Introduction

Evaporation of fluid off the skin’s surface represents the greatest avenue of heat loss during exercise in hot environments [1]; eccrine sweat production and evaporation is therefore the most important physiological mechanism for the maintenance of heat balance in such conditions. Indeed, the requirement for evaporative cooling to dissipate excess heat from the body in order to maintain heat balance (Ė_req), by definition, drives the steady-state sweating response [2–4]. This is supported by a recent study which found that steady-state sweating was similar across a range of absolute core and skin temperature conditions that elicited the same Ė_req [5]. Similarly, metabolic heat production (Ḣ_prod) is also positively related to the rate of whole-body sweating [6], which is intuitive as Ė_req is determined as the net difference between Ḣ_prod and the sum of respiratory and dry heat exchange [7]. Therefore, both Ḣ_prod and Ė_req are two modifiable variables responsible for driving the rate of evaporative cooling, which can be estimated using partitional calorimetry in a controlled laboratory setting [7]. Valid and reliable measurements of thermal sweating during exercise, and their determining factors (Ḣ_prod and Ė_req), are fundamental to the investigation and interpretation of human thermoregulation, yet the reliability of these measures has not been sufficiently reported.

There are numerous ways to assess sweating responses in humans [8,9], with many methods designed to determine whole-body or local sweat rate (LSR; [10]) and sweat gland activation (SGA; [11]). Local sweat rate is typically measured using the ventilated capsule method; additionally, absorbent patches affixed to the skin can be used by assessing pre- to post-exercise differences in patch mass, across a known time period [8,10,12]. This is a long-standing technique, which has been adapted from early work in the 1930/40s [13,14], leading to the more recent technical absorbent method [12,15,16] and absorbent patch technique [8,10]. The technical absorbent method was strongly associated with the ventilated capsule method (r = 0.74 to 0.95), across varying time durations and patch sizes (sample surface area) and locations (body region; [9]). Estimated whole-body sweat loss (WBSL) is most commonly reported by comparing changes in body mass pre- versus post-exercise [17,18]. Sweat gland activation can be assessed using the modified iodine-paper technique, which is a manual way of measuring sweat droplets on the skin surface and can subsequently be quantified using a computer programme. This technique produces repeatable, intra-trial values (coefficient of variation [CV%] = 11 ± 10%) during controlled exercising conditions [11]. However, this method does not account for clustering of sweat droplets from multiple glands. This could affect the consistency of results and might be resolved by expressing SGA according to the surface area covered, which has not been previously reported in the literature. Lastly, an important element of this technique is the application of iodine impregnated paper onto the skin surface. Varying sizes of iodine paper have been used in different studies and are reported inconsistently [19–23]. It is possible that the size of the paper significantly affects the repeatability of this manual technique, which assumes an even distribution of sweat gland activation across the measured areas. Thus, the effect of the paper size on measurement error requires further investigation.

While intra-trial (i.e. within a single trial day, recorded seconds-to-minutes apart) reliability studies have been conducted for some methods, such as the modified iodine-paper technique [11] and the absorbent patch technique for measuring LSR [10], none have analysed their inter-day reliability at rest and a range of relative exercise intensities. Considering that the majority of studies have multiple testing sessions, spanning several days (or weeks) and exercise intensity varies across studies, it is important to appreciate the variability under such differing circumstances and the subsequent reliability of the technique. In addition, there are daily fluctuations in individuals’ physiological responses to stimuli, which could potentially create noise when assessments are conducted across a longer time period [24]. Without quantifying the reliability (i.e. noise) of a technique across several days, its efficacy (e.g. detection limits) is unknown, and thus the capacity to identify meaningful changes in sweat (i.e. rate and SGA) as a result of an intervention, such as acclimation or dietary supplementation, cannot be determined [25]. Furthermore, the reliability of these techniques should be considered in conjunction with the established drivers of thermal sweating (Ḣ_prod and Ė_req) since error in sweating measures will be partly determined by variation in these factors.

