Temporal Stability, Reproducibility and Predictability of Whole-Body Sweat Sodium Concentration During Prolonged Cycling in the Heat with Ad Libitum and Programmed Drinking

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Thank you for the opportunity to review this manuscript, which provides a secondary analysis of a previously published study, in this case examining the temporal stability, re-producibility, and predictability of sweat sodium concentration, measured using the regional patch method, during 5 hours of steady state cycling with two different drinking strategies.

Overall, the paper is well written and logical, but some sections are more complicated and/or more difficult to read than they need to be. Here are my specific comments - most are related to improving the readability of certain sections, or assisting the reader in interpretation of the findings:

Introduction

The introduction lays out the study rationale well. It does, however, suggest that it is currently unknown if sweat sodium concentration changes over time during prolonged exercise, even though it cites a study from Montain et al. in the discussion that included this analysis. While the authors do highlight methodological differences to that of Montain et al, the existence of this study should be acknowledged in the Introduction, and the reason that this is considered inadequate evidence given. Currently, the statement “…little is known about the temporal behaviour of WBSSC during prolonged exercise…” seems contradictory to the comparison made later in the manuscript.

It is also worth mentioning that an increase in sweat sodium concentration from the first hour to subsequent hours of exercise has been reported in the literature before. This includes McCubbin et. al. 2019 (consistently increased sweat sodium from the first to second hour of exercise during three different interventions, measured across 5 regional sites and whole-body estimate), and Baker et al. 2024 (the euhydrated trial, change in sweat sodium concentration from beginning to end of 90 min exercise). Although these studies used exercise durations ≤ 3 hours, they do provide evidence that sweat sodium is expected to change from the first to subsequent hours of exercise, and could be used in the Introduction to justify that there is therefore merit investigating if further increases occur over longer exercise durations.

Likewise, multiple other studies have examined the reproducibility of sweat sodium losses during repeated trials of prolonged exercise, albeit with slightly different interventions. The current study varied fluid intake while sodium intake was fixed. Other studies from Barr et al, 1991 and Sanders et al. 2001 varied sodium intake with a fixed fluid intake during 6 and 4 hours of steady state cycling respectively, and McCubbin & Costa 2023 varied sodium intake with an ad libitum fluid intake during 5 hours of steady state running. While these studies did not directly report the sweat sodium concentration values obtained in their manuscripts, they did report both fluid and sodium losses, which allows estimation of mmol of sodium lost per L of sweat as an average across the total exercise duration. These previous findings would be worth mentioning in relation to the present findings.

Methods:

Some of the methodological description could be removed, as although it was undertaken in the primary study, it is not relevant to the aims and results of this manuscript. These include:

• Measurement of rectal temperature – not reported in this manuscript
• Indirect calorimetry measurements (you can just mention the intensity as a % of VO_2max that was achieved rather than reporting the results themselves)
• Measurement of haemoglobin/haematocrit for assessment of plasma volume change – not reported in this manuscript

The statistical analysis section was quite complicated to read. My suggestion here is to break it more clearly into paragraphs that describe the statistical methods used to answer each of the three research questions, and make these links more explicit:

• Does sweat sodium concentration change over time during exercise? This is answered simply with one type of analysis and shown in Figure 1. I’m not sure the additional analysis of typical variation, and CV shown in Table 3 is necessary here to answer this research question, since the linear effects model already included post hoc/pairwise comparisons in Figure 1 where significant effects were found, and the regression models are used to predict WBSSC across the whole duration from the first timepoint, and the typical variation described for this is what is used in the interpretation of the findings.
• Is sweat sodium concentration similar from one trial to another (i.e., test-retest reliability)? The inclusion of statistical methods commonly used in test-retest reliability studies would make this easier for readers to interpret (e.g., intraclass coefficients) and allow for comparison with other studies that have examined repeatability of sweat sodium measurements. I suspect though that the sample size might be underpowered for ICC – was this the reason for not using it?
• Can the sweat sodium concentration at 40min be used to adequately predict the weighted-average sweat sodium concentration of the entire exercise bout? This is achieved by the regression models, but linking this back to the research question could be made much clearer rather than stating that these models were used “…to examine relationships between variables.” The description of each model could also be simplified so the reader can more easily link them to the research question and the methods. If I’ve understood the Methods correctly:
  • Model 1: Simply looks at the use of WBSSC at the first sweat collection (i.e., 40mins) to predict WBSSC across the entire exercise duration (using the weighted average of all 4 sweat samples).
  • Model 2: Adds uncorrected whole-body sweat rate across the entire exercise duration (as would commonly be measured in laboratory and field settings) to see if this improves the prediction of WBSSC across the entire exercise duration.
  • Model 3: Adds the interaction between the predictor variables in Model 2 to see if this further improves the prediction of WBSSC across the entire exercise duration.

