Post-Exercise Sweat Loss Estimation Accuracy of Athletes and Physically Active Adults: A Review

The main purposes of this review were to provide a qualitative description of nine investigations in which sweat losses were estimated by participants following exercise and to perform a quantitative analysis of the collective data. Unique estimations (n=297) were made by 127 men and 116 women after a variety of exercise modalities in moderate to hot environmental conditions. Actual sweat loss exceeded estimated sweat loss (p<0.001) for women (1.072 0.473 vs. 0.481 0.372 L), men (1.778 0.907 vs. 0.908 0.666 L) and when all data were combined (1.428 0.806 vs. 0.697 0.581 L), respectively. However, estimation accuracy did not di er between women (55.2 51.5%) and men (62.4 54.5%). Underestimation of 50% or more of sweat losses were exhibited in 168 (54%) of estimation scenarios with heavier sweaters displaying a higher prevalence and trend of greater underestimations in general. Most modern guidelines for uid intake during and between training bouts are based on approximate sweat loss estimation knowledge. These guidelines will likely have minimal e cacy if greater awareness of how to determine sweat losses and accurate recognition of sweat losses is not increased by coaches and athletes. Keywords:hydration strategies; team sports; endurance; uid balance; sweat 1. Introduction Over the last two and a half decades, multiple scienti c cohorts and conclaves have devoted extensive e orts to provide physically active individuals with robust guidelines for

if greater awareness of how to determine sweat losses and accurate recognition of sweat losses is not increased by coaches and athletes. Keywords:hydration strategies; team sports; endurance; uid balance; sweat 1. Introduction Over the last two and a half decades, multiple scienti c cohorts and conclaves have devoted extensive e orts to provide physically active individuals with robust guidelines for uid intake strategies during and in the times between training bouts [1–7]. While some viewpoints have contended that thirst is an adequate indicator and stimulus for proper uid consumption [6,8], most other guidelines include detailed recommendations for ideal uid intake practices based on approximate estimates of sweat losses that will be incurred [1,3,4,7]. There is no lack of original research concerning hydration and physical activity, but one article in particular, Passe et al. [9], piqued the interest of the current authors in regards to whether the more nuanced guidelines have pragmatic value in real world practice. Passe et al. [9] reported that during a 16-km run trained runners severely under replaced sweat losses with uid intake (replacing 30.5% 18.1% of sweat loss) in comparison to formal contemporary hydration recommendations of that period suggesting uid intake match sweat losses [1]. Sports2020,8, 113; doi:10.3390 /sports8080113 /journal/sports

Sports2020,8, 113 2 of 12 This occurred despite uid intake opportunities that were designed to be as accommodating as possible. The recommendation of complete uid replacement during exercise [1] that the investigators used to interpret the adequateness of uid consumption has since been deemed excessive and that body mass loss during exercise should occur [4]. However, to our knowledge Passe et al. [9] also presented the rst evidence that reported athletes' perceptions of sweat losses, with an average underestimation of sweat loss volume of ~43%. Beverage intake adequateness changes drastically if replacement volume is viewed from the standpoint of the runners' perceived sweat loss volume. This nding is revelatory but almost universally disregarded when discussing the e cacy of nearly all contemporary, formal hydration guidelines that base uid replacement recommendations on individual sweat loss volume. Since the seminal article that reported sweat loss estimation accuracy, multiple attempts to document athletes' perception of sweat loss volume have been undertaken in the current authors' respective laboratories. However, there has been no formal review of these sweat loss estimation investigations. The two primary purposes of this review will be to provide a single, concise literature overview source of sweat loss estimation investigations and combine individual data from these studies for a comprehensive descriptive analysis of athletes' sweat loss estimation accuracy in eld and laboratory settings. A tertiary aim was to o er some brief, real world application considerations based on these ndings for coaches, athletes and sport medicine sta . 2. Methods The lead author initiated communications with all authors of investigations that had cited his original sweat loss estimation study [10] and collected original sweat loss estimation data in their respective publications. Although not a formal systematic review, each author then completed their own individual search of literature that cited their original sweat loss estimation studies. No additional published articles were discovered. Data from nine published investigations [10–18] in which sweat loss was estimated were shared by current authors through personal communication. A new spreadsheet database was created for comprehensive analysis. Data included a description of participants (age, sex and training background)