On the understanding that Ḣ_prod and Ė_req predominantly determine the magnitude of required evaporative cooling during exercise in hot environments [4,5], it stands to reason that any measurement used to determine the sweating response should be positively associated, and therefore constructively valid. This includes measures such as SGA, LSR and WBSL. Indeed, Ḣ_prod and Ė_req have been positively associated to whole-body sweat rate and LSR, determined using absorbent patches [26]. The relationship between Ė_req and whole-body sweat rate, measured via the ventilated capsule method has also been established [4–6]. However, the relationship between Ḣ_prod and Ė_req and all other techniques for assessing the sweating response have not.

Based on the above reasoning, the aims of the current study were to establish the inter-day reliability of: i) the modified iodine-paper technique for the measurement of SGA using two separate assessment methods (sweat gland counting and surface area covered); ii) the absorbent patch technique for the measurement of LSR; and iii) pre- vs post-exercise body mass changes for measurement of WBSL. Finally, the relationship between all measurements of thermal sweating and both Ḣ_prod and Ė_req was assessed to establish the construct validity of these measurements.

Methods

Preliminary testing

During visit 1, participants undertook an incremental exercise test to volitional exhaustion on a cycle ergometer (Monark Exercise AB, Ergomedic 874E, Varberg, Sweden) in hot ambient conditions (37.6 ± 0.4°C and 27.0 ± 5.9% RH) to determine their individual work rates required to elicit Ḣ_prod of ~200, ~250 and ~300 W/m² and their peak oxygen consumption (). Participants performed a 5-min warm-up at 80 W, followed by 5-min rest, before commencement of the exercise test. Oxygen consumption () was measured using breath-by-breath gas analysis (Jaeger Vyntus CPX, Hoechberg, Germany), with determined as the highest 30-s mean value, which occurred in the final stage of each participant’s test. Criteria for achieving was: 1) reaching volitional exhaustion; 2) respiratory exchange ratio (RER) > 1.15; 3) final HR within 10 beats/min of age-predicted maximum; and 4) rating of perceiving exertion (RPE) > 19 (6–20 Borg scale; [29]). The test was designed to progressively increase mechanical work rate on the ergometer, in a square-wave manner, to elicit a range of Ḣ_prod values, including those required for exercise in visits 2 and 3. The Ḣ_prod was determined by subtracting the rate of mechanical work (Wk) from the rate of metabolic energy expenditure (Ṁ; Eq 1).

(Eq 1)

Where metabolic energy expenditure (Ṁ) was determined using measured and within the final 1-min of each stage (Eq 2): (Eq 2)

To achieve the necessary Ḣ_prod values, the test was initiated at a mechanical work rate below that which was required to elicit the lowest desired Ḣ_prod (200 W/m²) and increased by 20 W every 5-min at a fixed cadence of 80 rpm until exhaustion. The Ḣ_prod (W/m²) at each stage was estimated based on participant body surface area (BSA; Eqs 3 and 4; [2]).

(Eq 3)(Eq 4)

Dubois & Dubois equation [30].

The mechanical work rate required to elicit each target Ḣ_prod (W/m²) for the exercise trials during visits 2 and 3 (i.e. ~200, ~250 and ~300 W/m²) was subsequently estimated based on the linear relationship (y = a + b), between Ḣ_prod (W/m²) and work rate during the incremental test (Eq 5).

(Eq 5)

Inter-day test-retest reliability

Pre-exercise instrumentation.

Participants were required to arrive euhydrated, as determined by a urine osmolality value < 600 mOsm kg/H₂O (Portable osmometer, Osmocheck, Vitech, Scientific Ltd). If the participant was not deemed to be euhydrated, they were asked to drink 500-ml of plain water and wait 30-min before their urine osmolality was re-measured. Participants wore standardised cycling shorts (94% polyester; 6% elastane), as well as a sports bra for female participants. To measure core body temperature (T_core), participants were instructed to insert a flexible rectal probe 10 cm past the anal sphincter. Skin thermistors (Grant Instruments Ltd., Cambridge, UK) were attached to four sites on the participant’s right side: upper-chest, mid-humerus, mid-calf and mid-thigh to measure mean skin temperature. Ramanathan’s equation [31] was used to calculate mean skin temperature: (Eq 6)

