to any of these factors can lead to significant background noise, errors, and/or misinterpretation of results [18,19]. However, there are few reviews available on sweat testing athletes, particularly with respect to personalized fluid replacement and best practices in the field.
The primary purpose of this paper is to review the methodological considerations for sweat-testing athletes. In addition, this review will briefly discuss within and between-subject factors that impact the variability in SR and sweat composition. Although many electrolytes and other constituents are lost in sweat, this review will primarily focus on sweat sodium concentration ([Na ? ]) since it is the electrolyte lost in the greatest quantities and has the most significant impact on body fluid balance (through the effects of Na ? on fluid retention and plasma volume maintenance/restoration) [1]. Finally, based on the information gleaned from this literature review, recommendations regarding best practices for sweat testing in the field, as well as considerations for interpretation and practical application of results, will be proposed.

Eccrine Sweat Glands and Thermoregulatory Sweating
During exercise, a large amount of heat is produced by the contracting muscles as a byproduct of metabolism, leading to body heat gain. In addition, if ambient temperature is greater than skin temperature (T sk ), heat is transferred from the air to the body. The resultant increase in body core temperature (T c ) is sensed by central and skin thermoreceptors and this information is processed by the preoptic hypothalamic region of the brain to stimulate sweating and cutaneous vasodilation to dissipate heat [20][21][22]. Evaporation of sweat is the primary avenue of heat loss during exercise. With sweating, heat is transferred from the body to water (sweat) on the surface of the skin. When this water gains sufficient heat, it is converted to water vapor, thereby removing heat from the body (580 kcal of heat per 1 kg of evaporated sweat) [23,24]. Sweat glands are classified into three main types: apocrine, apoeccrine, and eccrine [25,26]. Apocrine and apoeccrine glands are limited to certain regions of the body (e.g. the axillae region) and do not become active until puberty [27][28][29][30]. Eccrine sweat glands are located across most of the body surface, are primarily responsible for thermoregulatory sweating [25], and therefore will be the focus of this review. Humans have approximately 2-3 million eccrine sweat glands and this number is fixed by approximately 2-3 years of age [31,32]. Sweat gland density decreases with skin expansion during growth and is generally inversely proportional to body surface area (i.e. larger or more obese individuals have lower sweat gland density than their smaller counterparts) [33]. SR over the whole body is a product of the density of active sweat glands and the secretion rate per gland. At the onset of sweating (i.e. upon reaching the T c set point), the initial response is a rapid increase in sweat gland recruitment, followed by a more gradual increase in sweat secretion per gland [31,[34][35][36]. Most of the intra-and interindividual variability in steady-state SR is due to differences in sweat secretion rate per gland, rather than the total number of active sweat glands or sweat gland density. With habitual activation, sweat glands show some plasticity in their size and neural/hormonal sensitivity, which in turn impact SR and sweat [Na ? ] [37][38][39].
Eccrine sweat glands primarily respond to thermal stimuli, particularly increased T c [40] and also T sk and associated increases in skin blood flow [25,41,42]. Sweating is mediated predominately by sympathetic cholinergic stimulation; the nerves surrounding eccrine sweat glands are nonmyelinated class C sympathetic postganglionic fibers and acetylcholine is the primary terminal neurotransmitter [25]. Eccrine glands also secrete sweat in response to adrenergic stimulation, but to a much lesser extent (approximately 10%) than that of cholinergic stimulation [29,43]. Catecholamines, as well as other neuromodulators, such as vasoactive intestinal peptide, calcitonin gene-related peptide, and nitric oxide, have also been found to play minor roles in the neural stimulation of eccrine sweating [24,25,44]. In addition, eccrine sweat glands respond to non-thermal stimuli related to exercise (i.e. in the absence of or prior to changes in T c or T sk ). These non-thermal sudomotor responses are thought to be mediated by feed-forward mechanisms related to central command, the exercise pressor reflex (muscle metabo-and mechanoreceptors), osmoreceptors, and possibly baroreceptors [44,45].
The structure of the eccrine sweat gland consists of a secretory coil and duct made up of a simple tubular epithelium. Upon stimulation, primary or precursor sweat, which is an ultrafiltrate of the plasma, is secreted by clear cells of the secretory coil. Primary sweat is nearly isotonic with blood plasma (e.g. approximately 135-145 mmol/L Na ? , approximately 95-110 mmol/L Cl -, and approximately 4-5 mmol/L K ? ) [29,[46][47][48][49]. As sweat flows through the duct, Na ? is passively reabsorbed via epithelial Na ? channels (ENaCs) on the luminal membrane and actively reabsorbed via Na ? /K ? -ATPase transporters primarily on the basolateral membrane [25,50]. Chloride (Cl -) is passively reabsorbed via the cystic fibrosis transmembrane conductance regulator (CFTR) on the luminal and basolateral ductal cell membrane [25,51]. The result is a hypotonic (with respect to Na ? and Cl -) final sweat excreted onto the skin surface [25,29]. Na ? /K ? -ATPase activity is influenced by the hormonal control of aldosterone [52]. The rate of Na ? and Clreabsorption is also flow dependent, such that there is a direct relation between SR and final sweat [Na ? ] and [Cl -] [29]. As SR increases, the rate of Na ? and Clsecretion in precursor sweat increases proportionally more than the rate of Na ? and Clreabsorption along the duct, and therefore leads to higher final sweat [Na ? ] and [Cl -] [29,53]. The rate of sweat potassium (K ? ) loss has been reported to be indirectly related to sweat flow rate, but the underlying mechanism is unclear [29]. Nonetheless, final sweat typically has a [K ? ] similar, albeit with a slightly broader range (e.g. approximately 2-8 mmol/L), to that reported for blood plasma [54][55][56]. More details on the structure, function, and control of eccrine sweat glands can be found in several comprehensive reviews [19,26,29,32,44,57].