Other comments for the Methods

• Line 145: “and a capillary blood sample was collected”
• Line 164-165: “Capillary blood samples were collected via finger pricks.” – this is unnecessary here (in the section of how blood and urine samples were analysed), and the details should appear earlier at Line 145 (describing how the blood samples were collected).
• Line 165: “Whole-body sweat losses were measured..” this should be “Whole-body sweat losses were calculated…”, as the “measurement” was the collection of body mass before/after exercise as described earlier in Lines 138 and 157.
• Line 169: I assume the 15 cm² patches were the 6 x 10cm patches, but what is described is just the absorbent part? Perhaps make that clearer, as from my experience when people discuss or purchase the patches, the dimensions described include the film. So you could say “Local sweat production was collected using regional patches (Tegaderm + Pad 6 x 10cm, 3M, USA) with a 15cm² absorbent pad” or something similar.
• Line 170: By “sterile water” to you mean deionised water?
• Lines 199-200: The second half of this sentence doesn’t make grammatical sense. Consider re-wording this.

 

Results

As mentioned above, there are data here that are probably unnecessary to be reported:

• Details of VO₂ and % VO_2max in both trials (this could just be described in the methods)
• The paragraph/analysis at lines 275-281, and Table 3, as mentioned above.

In Section 3.3 it would be helpful here to report the total fluid deficit (in L and/or % BM loss) and final plasma osmolality and or sodium in addition to % fluid replacement, as most readers are more familiar with these data and can more easily understand the effect of the ad libitum vs planned drinking on hydration status (which becomes relevant to your discussion later on). I know this data was reported in the original paper, but you don’t want to expect readers to go and find that paper to understand the context of hydration status in this study. This could be added to an expanded Table 2.

The section of text from lines 311-318 is difficult to follow because of the inconsistency in language and terminology used. The first part is described but not labelled in relation to any model (i.e., Model 1, Model 2, or Model 3). Using “R²” for two models and “coefficient of determination” for another in the text makes it harder for the reader to easily pick out the main outcomes and directly compare them. And the R² value is provided for Model 1 and Model 3, but for Model 2 it is only described how much the R² is increased from Model 1. My suggestion here is to clearly articulate each model (i.e. “Model 1…”, “Model 2…”, etc.), and present the outputs of each model (regression equation or the beta coefficients, R², p value and/or 95% CI’s for the beta coefficient, SEE) in exactly the same way so they can be easily understood and compared side-by-side (a table would probably be the easiest way to do this).

Figure 2 with the two x axes would be much easier to read as two separate panels, side-by-side, with only one x-axis on each. Currently it requires the reader to very carefully go back and forth between the x axis labels, the trend lines, the captions, and the regression equations to make sure they’re reading it correctly. Having two separate panels would simplify this immensely, and the reader could understand both relationships in seconds without having to check and double check they’re reading the right thing.

 

Discussion

Overall the discussion does a good job of summarising the findings and their usefulness in practice.

One suggestion to consider that wasn’t mentioned, is whether the exercise intensity and/or environmental conditions are likely to alter the time it takes to achieve steady state sweating during exercise, and if the 40 minute timeframe would be adequate (or more than adequate) in different conditions/different rates of metabolic heat production?

For the paragraph in lines 408-428 and the Abstract and Conclusion, it wasn’t clear whether you are advocating that practitioners apply the regression equation from Model 1 (i.e., y = 1.3397x + 0.2738), or that they could simply use the raw WBSSC value obtained at 20-40min as a proxy for WBSSC for the rest of the exercise duration (i.e. y ≈ x with no correction)? You mention “…can be predicted…” several times, but it’s not clear if the regression equation is considered necessary to make that prediction.

 

 

References mentioned above:

Baker et al. 2024. Physiol Rep. 12, e16174. (already cited elsewhere in the manuscript)

McCubbin et. al. 2019. Eur J Appl Physiol. 119(9):2105-2118.

Barr et al, 1991. Med Sci Sports Exerc. 23(7):811-817.

Sanders et al. 2001. Eur J Appl Physiol. 84(5):419-425.