Prior to application of the skin thermistors, the skin was dry-shaved, cleaned with soap and water and allowed to air-dry. Both core and skin temperature were continuously recorded using a data logger (SQ2010; Grant Instruments Ltd., Cambridge, UK). Heart rate (HR) was continuously recorded throughout each exercise trial (Polar Heart Rate Monitor M400, Warwick, UK). Each participant’s body mass was measured (whilst wearing cycling shorts, a sports bra for women, the HR monitor, the inserted rectal thermistor and the skin thermistors) using a calibrated scale (resolution 50 g; Seca 711, Hamburg, Germany). This was necessary because of the repeated body mass measurements throughout the trial.

Exercise trials.

Participants initially rested for 30-mins in a seated position in a temperature-controlled room that was regulated to an air temperature of 37.6 ± 0.4°C and a %RH of 27.0 ± 5.9%. Environmental conditions, such as ambient dry bulb temperature (°C), RH and air velocity (m/s), were continuously monitored (Kestrel 5400 Heat Stress Tracker, Kestrel Meters, Boothwyn, PA, US). An electric fan (SIP 24” Drum Fan, Loughborough, UK) was placed adjacent to the participant during the rest period and diagonally in-front during the exercise periods, providing an airflow of 1 m/s directed at the torso. After 30-min of rest, the participants were seated on the ergometer and performed three exercising periods of 30-min, at the three pre-determined fixed rates of Ḣ_prod (~200, ~250 & ~300 W/m²). During exercise, oxygen consumption was measured using the same breath-by-breath expired gas analyser (Jaeger Vyntus CPX, Hoechberg, Germany). Participants’ body mass was measured, and they were provided with 200-ml of plain water (maintained at room temperature [~20°C]) to drink between each 30-min period. Rating of perceived exertion was recorded using a 6–20 point Borg scale [29], while thermal comfort (TC) was recorded using a 7-point scale (where -3 = “much too cool”, 0 = “comfortable” and 3 = “much too warm” [32]). Thermal sensation (TS) was recorded using a 9-point scale (where -4 = “very cold”, 0 = “neutral” and 4 = “very hot” [33]). All perceptual data (RPE, TC and TS) were recorded at 5-min intervals during the rest and exercise periods, and upon completion of the trial. Local sweat rate and SGA measurements were taken using the techniques described below.

Partitional calorimetry.

As detailed in supplementary material, heat balance parameters such as Ḣ_prod and Ė_req (Eq 7) were estimated for each 30-min exercise period via partitional calorimetry [7]. Ḣ_prod was also expressed relative to body surface area [30].

(Eq 7)

Local sweat rate measurement.

Local sweat rate was determined using the absorbent patch technique [10], on the left scapula. Measurements were taken during the final 5-min of each 30-min exercising period. Prior to the trial, the area of interrogation on the skin’s surface was shaved and cleaned using water and gauze. A template matching the size of the absorbent patch was pressed to the skin’s surface and outlined in indelible ink to identify the area of measurement and ensure consistency of application across the exercise stages and the two inter-day reliability trials. The patch (Medipore + Pad [3M]) was 5 cm x 5.5 cm, with an absorbent capacity of ~7 g. Immediately before patch application, the skin was wiped dry with gauze. The patch was weighed (resolution 0.01 g; Ohaus, Navigator N24120, Nänikon Switzerland) prior to and after the 5-min skin application. Local sweat rate (mg/cm²/min) was determined as: pre vs post change in patch mass in milligrams, divided by the surface area of the patch (5 cm x 5.5 cm) and the duration of application (5-min; Eq 8).

(Eq 8)

Modified iodine-paper technique.