Sweating Rate
Local methods to measure SR include hygrometry and gravimetry [19]. With hygrometry, or the ventilated sweat capsule technique, dry air with a known temperature is pumped at a constant flow rate through a capsule affixed to the skin. Local SR (LSR) onto the skin surface under the capsule (approximately 1-20 cm 2 ) is determined from the change in the temperature and water vapor content of the effluent compared with the influent air of the capsule [58,59]. The ventilated capsule technique is highly reliable (coefficient of variation [CV] of 2%) [60] and is considered the reference technique for measuring LSR [60][61][62]. However, hygrometry can overestimate sweat flow rates because the forced ventilation and maintenance of dry skin (which facilitates sweating) under the capsule is not representative of ambient conditions in some situations (e.g. the microclimate under clothing, or exercise in humid or still air).
Gravimetric techniques involve the collection of sweat directly from the skin surface (approximately 4-100 cm 2 ) using filter paper [63], absorbent patches [18,54,64,65], Parafilm-M Ò pouches [66,67], cotton gloves/socks [68], latex gloves [68], or plastic sweat collectors [69]. With these methods, LSR is determined from the mass change of the collection system. Gravimetric techniques, particularly absorbent patches, are more practical than hygrometry for sweat-testing athletes in the field. Nonetheless, a limitation of gravimetry is that the collection system can modify the local environment and consequently alter the flow rate of sweat onto the skin surface. The lack of ventilation caused by an occlusive covering increases moisture accumulation on the skin, thereby leading to progressive blocking of sweat ducts and sweat suppression (i.e. hidromeiosis [70][71][72][73][74][75]). However, it has been proposed that hidromeiosis can be minimized by limiting the duration of the collection system on the skin and/or using patches made of a material with a high absorbent capacity [19,64].
Two studies have compared LSR results using gravimetry versus hygrometry during steady-state cycling. Morris et al. [61] reported significantly (6-37%) greater LSR with ventilated capsules (4 cm 2 ) than absorbent patches after 10 and 30 min, but there were no differences after 50 and 70 min of exercise. The results were not impacted by the anatomical site (forearm and midback) or absorbent patch size (4 cm 2 and 36-42 cm 2 ) [61]. Boisvert et al. [62] reported that LSR was significantly (27%) greater with ventilated capsules versus the Parafilm-M Ò pouch technique in the first 20 min of exercise, but there were no differences from 20 to 60 min [62]. Several studies have also reported strong correlations between LSR at various sites across the body (using either gravimetry or hygrometry) and whole-body SR (WBSR) during exercise [64,76,77]. Taken together, it seems that gravimetric techniques are a reliable, portable, cost-effective alternative to hygrometry for measuring the rate of sweat appearance on the skin surface, but only after steady-state sweating has been established (e.g. 20-30 min into exercise) [61,62].
The simplest and most accurate method to assess WBSR is via changes in nude body mass from before to after exercise [2,4,78]; however, corrections for non-sweat sources of body mass change should be considered. The SR calculation should be corrected for fluid intake and urine output. In addition, if athletes consume food or void their bowels during the training session, these non-sweat body mass changes should also be taken into account. It is also important to note that a portion of body mass loss during exercise occurs due to metabolic mass loss (substrate oxidation) and respiratory water loss (approximately 5-15% combined [54,[79][80][81][82][83]). SR calculations based on change in body mass should be corrected for metabolic mass loss and respiratory water loss, particularly when exercise lasts several hours (e.g, [2-3 h) [2], is high-intensity, and/or is performed in a cool/dry environment [79,82,83].
Oftentimes, nude body mass measurements are not practical in the field, thus athletes are weighed while wearing clothing; however, trapped sweat in clothing can lead to underestimations of SR. Cheuvront et al. [83] tested women wearing a racing singlet, shorts, socks, and running shoes during a 30-km treadmill run at 71% of maximal oxygen uptake in warm (30°C dry bulb temperature, 20°C dew point temperature) or cool (14°C dry bulb temperature, 7°C dew point temperature) conditions (2.1 m/s wind). When corrections were not made, trapped sweat in clothing caused an 8-10% underestimation of sweat loss (no difference between warm and cool conditions), while urine loss caused an overestimation of sweat loss by 16% (warm) or 37% (cool), and combined respiratory water loss and metabolic mass loss led to an overestimation of sweat loss by 9% (warm) or 20% (cool) [83].
For more information on determining WBSR, the reader is referred to detailed, step-by-step instructions described by Armstrong and Casa [84]. In addition, the following papers provide equations to calculate mass loss from substrate oxidation [79,85,86] and respiratory water [79,85]. No equations are currently available to correct for trapped sweat in different clothing/uniform ensembles.