their own individual search of literature that cited their original sweat loss estimation studies. No additional published articles were discovered. Data from nine published investigations [10–18] in which sweat loss was estimated were shared by current authors through personal communication. A new spreadsheet database was created for comprehensive analysis. Data included a description of participants (age, sex and training background) and physical activity conditions (exercise modality, environmental conditions and duration of exercise). Additionally, pre-activity body mass, post activity body mass, uid intake and void output (if applicable) were used to calculate actual sweat loss, post-exercise estimated sweat loss volume and sweat loss as a percent of body mass. Sweat loss was calculated as change from pre- to post-exercise body mass with adjustment for uid intake or urine production and under the assumption that 1 kg of body mass change was equal to 1 L of sweat loss. The authors recognize that sweat loss is more complex than this simple model due to the presence and density of electrolytes and metabolites found in sweat [19] and respiratory water and gas exchanges related to metabolic processes [20]. However, due to the variety of exercise modalities and duration of exercise for most of the investigations, the authors have chosen to represent sweat loss volume in liters as equivalent to change in body mass in kilograms. Each investigation is summarized in Table. Two running studies [ 11,18] and one basketball study [12] incorporated designs in which sweat loss estimations were made twice by the same participants during separate training sessions. In these studies, participants were not informed of their actual sweat losses until after study completion. Thigpen, Green and O'Neal [12] examined sweat loss estimation in short conditioning practices and longer, in-season practices, and Davis et al. [18] looked at sweat loss estimation during treadmill running under temperate versus hot conditions. Both of these studies subcomponents are presented in separate rows in Table di erence in testing conditions. The ndings of Shaver et al. [11] are not presented in two rows due to the consistency in exercise modality, duration and environmental conditions. Muth et

Davis et al. [18] looked at sweat loss estimation during treadmill running under temperate versus hot conditions. Both of these studies subcomponents are presented in separate rows in Table di erence in testing conditions. The ndings of Shaver et al. [11] are not presented in two rows due to the consistency in exercise modality, duration and environmental conditions. Muth et al. [16] also incorporated multiple sweat loss estimations within subjects, but participants were informed of their sweat loss volume after the rst estimation session. Thus, only rst sweat loss estimation accuracy scores are presented in Table 16], but data has been presented separately for male and female rugby players.

Sports2020,8, 113 3 of 12 Table 1.Synopsis of sweat loss estimation investigations. Study Participants Exercise Condition Sweat Loss Sex=n Modality: Duration (M SD; Range) Actual (L) Age (Years) Environmental Conditions (M SD; Range) Predicted (L); Prediction Method Training Status Sweat Loss as % of Body Mass % of Actual Campbell et al. (2017) [15] F=18 M=3 33 11 Recreational Hot Hatha yoga; 60 min Dry=38.7 2.7 C Relative humidity=35.8 13.3% 1.11 0.57% 0.814 0.539 0.512 0.335; 1000 mL bottles 72.5 49.0% Cronin et al. (2016) [14] F=20 M=30 30 9 Recreational CrossFit — ; 34.3 5.5 min WBGT=20.1 2.8; 16.0–22.7 C 0.91 0.31% 0.746 0.305 0.655 0.404; 250 mL bottles 96.0 60.8% Davis et al. (2019) [18] A M=12 22 2 Recreational Treadmill running-temperate environment; 60 min WBGT=18.1 0.2 C 1.72 0.29% 1.348 0.282 0.793 0.333; paper race cups 60.5 26.6% Davis et al. (2019) [18] B M=12 22 2 Recreational Treadmill running-hot environment; 60 min WBGT=25.6 0.5 C 2.43 0.56% 1.907 0.529 1.198 0.550; paper race cups 66.7 35.4% O'Neal et al. (2012) [10] F=20 M=19 41 11 Recreational to collegiate Outdoor running: 59.14 3.46 min WBGT=24.1 1.5; 21.3–27.7 C 2.11 0.52% 1.468 0.484 0.738 0.470; paper race cups 50.4 23.0% Shaver et al. (2018) [11] F=23 (2 trials) 26 6 Recreational Outdoor running 15-km time trial; 79.98 7.00 min WBGT=20.0 2.8; 12.8–25.6 C 2.20 0.48% 1.306 0.307 0.498 0.516; water bottle reference 39.9 40.9% O'Neal et al. (2014) [13] F=8 M=12 20 2 Recreational to collegiate Outdoor running; 59.91 3.42 min WBGT=20.1 2.8; 16.0–22.7 C 1.96 0.51% 1.374 0.423 0.697 0.600; paper race cups 50.1 39.1% Thigpen et al. (2014) [12] A F=11 M=11 20 1 NCAA Division II Collegiate basketball conditioning practice; (F=95 min, M=46 min) WBGT F=20.4, M=20.0 C F=1.47 0.27, M=1.13 0.27% F=1.112 0.271 M=0.969 0.250 F=0.394 0.242 M=1.316 0.847; standard practice water bottle F=37.5 28.3 M=142.6 102.4% Thigpen et al. (2014) [12] B F=11 M=11 20 1 NCAA Division II Collegiate basketball in-season practice; (F=170 min, M=170 min) WBGT F=18.5; M=17.2 C F=2.53 0.43, M=2.90 0.50% F=1.910 0.441 M=2.471 0.495 F=0.632 0.284 M=1.740 1.201;