The modified iodine-paper technique [11] was used to determine SGA on the right scapula. 100% cotton paper (Southworth, Agawam, MA, US) was cut to 9 x 9 cm and further divided into 6 x 6 cm, 3 x 3 cm and 1 x 1 cm sections using a fine-point pencil (Fig 1). The paper was placed in an air-tight sealed box, containing iodine in solid form (Sigma-Aldrich, St. Louis, MO). Each 9 x 9 cm piece of cotton paper was suspended from the lid of the container to avoid direct contact with the iodine. The pieces of paper were impregnated with iodine after ~48-h, after which they were removed and placed in sealed bags. Double-sided tape was used to affix the cotton paper to a hard-flat plastic surface to ensure uniform application to the skin. Prior to testing, a 9 x 9 cm template was pressed to the skin’s surface at the designated site and outlined in indelible ink to ensure consistency of application across the exercise stages and the inter-day trials. At the end of the rest period and each 30-min exercise period, the skin’s surface was blotted dry using a small towel before the iodine-impregnated cotton paper was firmly applied for 5-s. Visually identifiable blue colourations formed on the paper, indicating excreted sweat from active sweat glands during the 5-s application (Fig 1). A high-resolution (3024 x 4032) photograph (image) was taken of the paper and subsequently analysed using ImageJ [34].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1.

(a) Cotton paper divided into pre-determined areas and (b) Iodine impregnated paper displaying sweat gland activation within the pre-determined and optimal (red) areas.

https://doi.org/10.1371/journal.pone.0278652.g001

Iodine paper image processing and analysis.

ImageJ was used to determine i) the number of active sweat glands and ii) the percentage surface area covered with sweat within each pre-determined section of the paper (9 x 9 cm, 6 x 6 cm, 3 x 3 cm, 1 x 1 cm). Further analysis was also performed to determine the optimal area of sweat gland density within 3 x 3 cm and 1 x 1 cm areas within the 9 x 9 cm iodine paper area. This was defined as the area (3 x 3 cm and 1 x 1 cm) with the highest density of recruited glands. All images were taken in a well-lit area to avoid colour contrasts on the paper. For the purposes of replication, please see supplementary files for a detailed guide on ImageJ functions. The number of active sweat glands and the percentage surface area covered in each pre-determined area were later normalised to the maximum value for each individual across all exercise periods. This was to account for inter-individual variation in both the maximal SGA and the point of uncompensability (i.e. individual differences in the balance between evaporative heat transfer [Ė_req] and the maximum evaporative capacity of the environment [Ė_max]; [7]).

Results

Whole-body sweat loss and local sweat rate

Participants’ mean values for WBSL and LSR during both trials and between-trial reliability (CV%) are presented in Table 1. There were no trial main effects or trial x period interaction effects for WBSL (P = .816; P = .272, respectively) and LSR (P = .468; P = .439, respectively).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Mean, standard deviation and reliability of whole-body sweat loss and local sweat rate at rest and a range of exercise intensities.

https://doi.org/10.1371/journal.pone.0278652.t001

Sweat gland activation

Participants’ mean values for SGA (gland number and surface area covered) during both trials and between trial reliability in the form of coefficient of variation (CV%) are presented in Tables 2 & 3. There were no trial main effects or trial x period interaction effects for gland number at 9 x 9 cm (P = .085; P = .488, respectively), 3 x 3 cm (P = .210; P = .407, respectively), optimal 3 x 3 cm (P = .056; P = .526, respectively), 1 x 1 cm (P = .125; P = .217, respectively) and optimal 1 x 1 cm (P = .212; P = .115, respectively) and surface area covered at 9 x 9 cm (P = .785; P = .372, respectively), 6 x 6 cm (P = .227; P = .357, respectively), 3 x 3 cm (P = .709; P = .407, respectively), optimal 3 x 3 cm (P = .133; P = .754, respectively), 1 x 1 cm (P = .808; P = .282, respectively) and optimal 1 x 1 cm (P = .571; P = .100, respectively). A significant trial main effect was found for gland number at 6 x 6 cm (P = .021), but no trial x period interaction effect (P = .571).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Mean, standard deviation and reliability of sweat gland activation (gland number) at rest and a range of exercise intensities.

https://doi.org/10.1371/journal.pone.0278652.t002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Mean, standard deviation and reliability of sweat gland activation (area covered) at rest and a range of exercise intensities.