Sweat [Na 1 ]
The reference method for determining sweat Na ? loss is whole-body washdown (WBW) [84,87,88]. The recovery of water and Na ? using the WBW method has been reported to be 102 ± 2% and 99 ± 2%, respectively [87]. The WBW method is considered the most accurate measure of wholebody sweat electrolyte loss because all sweat runoff is collected and accounted for and it does not interfere with the normal evaporative sweating process; however, WBW requires a controlled laboratory setting and the mode of exercise is primarily limited to stationary cycling (for more methodological details see Armstrong and Casa [84] and Shirreffs and Maughan [87]). Thus, local methods for estimating sweat [Na ? ] are more commonly used because they are relatively simple and more practical for field studies. Local sweat [Na ? ] methods are similar to that of the gravimetric LSR techniques discussed in Sect. 3.1, i.e. filter papers [89,90], absorbent patches [91,92], Parafilm-M Ò pouches [66,67], arm bags/gloves [91,93,94], plastic sweat capsules [90,95], and sweat collectors [69]. Quality-control analyses have reported a high percentage of Na ? recovery (99% [92]) and negligible background Na ? (approximately 0-3 mmol/L [92,96]) in the absorbent patch technique. Still, local sweat [Na ? ] is usually not a valid direct surrogate for whole-body [Na ? ] due in part to the creation of a microenvironment (i.e. increased local humidity, skin wettedness, and possibly T sk ). The impact of hidromeiosis on LSR and, in turn, on sweat [Na ? ], has been discussed in Sect. 3.1. In addition, leaching of electrolytes (from the stratum corneum of the skin to the local sweat sample) and/or absorption of water (from the sweat into the stratum corneum) can lead to falsely high sweat electrolyte (including Na ? and K ? ) concentrations from samples collected within occlusive coverings [27,88,97] [54,55,87]. This regional variation in sweat [Na ? ] can be explained in part by regional variations in LSR. Not surprisingly, inter-regional differences in LSR and sweat [Na ? ] have been reported to follow the same general pattern (e.g. [55]. In addition, the arm bag or rubber glove technique tends to overestimate sweat electrolyte concentrations to a greater extent than other local methods [88,98,99]. Nonetheless, studies have shown that local sweat [Na ? ] is highly and significantly correlated with WBW sweat [Na ? ], thus regression equations are available to estimate WBW sweat [Na ? ] from local sweat [Na ? ] using absorbent patches [54] and Parafilm-M Ò pouches [55]. Laboratory studies have shown that using a composite of sweat [Na ? ] from multiple regions does not improve the predictability of WBW sweat [Na ? ] [54,55]; however, it may still be prudent to collect sweat from multiple sites when sweat testing in the field (as a measure of quality control or to have a back-up in the event that one patch falls off).