F=20.4, M=20.0 C F=1.47 0.27, M=1.13 0.27% F=1.112 0.271 M=0.969 0.250 F=0.394 0.242 M=1.316 0.847; standard practice water bottle F=37.5 28.3 M=142.6 102.4% Thigpen et al. (2014) [12] B F=11 M=11 20 1 NCAA Division II Collegiate basketball in-season practice; (F=170 min, M=170 min) WBGT F=18.5; M=17.2 C F=2.53 0.43, M=2.90 0.50% F=1.910 0.441 M=2.471 0.495 F=0.632 0.284 M=1.740 1.201; standard practice water bottle F=35.1 20.1 M=70.6 56.9% Muth et al. (2019) [16] A F=16 20 1 NCAA Rugby Union Outdoor rugby practice; 95 min Dry=11.0 1.4 C Relative humidity=57.5 3.5% 1.19 0.55% 0.928 0.446 0.335 0.243; 532 mL cup 53.0 69.2% Muth et al. (2019) [16] B M=20 20 1 NCAA Rugby Union Outdoor rugby practice; 90 min Dry=8.5.0 2.1 C Relative humidity=46.5 9.2% 2.00 0.54% 1.917 0.516 0.373 0.183; 532 mL cup 20.9 12.5% Love et al. (2018) [17] M=20 25 4 Professional Outdoor rugby practice; 120 min Dry=35.0 C Relative humidity=40% 3.2 0.9% 3.291 0.948 0.898 0.575; 532 questionnaire 27.4 17.2% Abbreviations: F=Female, M=Male, NCAA=National Collegiate Athletic Association, WBGT=Wet Bulb Global Temperature. A, B =sub sets of data from same study.

Sports2020,8, 113 4 of 12 Statistical Analysis Following testing for equal variance, independent samplesttest were used to analyze di erences between sexes for age, weight, duration of exercise, sweat loss (absolute and as percent body mass) and sweat loss estimation accuracy. Dependentttests were used to compare absolute and estimated sweat losses within each sex and for all data combined. Three scatterplots were prepared using data from all studies. In the rst scatterplot, estimated sweat loss ((estimated sweat loss/actual sweat loss) 100) is plotted relative to actual sweat loss. The second and third plotted sweat loss estimation accuracy as a percentage against sweat loss as a percent body mass ((sweat loss/body mass) 100) with markers identifying participants by study or by sex. Additionally, a contingency table was prepared to categorize outcomes from the scatterplots. All data were analyzed and gures were created using Microsoft Excel. Because of the large sample size, an alpha of 0.01 was selected a priori. All data are presented as mean SD. 3. Results The nine studies included 243 participants (males=127; females=116) and produced 297 unique samples (Table). All participants were at least 18 years of age and varied in experience from recreational exercisers to professional athletes (Table). Exercise modalities included hot yoga, running, rugby, basketball and CrossFit — training (Table). Female particpants exhibited lower ( p<0.001) body masses, sweat loss as a percent body mass, absolute sweat losses and sweat loss estimations than male participants despite no di erence in exercise duration (Table). Actual sweat losses were approximately double the estimated sweat loss volumes and signi cantly di ered for female, male and all participants, but there were no signi cant di erences in the accuracy of predictions by percentage between male and female participants (Table). Table–3 the high prevalence of underestimation with participants experiencing greater absolute and relative sweat losses more heavily underestimating sweat loss volume. Table 2. Cumulative comparisons of sweat loss estimation variables for all participants and by sex (mean SD). Female (n=147 *) Male (n=150 **) All ( n=297) Age (years) 27.3 9.7 26.4 9.6 26.8 9.7 Weight (kg) 66.2 11.1 87.1

Table–3 the high prevalence of underestimation with participants experiencing greater absolute and relative sweat losses more heavily underestimating sweat loss volume. Table 2. Cumulative comparisons of sweat loss estimation variables for all participants and by sex (mean SD). Female (n=147 *) Male (n=150 **) All ( n=297) Age (years) 27.3 9.7 26.4 9.6 26.8 9.7 Weight (kg) 66.2 11.1 87.1 13.5y 76.7 16.2 Exercise duration (min) 77 33 74 38 75 36 Sweat loss % body mass 1.64 0.71 2.04 0.92y 1.84 0.84 Absolute (L) 1.072 0.473z 1.778 0.907yz 1.428 0.806z Estimated (L) 0.481 0.372 0.908 0.666y 0.697 0.581 Accuracy (%) 55.2 51.5 62.4 54.5 58.9 53.1 *=147 estimations were made by 116 female participants. **=150 estimations were made by 127 male participants. y=p<0.001 for comparison between sexes.z=p<0.001 for comparison made between estimated and actual sweat loss volume.