https://doi.org/10.1371/journal.pone.0278652.t003

There were also no trial main effects or trial x period interaction effects for normalised gland number at 9 x 9 cm (P = .209; P = .424, respectively), 3 x 3 cm (P = .138; P = .459, respectively), optimal 3 x 3 cm (P = .442; P = .625, respectively), 1 x 1 cm (P = .178; P = .152, respectively) and optimal 1 x 1 cm (P = .484; P = .367, respectively) and normalised surface area covered at 9 x 9 cm (P = .551; P = .456, respectively), 6 x 6 cm (P = .123; P = .123, respectively), 3 x 3 cm (P = .351; P = .270, respectively), 1 x 1 cm (P = .370; P = .325, respectively) and optimal 1 x 1 cm (P = .136; P = .154, respectively). There was a significant trial main effect, but no trial x period interaction effect for normalised gland number at 6 x 6 cm (P = .048; P = .362, respectively) and surface area covered at optimal 3 x 3 cm (P = .046; P = .430, respectively).

Partitional calorimetry

Participants’ mean values for absolute Ḣ_prod (W), Ḣ_prod (W/m²), Ė_req (W) and Ė_req (W/m²) during both trials and between trial reliability in the form of coefficient of variation (CV%) are presented in Table 4. There were no trial main effects or trial x period interaction effects for Ḣ_prod (P = .590; P = .112, respectively), Ḣ_prod (W/m²; P = .603; P = .126, respectively), Ė_req (P = .904; P = .187, respectively) and Ė_req (W/m²; P = .946; P = .211, respectively).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Mean, standard deviation and reliability of partitional calorimetry variables (Ḣ_prod and Ė_req) at a range of exercise intensities.

https://doi.org/10.1371/journal.pone.0278652.t004

Correlations

Absolute Ḣ_prod (Fig 2), relative Ḣ_prod (Fig 3) and absolute Ė_req (Fig 4) were significantly correlated with WBSL, LSR and SGA (gland number and surface area covered).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. The relationship between absolute heat production (Ḣ_prod) and measures of the sweating response.

https://doi.org/10.1371/journal.pone.0278652.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. The relationship between relative heat production (Ḣ_prod) and measures of the sweating response.

https://doi.org/10.1371/journal.pone.0278652.g003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. The relationship between absolute evaporative requirement for heat balance (Ė_req) and measures of the sweating response.

https://doi.org/10.1371/journal.pone.0278652.g004

Discussion

The current study assessed the inter-day reliability of components of thermal balance and multiple measures of the sweating response at rest and various fixed exercise intensities. The pre-to-post exercise body mass change for measurement of WBSL, the absorbent patch technique for the measurement of LSR (Table 2) and the modified iodine-paper technique for the measurement of SGA (both gland counting [Table 3] and area covered [Table 4]), were all found to have greater inter-day reliability during the exercising periods compared to rest. Relative and absolute Ḣ_prod and Ė_req were found to have very low CVs (Table 1), indicating a controlled environment between the two trials. As expected, all measures of the sweating response had strong or moderate correlations with Ḣ_prod (Fig 2) and Ė_req (Fig 3; the drivers of sweat production).

It is important to evaluate the systematic bias (unidirectional error) of a test to understand the potential threats to reliability of a measurement technique [35]. For example, in the current study, it was feasible that participants would demonstrate early-phase adaptation to the heat and exercise stimulus across days, or that the user would improve their technique while using manual methods to assess sweating responses. However, there were no systematic biases detected across all but one of the sweating and calorimetry variables, thus indicating that the trials were conducted in a repeatable manner and the techniques used to characterise the participants’ responses were not affected by learning, familiarisation or adaptation effects. Therefore, these techniques, when performed four days apart, can be employed by researchers to control thermal balance and monitor sweating responses among unfamiliarised participants in a laboratory, without concern for systematic adaptations or improvements in the testing procedure. The systematic biases observed for absolute (P = .021) and normalised (P = .048) sweat gland number using iodine paper at 6 x 6 cm were unanticipated but were largely attributable to two of the participants’ measurements. This is likely to be caused by the natural variation in sweat gland recruitment between testing days, which we discuss in later sections of the current discussion in relation to random error.