Methodological Sources of Variability
in Sweating Rate and Sweat [Na 1 ]

Method of Sweat Stimulation
There are generally three methods by which sweating can be induced to enable sample collection: (i) pharmacological; (ii) passive thermal (heat) stress; and (iii) exercise. In studies investigating sweat gland function and responsiveness, local sweating is typically stimulated pharmacologically. This method involves the use of a small electrical current (iontophoresis) to propel charged cholinergic agonists (usually pilocarpine) transdermally to stimulate the muscarinic receptors on the sweat glands and induce sweat secretion. Pilocarpine iontophoresis was standardized by Gibson and Cooke in 1959 (Quantitative Pilocarpine Iontophoretic Test [or QPIT] method [89]). Subsequently, in 1983, a simplified version of sweat collection via pilocarpine iontophoresis was introduced (Wescor Macroduct system [100]). Pilocarpine iontophoresis is the gold-standard sweat-testing method for diagnosing cystic fibrosis (CF) [101]; however, it is important to note that the sweating response differs significantly between pharmacological and thermal and/or exercise-induced sweating. Most notably, LSR is consistently higher with exercise and thermal stress compared with pilocarpine iontophoresis [19,102,103]. The reason for the discrepancy could be related to the different mechanisms involved in the sudomotor response between methods. With pilocarpineiontophoresis, sweat secretion is only induced via local cholinergic stimulation of sweat glands, whereas with exercise and/or heat stress other local (e.g. T sk , skin blood flow, adrenergic stimulation, and other neuromodulators) and central mediators (e.g. T c , central command, and exercise pressor reflex) are involved in sweat stimulation [44,45].
Regarding the impact of sweat stimulation methodology on sweat composition, two studies have shown higher [Na ? ] and [Cl -] with exercise versus passive heating [98,104]. There have been mixed results when comparing the composition of pharmacological and thermal sweat. Separate studies have reported higher [105,106], lower [106], or similar [107] electrolyte concentrations in pharmacological versus thermal sweat, therefore more work is needed in this area. Nonetheless, it is clear that sweat tests with athletes should be conducted during exercise and in conditions (i.e. thermal environment) specific/relevant to their sport (see Table 1 for more information).

Skin Surface Contamination and Initial Sweat
Several studies have shown that when serial samples are collected during exercise, initial sweat often has higher mineral concentrations than subsequent samples [99,[108][109][110][111]. This is likely because initial sweat mixes with minerals trapped in the sweat pore or residing in the epidermis [112][113][114][115]. The decrease in mineral concentrations observed throughout exercise is likely due to flushing of surface contamination. As Ely et al. [99] have shown, meticulously cleaning the skin (with distilled water, soap, and a surgical scrub brush) prior to collection results in stable sweat mineral concentrations during 3 h of exercise (using the Parafilm-M Ò pouch technique on the back).
Nonetheless, it is important to note that skin surface contamination from skin desquamation and mineral residues has only been reported with trace minerals (e.g. iron  [99, 108-111, 114, 115]. Studies have shown no change [108] or an increase [99] in sweat [Na ? ] and [K ? ] throughout exercise, which may be explained in part by the significantly higher concentrations of Na ? and K ? in sweat (i.e. less impact on the signal-to-noise ratio) compared with trace minerals; however, more work is needed to confirm this hypothesis. Taken together, it seems prudent to clean the skin surface prior to application of the sweat collection system. This commonly includes wiping the skin surface with alcohol and/or rinsing with distilled water. Meticulous cleaning, such as shaving and/or scrubbing with a surgical brush [99], may only be necessary when conducting laboratory research, using the arm-bag technique (to remove dirt and other material under finger nails), or when measuring trace mineral concentrations in sweat (see Table 1 for more information).