Sports2020,8, 113 5 of 12 Table 3. Contingency table for sweat loss estimation accuracy comparisons (values=n(% of total sample)). Absolute Volume Comparisons Estimated Sweat Loss (L) Actual Sweat Loss (L) 0–1 1–2 2–3 3 + 0–1 82 (27) 122 (41) 29 (9) 12 (4) 1–2 7 (2) 22 (7) 9 (3) 5 (1) 2–3 1 (0.3) 3 (1) 4 (1) 0 (0) 3+ 0 (0) 0 (0) 1 (0.3) 0 (0) Relative comparisons Estimation accuracy (%) Sweat loss as a % of body mass 0–1 1–2 2–3 3 + 0–50 17 (5) 74 (24) 55 (18) 22 (7) 50–100 16 (5) 33 (11) 21 (7) 7 (2) 100–150 10 (3) 17 (5) 4 (1) 3 (1) 150+ 12 (4) 4 (1) 2 (0.6) 0 (0)Sports 2020, 8, x FOR PEER REVIEW 6 of 13 Table 3. Contingency table for sweat loss estimation accuracy comparisons (values = n (% of total sample)). Absolute Volume Comparisons Estimated Sweat Loss (L) Actual Sweat Loss (L) 0–1 1–2 2–3 3+ 0–1 82 (27) 122 (41) 29 (9) 12 (4) 1–2 7 (2) 22 (7) 9 (3) 5 (1) 2–3 1 (0.3) 3 (1) 4 (1) 0 (0) 3+ 0 (0) 0 (0) 1 (0.3) 0 (0) Relative comparisons Estimation accuracy (%) Sweat loss as a % of body mass 0–1 1–2 2–3 3+ 0–50 17 (5) 74 (24) 55 (18) 22 (7) 50–100 16 (5) 33 (11) 21 (7) 7 (2) 100–150 10 (3) 17 (5) 4 (1) 3 (1) 150+ 12 (4) 4 (1) 2 (0.6) 0 (0) Figure 1. Individual (n = 297) sweat loss estimations versus actual sweat losses across all studies. The middle blue line represents 100% accuracy in prediction. Markers below the red line represent 50% or greater underestimation. Markers above the green line represent 50% or greater overestimation. Dashed, vertical and horizontal lines separate absolute and estimated sweat loss volumes into 1 L increments. Figure 1. Individual (n=297) sweat loss estimations versus actual sweat losses across all studies. The middle blue line represents 100% accuracy in prediction. Markers below the red line represent 50% or greater underestimation. Markers

Markers above the green line represent 50% or greater overestimation. Dashed, vertical and horizontal lines separate absolute and estimated sweat loss volumes into 1 L increments. Figure 1. Individual (n=297) sweat loss estimations versus actual sweat losses across all studies. The middle blue line represents 100% accuracy in prediction. Markers below the red line represent 50% or greater underestimation. Markers above the green line represent 50% or greater overestimation. Dashed, vertical and horizontal lines separate absolute and estimated sweat loss volumes into 1 L increments.

Sports2020,8, 113 6 of 12Sports 2020, 8, x FOR PEER REVIEW 7 of 13 Figure 2. Relationship between sweat loss presented as a percentage of body mass and sweat loss estimation accuracy (%) (n = 297) across all studies with coded identification by study. Figure 3. Relationship between sweat loss presented as percentage of body mass and sweat loss estimation accuracy (%) (n = 297) across all studies with coded identification by sex. 4. Discussion The first aim of this paper was to provide a review of the authors’ individual investigations concerning sweat loss estimation accuracy under a variety of exercise modalities and environments in a single source (Table 1). The second goal was to determine if sweat loss estimation trends across Figure 2. Relationship between sweat loss presented as a percentage of body mass and sweat loss estimation accuracy (%) (n=297) across all studies with coded identi cation by study.Sports 2020, 8, x FOR PEER REVIEW 7 of 13 Figure 2. Relationship between sweat loss presented as a percentage of body mass and sweat loss estimation accuracy (%) (n = 297) across all studies with coded identification by study. Figure 3. Relationship between sweat loss presented as percentage of body mass and sweat loss estimation accuracy (%) (n = 297) across all studies with coded identification by sex. 4. Discussion The first aim of this paper was to provide a review of the authors’ individual investigations concerning sweat loss estimation accuracy under a variety of exercise modalities and environments in a single source (Table 1). The second goal was to determine if sweat loss estimation trends across Figure 3. Relationship between sweat loss presented as percentage of body mass and sweat loss estimation accuracy (%) (n=297) across all studies with coded identi cation by sex.

A review analyzing sweat loss estimation accuracy in athletes and physically active adults.