Determining the pre vs post difference in body mass is the most basic, yet most common method of measuring sweat loss and, thereby, estimating evaporative cooling [37]. Despite this, we are unaware of a study that has investigated the inter-day reliability of WBSL at rest, and multiple fixed exercise intensities. The CVs for WBSL ranged between 6.0 ± 2.0% and 33.5 ± 23.8%. When all exercise work-rates were combined, this was 7.8 ± 1.8%. The CVs decreased (i.e. reliability improved) with increasing exercise intensity, with resting values demonstrating the poorest reliability (33.5 ± 23.8%) and the highest exercise intensity producing the best reliability (300 W/m² Ḣ_prod; 6.0 ± 2.0%). These results suggest that WBSL has greater reliability at higher work-rates and, consequently, higher sweat rates. This is consistent with the understanding that a greater afferent stimulus for sweating caused by increasing exercise intensities and ambient temperatures elicits a higher frequency and amplitude of sweat gland recruitment [38] and, thus, more pronounced effect on sweat production [39]. It is likely that the magnitude of stimulus leads to a more sustained effect of sweat production, and therefore more consistent result between trials. On the other hand, the larger CV observed at rest and lower exercise intensities indicates a greater random error or measurement ‘noise’, which can be due to technical (instrument dependent variation), biological variation or human error [25]. The majority of body mass scales have a resolution of ~100 g and therefore smaller body mass changes, as reported at rest, are more difficult to accurately quantify, as they may be just outside the sensitivity of the measurement device. It is important to consider error in relation to ‘signal’ change; for example, WBSL has been reported to vary by 23% due to heat acclimation [22] and by 13% following supplementation of taurine [40]. These signal changes are greater than the measurement noise (CV%) found during exercise in the current study but are smaller than the CV at rest. Therefore, changes in WBSL should be detectable using this technique during exercising periods but might not be identifiable during rest, owing to the inherent noise of the tests between weeks. It should be mentioned that there are alternative techniques for measuring WBSL with greater accuracy, such as a potter balance, however, these are rarely used due to practicality and limited availability [37]. Scales with a resolution of 1 g and accuracy of up to 10 g are also available, though, their high cost can limit their accessibility.

The CVs for LSR, measured using absorbent patches, were 30.1 ± 12.8% overall and 13.8 ± 3.3% for all exercise work rates combined and ranged between 10.1 ± 5.6% and 77.9 ± 39.1%. While the largest CV was also measured during rest (77.9 ± 39.1%), unlike WBSL, these values did not decrease with exercise intensity. The larger CV% found at the highest exercise intensity (Table 1) could be attributed to both technical and human sources of error. For example, at higher sweat rates, the absorbent patches’ adherence to the skin surface can be compromised, which could affect the amount of sweat absorbed by the patch. In addition, it is well-known that sweating varies across regions of the body, with sweat rates typically higher in the upper-back [41]. Even within the same region of the body, sweating can be highly variable, depending on specific locations [15]. Perhaps less-well recognised is the pulsatile manner of sweat gland innervation and secretion, which varies in amplitude, shape and temporal spacing [42]. Therefore, it is possible that the measurements of localised sweat, such as absorbent patch techniques, could be affected by factors such as the timing of measurement and the consistency of patch application on the upper-back. Additionally, the size of the patches applied, and the resolution of the scales used to weigh them, could further affect the results. Overall, these CV% values suggest that the measure of LSR is slightly less reliable than that of WBSL and could make changes in LSR more difficult to detect, particularly during resting conditions across short time periods. However, the CVs are similar to previous findings, where [10] reported inter-day reliability of local sweat rate at the scapula during 90-min of cycling at 75–80% HR_max was 14.5%. The ‘noise’ reported during exercise in the current study (13.8%) is markedly less than the expected signal change in LSR on the back with a stimulus such as heat acclimation (~26.7–30%; [23,43,44]).