Timing and Duration of Sweat Collection
As discussed in Sects. 2 and 3.1, LSR gradually increases from the onset of exercise until a steady state is reached. Applying absorbent patches approximately 20-30 min into the training session will provide LSR and sweat [Na ? ] results more indicative of steady-state sweating than initial sweat [61,62]. Another important factor to consider is the timing of absorbent patch removal from the skin. A wide range in the duration of sweat collection has been reported, with some laboratory studies suggesting a maximum of 5 min [61], while approximately 15-30 min [116][117][118][119], or even up to approximately 90 min [10,11,13] has commonly been reported in field studies. Few studies have investigated the impact of patch adherence time on LSR or sweat [Na ? ]. Brebner and Kerslake [71] showed that a decline in sweating produced by wetting the skin with water or sweat occurs within 15 min and proceeds exponentially for at least 5 h. However, Dziedzic et al. [18] found no significant difference in forearm sweat [Na ? ] whether absorbent patches were on the skin for 30 or 70 min (although there was a non-significant 8 mmol/L or 13% increase over time). Future research is needed to determine the effects of patch adherence time (and possible interactions with LSR and patch absorbent capacity/size) on local sweat [Na ? ] and to cross-validate results against values obtained with the WBW technique (see Table 1 for more information).

Sample Storage
Another potential source of variability in sweat electrolyte concentrations is the method and duration of sample storage. Dziedzic et al. [18] found that, compared with immediate analysis, there was an approximately 7% decrease in sweat [Na ? ] after freezing (-80°C) and an approximately 14% increase with refrigeration (7°C) for 7 days. There were no significant differences in sweat [Na ? ] among samples analyzed immediately versus after being stored at room temperature (21°C) or in an incubator (32°C) for 7 days [18]. However, another paper reported that, after 5 days of sample storage, increases in sweat [Cl -] occurred to a greater extent at room temperature (21-23°C, approximately 21-66% increase) than at refrigerated temperature (2-8°C, approximately 3-19% increase) [120].
The primary concern associated with sample storage is evaporation; i.e. the loss of water in excess of electrolytes, leading to increases in sweat electrolyte concentrations. A few studies have measured the mass change of sweat samples to assess the impact of different storage methods on sample evaporation. In one study, 72 h of storage at room temperature (25°C) or in refrigeration (4°C) led to minimal sample weight change, while incubation (37°C) led to significant sample evaporation [121]. Another study investigated sample weight changes and the impact of storage temperature (room vs. refrigeration) and the sample sealing method (Parafilm-M Ò seal vs. plastic bag vs. none) over 5 days [120]. The authors concluded that the least evaporation occurred with refrigerated, Parafilm-M Ò sealed  [120]. Accordingly, sweat-testing guidelines established for the diagnosis of CF recommend samples are stored at 4°C for a maximum of 3 days in airtight containers [122,123]. However, the studies on which these guidelines are based did not measure sweat [Na ? ] or investigate longer durations of sample storage (i.e. [3-5 days) or the impact of temperature fluctuations (e.g. shipping across different climates). Therefore, more research is needed to determine best practices for sample transportation and sample stability related to sweat [Na ? ] (see Table 1 for more information).  [122,127], as well as the reference method used in many sweat tests during exercise or thermal stress [13,55,92,105,118,124,128]; however, flame photometers have become outdated and can be difficult to service or replace [129]. Contemporary laboratory reference techniques for sweat electrolyte analysis include IC and ICP-MS, both of which require only small sample volumes and have been found to be highly sensitive, accurate, and reliable (CV approximately 1-5%) [129][130][131]; however, IC, and particularly ICP-MS, involves expensive equipment, labor-intensive sample preparation/analysis, and expertise [123,129].

Sample Analysis
Techniques that are more amenable to sweat analysis in the field include ISE and conductivity. Sweat-testing guidelines for the diagnosis of CF caution against the use of conductivity because it measures the concentration of all ions (and is therefore not specific for the particular ion of interest, such as [Na ? ] or [Cl -]) [127,132]. ISE techniques, while a measure of ion activity rather than a direct measure of concentration, are considered acceptable for CF diagnosis [122,127,133]. Studies measuring sweat [Na ? ] have found that ISE (via a compact Na ? analyzer; HORIBA Scientific, Irivine, CA, USA) has similar reliability to that of IC (CVs approximately 1-4% for both methods) [125,126]. These studies also reported that compared with IC, the ISE technique produced sweat [Na ? ] values with a mean bias of approximately 2-4 mmol/L (or approximately 4-10%) [125,126]. This small but significant discrepancy between ISE and IC could be due to the limited measurement resolution and range of ISE [125,126,129]. Another potential drawback of ISE is the need for larger sample volumes compared with IC and ICP-MS [126,129]. However, investigators should weigh the shortcomings of sweat analysis in the field versus the potential error (e.g. sample evaporation/contamination) and practical inconveniences (e.g. cost, delay in obtaining results) introduced by storing and transporting samples to the laboratory for subsequent analysis. There may be situations where sweat analysis in the field is the best practice (e.g. if sample storage duration and conditions during transportation cannot be well-controlled). Nevertheless, more studies directly comparing different analytical techniques are needed (see Table 1 for more information).

Intra/Interindividual Sources of Variability
in Sweating Rate and Sweat [Na 1 ]

Day-to-Day
Even when sweat-testing methods are well-controlled, a certain amount of variability is observed within subjects. From day-to-day, WBSR has been shown to vary by approximately 5-7% [54,134], while LSR can vary by approximately 6-17% using Parafilm-M Ò pouches [134] and up to approximately 22% with ventilated capsules [60]. The reported day-to-day variability of WBW [Na ? ] is approximately 11-17% [54,87] and for local sweat [Na ? ] is approximately 5-16% for absorbent patches [54] and approximately 8-12% for Parafilm-M Ò pouches [134]. This variability in the sweating response should be taken into account when interpreting study results. For example, differences in results between tests may only have practical significance (e.g. warrant changes in fluid replacement strategy) when a change in conditions (e.g. exercise intensity, environment, equipment/clothing, etc.) elicits changes in WBSR by more than approximately 5% and sweat [Na ? ] by more than approximately 15% (see Table 2 for more information).

Regional
The substantial variability in LSR and local sweat [Na ? ] across the body has been extensively researched [54, 55, 64, 68,77] and reviewed elsewhere [19]. In brief, LSR can vary by up to approximately 360% [55,64,68,77] and tends to be higher on the torso versus extremities [68,135], and, within the torso, higher posteriorly versus anteriorly, and medially versus laterally [68]. In addition, local sweat [Na ? ] across the body can vary by as much as approximately 80-120% within subjects [54,55]. Studies report no significant bilateral differences in scapula LSR [61], forearm LSR [60,61,92], or forearm sweat [Na ? ] [18,96]. There may be small differences in LSR between the dorsal and ventral forearm (7-12%) [68] but this comparison has not been well-studied. In addition, more work is needed to determine the impact of subtle differences in patch placement in a given location (e.g. proximal vs. distal forearm) on LSR and sweat [Na ? ] (see Table 2 for more information).

Intra-and/or Interindividual Variability
Several papers have reported on the wide variability in sweat-testing results among athletes [2,96,136]. For example, Figs. 1 and 2 show WBSR and sweat [Na ? ], respectively, from approximately 500 athletes tested by the Gatorade Sports Science Institute [96]. As this and other studies have shown, WBSR typically ranges from approximately 0.5 to approximately 2.0 L/h [2,24,96]. WBSR can be [3.0 L/h [9,15,96,[137][138][139][140][141] but this is relatively rare (approximately 2% of athletes in Fig. 1a) and is usually associated with extreme circumstances (related to environment, exercise intensity, and/or large body mass) [96]. Local sweat [Na ? ] typically ranges from approximately 10 to approximately 90 mmol/L ( Fig. 2a; see also Maughan and Shirreffs [46], Baker et al. [54], Patterson et al. [55], Shirreffs and Maughan [87], and Verde et al. [92]), while whole-body sweat [Na ? ] is approximately 10-70 mmol/L (predicted, Fig. 2b; measured in Baker et al. [54], Patterson et al. [55], and Shirreffs and Maughan [87]). Many factors are involved in the variability in sweattesting results. These factors that can lead to acute changes in the sweating response include exercise intensity, exercise duration, environmental conditions, clothing/ equipment, hydration status, and circadian rhythm, while longer-term adaptations in the sweating response can occur with heat acclimation, aerobic training, and modifications in dietary sodium. Host factors, such as body mass, body composition, sex, menstrual cycle phase, maturation, aging, medications, medical conditions, and genetics can also play a role in SR and/or sweat [Na ? ] variability. The following section provides a brief summary of the factors accounting for intra-and interindividual differences in SR and sweat [Na ? ].