A poorer reliability at rest was also reflected in the SGA measurements, with resting CV% values ranging from 41.2 ± 22.8% to 70.0 ± 30.5% across the SGA variables and measurement areas. In the exercising periods, however, the CV% values ranged from 15.5 ± 5.4% to 30.0 ± 7.4%, which we attribute to the same reasons provided for other reliability values of resting sweat measures. These values are slightly higher than the intra-trial CV% values (11 ± 10%) previously reported [4], suggesting that the reliability of this technique is slightly less between- than within-day, which could be attributed to greater biological variability across the two separate trial days. Sweat gland number and surface area covered at 9 x 9 cm, optimal 3 x 3 cm and optimal 1 x 1 cm were the most reliable (Tables 2 & 3). For the optimal 3 x 3 cm and 1 x 1 cm areas this was anticipated, because as part of this new approach, the maximum sweat gland density across all regions of the iodine paper was identified. As discussed herein, sweat gland recruitment patterns are pulsatile in nature and vary across anatomical regions [15,41,42]. As such, the pre-designated anatomical regions used in the traditional method are less likely to capture the optimal area of gland activity at any one time-point, leading to increased random error. Searching for the area of optimal gland activity is a more flexible approach, accounting for this variation. The resulting CVs found in this study across exercise intensities supports this. Of course, from a practical perspective, this method is more time-consuming than the traditional measures and researchers should decide whether the incremental improvements (~2%) in reliability warrant the time burden. This decision should be guided by the researcher’s analytical goals [35] but a typical change in SGA as a result of heat acclimation is approximately 27.9% [23] and the CV% values reported here for 9 x 9 cm and optimal 3 x 3 cm and 1 x 1 cm (15.5–17.6%) would permit detection of these changes. In regard to the reliability of the fixed paper sizes of 9 x 9 cm, 6 x 6 cm, 3 x 3 cm and 1 x 1 cm, this was generally poorer with decreasing paper size for both sweat gland number and area of the paper covered in the exercising periods. Moreover, there is variation in the sweat gland number identified per cm² in the larger paper sizes compared to those identified in the smaller ones. This further indicates that SGA is affected by the paper size and, therefore, the results from varying paper sizes are not directly comparable. The explanation for these findings is similar to that provided above, as a larger surface area will more likely include the optimal area of gland activity and account for variability across the body region, which may not be included in smaller paper sizes.

We investigated, for the first time, measurement of the surface area covered, rather than activated sweat gland number. We considered that this could be more reliable than gland number on the basis that co-joining of neighbouring glands could be incorrectly measured as single glands using the counting method, leading to measurement error. We theorised this would most likely be the case at higher sweat rates (i.e. Ḣ_prod of 300 W/m²), where dots on the paper would be larger and output per gland would be taken into account, as well as number of activated glands. However, findings were equivocal with some paper sizes more or less reliable for gland number vs surface area covered. Nevertheless, these measures performed very similarly in regard to reliability, with the surface area covered marginally less reliable than sweat gland number by approximately ~1.9% across the paper sizes (Tables 2 & 3). Therefore, we conclude that either surface area covered, or the counting method could be used but recognise that the counting method may be preferable for researchers intending to combine sweat gland recruitment with LSRs to determine output per gland [23].

Indeed, it has been shown that progressive recruitment of sweat glands occurs during early exercise in the heat, but at latter stages, LSR is reliant upon sweat production per activated gland [39]. This might also support the current findings, where the 250 and 300 W/m² Ḣ_prod exercise intensity produced the most reliable sweat gland measurements. Owing to our study design, the highest work-rates occurred during later stages of exercise and would naturally have recruited most or all of the available sweat glands, leading to greater consistency in the results. Nevertheless, the noise associated with this technique should be considered by potential users and the manual nature of the paper application, particularly during exercise, will inevitably lead to some human error and some unreliability, which might prevent detection of ‘small’ changes in sweat gland activity.