Rate of Evaporation Required for Heat Balance
According to heat-balance theory, the rate of sweat evaporation from the skin is directly determined by the evaporative requirement for heat balance (E req ), which is represented by the following equation: where M is metabolic energy expenditure, W is external work, R is radiant heat exchange, C is convective heat exchange, and K is conductive heat exchange [142,143]. The primary means by which the body gains heat is from metabolism (which is directly proportional to exercise intensity) and the environment, therefore these factors are also the primary determinants of sudomotor activity [142,144]. Factors such as body size [145][146][147][148], body composition [145], sex [2,68,[149][150][151], or wearing protective clothing/equipment [152][153][154], which (directly or indirectly) impact metabolic heat gain and/or heat loss capacity, can modify SR. Ambient temperature [142,143], solar radiation [155][156][157], and wind [158,159] impact the sudomotor response through their effects on body heat exchange with the environment (see Table 2 for more information). It is also important to note that sweating is not 100% efficient because some secreted sweat can drip from the body and not be evaporated. Therefore, in certain situations (e.g. humid environments), higher SR than calculated from E req may be needed to satisfy the demands for cooling [160,161]. BML body mass loss, CV coefficient of variation, E req required rate of evaporation for heat balance, NA not applicable, [Na ? ] sodium concentration, SR sweating rate, Tc body core temperature, VO 2 oxygen uptake, VO 2max maximal oxygen uptake, WB whole body, WBSR wholebody sweating rate, : indicates increase in the sweating response, ; indicates decrease in the sweating response, $ indicates no effect on the sweating response, ? indicates limited data available

Structural Differences in Sweat Glands
Some intra-and interindividual variability in SR can be explained through differences in the structure of sweat glands. Sato and colleagues have shown that glandular size (volume) can vary by as much as fivefold between individuals [25,162], and there is a significant positive correlation between the size of isolated sweat glands and their methacholine sensitivity and maximal secretory rate [162]. Sweat gland hypertrophy and increased cholinergic sensitivity have been reported to occur with aerobic training [162] and heat acclimation [38] (see Table 2 for more information).

Central and Peripheral Control of the Sweating Response
As discussed in Sect. 2, sweating occurs primarily in response to increases in T c [40]. Two important aspects of thermoregulatory sweating are the onset (i.e. T c threshold) and sensitivity (i.e. slope of the relation between SR and the change in T c ) of the sweating response to hyperthermia   [163]. Shifts in the sweating temperature threshold are thought to be central (hypothalamic) in origin, whereas changes in sensitivity are peripheral (at the level of sweat glands) [40]. Several intra-and interindividual factors can modify the control of sweating [163]. For example, the enhancement of sweating with heat acclimation [37,95,164,165] and aerobic training [166][167][168][169] has been associated with both an earlier onset and greater responsiveness of sweating in relation to T c [57,164,[170][171][172][173][174]. By contrast, dehydration has been shown to delay/alter the sweating response [175,176], such that hyperosmolality increases the T c threshold for sweating onset [177] and hypovolemia reduces sweating sensitivity [178]. The decline in SR with aging is thought to be attributed in part to factors (decline in aerobic fitness) related to decreased sensitivity of sweat glands to cholinergic stimulation [167,[179][180][181]. In addition, other factors, such as maturation [181][182][183][184], altitude [185][186][187], circadian rhythm [188,189], and menstrual cycle [189][190][191][192] have been shown to modify the onset and/or sensitivity of the sweating response. However, it is important to note that modifications in the onset and/or sensitivity of local sweating in relation to T c are not always associated with significant differences in overall whole-body sweat losses during exercise (e.g. across menstrual cycle phases [190,[193][194][195]) (see Table 2 for more information).

Rate of Ductal Na ? Reabsorption
Predominate factors involved in Na ? reabsorption rate in the sweat duct include ion transporter activity and sweat flow rate. Na ? /K ? -ATPase activity is dictated in part by aldosterone, which can be influenced by heat acclimation and dietary Na ? . It is well known that adaptation to the heat leads to improved salt conservation through a decrease in sweat [Na ? ] [37,95,98,196,197], with the underlying mechanism related to increased sensitivity of the sweat gland to circulating aldosterone [37]. In addition, changes in dietary Na ? can modify aldosterone secretion, and thereby modify sweat [Na ? ] [198,199]. Studies have shown that, compared with moderate Na ? intake (3-4 g/day), 8-14 days of high (8-9 g/day) or low (1-2 g/day) Na ? intake are associated with significant increases [128,199] or decreases [128,199,200], respectively, in sweat [Na ? ]. As explained in Sect. 2, sweat flow rate and sweat [Na ? ] are directly related [53,197], therefore acute changes in sweat flow rate, such as an increase due to more vigorous exercise [53,197,201] or warmer ambient temperatures [18], can also lead to increases in sweat [Na ? ].
The potential effect of other factors on sweat [Na ? ] has also been studied. Maturation may impact sweat [Na ? ] as some studies have found higher values in adults than youth, particularly among male subjects [96,183]; however, within adults, neither sex [96,202] nor aging [167] seem to significantly impact sweat [Na ? ]. One study [203] has reported an increase in sweat [Na ? ] with dehydration, but more research is needed to corroborate this finding and elucidate potential mechanisms. The effect of menstrual cycle phase on sweat [Na ? ] is also unclear [195,204,205] (see Table 2 for more information).

Other Factors
Another factor that can impact the sweating response is exercise duration. Prolonged heavy sweating, leading to elevated skin wettedness and hidromeiosis, can decrease SR and sweat [Na ? ] [70,75,135,206]; however, studies have reported no decline in SR or sweat [Na ? ] throughout prolonged low-intensity exercise [99,108]. Interestingly, with heat acclimation the sweat glands become resistant to hidromeiosis (because of gland hypertrophy and sweat dilution) such that higher SR can be maintained [57,161,207,208].
Many investigators have reported significant differences in heat tolerance with racial/ethnic variation; however, it is thought that long-term adaptation to indigenous environmental factors is more important than race or ethnicity per se in the physiological responses to heat stress [57]. For example, heat habituation, characterized in part by lower, more efficient sweating (less dripping) may occur in people indigenous to hot or tropical climates [209][210][211]. For more thorough reviews on race/ethnicity and thermoregulation the reader is referred elsewhere [57,98,212,213].
Finally, some medical conditions and medications can impact the sweating response. Hypohydrosis can occur as a result of pore occlusion from miliaria rubra [214,215] or sunburn [216], as well as from medications that interfere with neural sudomotor mechanisms (e.g. anticholinergics and antidepressants such as amitriptyline) [217]. Hyperhidrosis can occur with menopause, a genetic predisposition [218], or when taking anticholinesterases and antidepressants (e.g. bupropion and venlafaxine) [217]; however, hypo-and hyperhidrosis are often localized and/ or episodic and the impact on thermoregulation during exercise is not well-studied. Increased sweat [Na ? ] is associated with CF, Addison's disease, and renal dysfunction [219]. Individuals with CF have three to five times higher sweat [Na ? ] and [Cl -] than normal because of a genetic absence of a functioning CFTR (two defective genes, homozygote) [220][221][222][223]. Additionally, individuals with one defective gene for CFTR (heterozygote), which is relatively common among White individuals (1/20 people), have elevated sweat [Na ? ] and [Cl -] [222,224]. For more details the reader is referred to the following reviews on CF [123,222,225,226] and other sweat gland disorders [217-219, 227, 228].   Tables 3 and 4 show a list of suggested best practices to consider when measuring WBSR and sweat [Na ? ], respectively, in athletes. These recommendations include options that allow for some methods to be adapted for the specific context of interest (e.g. laboratory vs. field-based testing). It is acknowledged that additional work is needed in some areas to corroborate or refine these best practices. Although these recommendations are intended to be used simply as a guide or educational tool, significant deviations from these methods may warrant the need for sweat testing to be repeated or, at the very least, for the results to be interpreted with caution. Furthermore, even when best practices are followed, some natural within-subject variability in WBSR and sweat [Na ? ] (e.g. approximately 5 and approximately 15% day-to-day, respectively) is still expected. For all of these reasons, categorizing athletes' sweat water and Na ? losses using a low, moderate, and high categorical scheme (or something similar) and recommending a range of fluid replacement options (rather than attempting to pinpoint exact values) may be the most appropriate strategy.

Conclusions
It is clear that methodological practices can be a significant source of unwanted variability in sweat-testing results. Thus, efforts should be made to understand these factors and use appropriate control and standardization when sweat-testing athletes in order to minimize errors associated with methodology. It is also important that sweat tests are interpreted in the appropriate context, i.e. (i) the results are only applicable to the specific conditions (i.e. environment, exercise intensity, etc.) in which the testing was conducted; and (ii) comparisons among sweat tests are only valid if the same methods were used. In summary, sweat testing can be a useful tool to estimate athletes' WBSR and sweat Na ? loss to help guide fluid/electrolyte replacement strategies, provided that data are collected, analyzed, and interpreted appropriately.