To the best of our knowledge, we are the first to directly report on the inter-day reliability of selected components of thermal balance using partitional calorimetry; namely, absolute and relative Ḣ_prod and Ė_req. Both absolute Ḣ_prod (CV% = 1.8 ± 0.5) and Ė_req (CV% = 1.7 ± 0.5) and relative Ḣ_prod (CV% = 1.9 ± 0.5) and Ė_req (CV% = 1.7 ± 0.5) were reliable across all exercise intensities. Given that Ḣ_prod and Ė_req are determined by a combination of measured variables, such as (CV% = 1.7 ± 0.5), Wk and ambient conditions (°C [CV% = 0.2 ± 0.1] and %RH [CV% = 4.3 ± 1.4]), which are commonly used laboratory measures, these results were anticipated. It was necessary to establish the reliability of these variables because of their reported relationship with selected sweat measurements [4,5]. Indeed, as drivers of thermal sweating, consistent control of these thermal balance components is necessary for reliable measurement of sweat responses.

As anticipated, WBSL showed strong significant positive correlations with both absolute Ḣ_prod and Ė_req (r_rm = 0.81; r_rm = 0.81, respectively). There were also moderate to strong, significant positive correlations with Ḣ_prod and Ė_req for LSR (r_rm = 0.78; r_rm = 0.78, respectively). This is consistent with previous reports, where whole-body sweat rate (adjusted R² = 0.64; adjusted R² = 0.78, respectively; [4,6]) and LSR [5], as measured by the ventilated capsule technique have been associated with absolute Ḣ_prod and Ė_req. Additionally, LSR at the arm (r = 0.62; P = 0.03; r = 0.38; P = 0.23) and forehead (r = 0.56; P = 0.06; r = 0.31; P = 0.33), determined using absorbent patches, has been moderately correlated with Ė_req and Ḣ_prod respectively [26]. To expand upon these findings, we evaluated the relationship between other sweat measures, such as SGA, and Ḣ_prod and Ė_req, using an appropriate statistical technique that accounts for repeated observations across the three exercise intensities. Overall, SGA (both gland number and surface area covered), normalised to maximum values, were related to absolute Ḣ_prod and Ė_req, with surface area covered at 9 x 9 cm (r_rm = 0.83; r_rm = 0.83, respectively) and 6 x 6 cm (r_rm = 0.80; r_rm = 0.80, respectively) demonstrating the strongest relationships. All other SGA variables demonstrated moderate correlations (Figs 2 & 3). The larger iodine paper sizes showed stronger relationships with Ḣ_prod and Ė_req, across both methods of establishing SGA. In addition, the relationships between Ḣ_prod and Ė_req were consistently greater for surface area covered compared to number of active sweat glands counted. Collectively, these results suggest that larger paper sizes and using the surface area covered method were more consistently associated to the pre-established drivers of thermal sweating. However, all SGA variables related positively with Ḣ_prod and Ė_req, and, therefore thermal stimuli, and are consequently valid. The reliability of the method should also be considered when deciding upon the method used to establish SGA.

Conclusion

As was anticipated, all sweating response variables were positively related with Ḣ_prod and Ė_req. Absolute and relative Ḣ_prod and Ė_req demonstrated inter-day reliability, adequate to control the thermal sweating response. The pre-to-post exercise body mass change for measurement of WBSL, the absorbent patch technique for the measurement of LSR and the various methods of establishing SGA were all also found to have inter-day reliability during the exercising periods, sufficient to detect changes in thermal sweating that might occur following an intervention, such as heat acclimation or dietary manipulation. This was not the case at rest, however, and therefore these methods would be unlikely to be able to detect any changes to the resting sweating response. The modified iodine-paper technique was marginally more reliable at 9 x 9 cm and when the 3 x 3 cm and 1 x 1 cm area of optimal sweat gland density was determined, particularly when measuring the sweat gland number, as opposed to the surface area covered. The larger paper sizes (9 x 9 cm and 6 x 6 cm) had the strongest relationships with Ḣ_prod and Ė_req, especially when SGA was measured using surface area covered. We therefore recommend that to establish SGA at the upper-back, 9 x 9 cm paper sizes are used, with the option of identifying the 3 x 3 cm or 1 x 1 cm optimal areas if deemed necessary. The method of analysis applied (i.e. gland counting or surface area covered) should be chosen based on the research aim.

Supporting information