McCubbin & Costa 2023. Int J Sports Physiol Perf. 19(2):105-115.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Dear Authors,

Your manuscript investigates the temporal stability, reproducibility, and predictability of whole-body sweat sodium concentration (WBSSC) during 5 hours of cycling in a warm environment. Your study addresses a significant gap in sports nutrition for ultra-endurance athletes, providing rare longitudinal data. While the experimental execution and writing quality are commendable, there are critical statistical flaws in the predictive modeling and methodological limitations regarding data extrapolation that must be addressed to ensure the validity of the conclusions.

MAJOR ISSUES
1. Mathematical Part-Whole Correlation Bias (Model 1): The predictability analysis (Section 3.5) contains a fundamental logical error. The authors regressed the 40-minute sodium value ($x_1$) against the time-weighted mean ($y$). However, y is calculated using the 40-minute point as one of its constituents. This creates an artificially inflated correlation (R2 = 0.87) because x1 is being used to predict itself. For a valid claim of 'predictability,' you must redo the regression by using the 40-min sample to predict the mean of the subsequent time points only (130, 220, and 290 min).

2. Risk of Overfitting (Model 3): Model 3 includes three predictor terms (WBSSC 40 min, Sweat Rate, and their Interaction) for a sample size of only 8 participants. Multivariate statistical standards generally require a minimum of 10–15 observations per predictor to avoid fitting the model to random noise. With n = 8, the reported R2 of 0.93 is likely an artifact of overfitting and may not be generalizable. Please, should present this model as exploratory or remove it.

3. Extrapolation of Baker (2009) Equations: Your study converts regional sweat (forearm) to WBSSC using equations validated for 90-minute protocols. The authors apply these same equations to samples collected after 4 or 5 hours of exercise. There is no evidence in current literature that the regional-to-whole-body ratio remains constant or linear during ultra-endurance efforts, where heterogeneous sudomotor fatigue may occur. This limitation must be explicitly discussed.

4. Justification for the Secondary Analysis: As this is a secondary analysis of a 2021 trial (Jeker et al.), you must more clearly differentiate the "added value" of this manuscript (nutrients-4151267). Since sweat sodium concentrations were briefly mentioned in the primary paper as having no differences between conditions, you must emphasize that the novelty here lies specifically in the temporal kinetics and methodological reproducibility to avoid the perception of redundant publication.

MINOR ISSUES
5. Sample Generalizability: Your study is restricted to highly trained men (VO2peak of 67 mL/kg/min). This elite fitness level and the absence of women or recreational athletes limit the universal applicability of the predictive equations. Your conclusions regarding practical implementation should be more conservative.
6. Handling of Missing Data: Your manuscript mentions one missing value for osmolality and two for WBSSC. You should specify in the Methods section how these were handled within the Linear Mixed-Effects Models (e.g., listwise exclusion or imputation).
7. Acclimatization State: Your participants were tested outside of summer months, and you described them as unacclimatized. The observed WBSSC stability might differ in athletes undergoing active heat acclimation, where ductal sodium reabsorption is dynamic. A brief mention of this in the limitations is suggested and recommended.
8. Patch Saturation and Local Mechanics: 20 minutes of application at sweat rates of ~1.2 L/h could potentially lead to local skin hydromeiosis or patch saturation. You confirm that the absorbent capacity was not exceeded? Please, explain.

In my opinion, your study, entitled "Temporal Stability, Reproducibility and Predictability of Whole-Body Sweat Sodium Concentration During Prolonged Cycling in the Heat with Ad Libitum and Programmed Drinking," has the potential to serve as a benchmark due to the rarity of 300-minute physiological data. However, the clinical utility of the proposed equations depends entirely on correcting for regression bias and being transparent about the statistical limitations imposed by small sample sizes.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

This manuscript examines the temporal stability, reproducibility, and predictability of whole-body sweat sodium concentration during prolonged cycling in the heat using a randomized crossover design in highly trained endurance athletes. The research question is practically relevant for sodium replacement strategies in endurance exercise, and the controlled laboratory design with repeated within-subject measurements represents a methodological strength. The observation of relatively small within- and between-trial variability in model-derived WBSSC, together with the potential utility of an early-exercise sweat sample for estimating mean exercise values, may provide useful applied insight for practitioners.

 

However, several methodological and interpretative limitations currently restrict the robustness, inferential strength, and generalizability of the conclusions. In particular, the very small sample size, the reliance on predictive modelling without independent validation, and the uncertainty inherent in the local-to-whole-body estimation of sweat sodium concentration require more cautious interpretation. Clarifying these aspects and more explicitly framing the findings as context-specific and preliminary would substantially strengthen the scientific rigor and applied relevance of the manuscript.

 

Overall, the study addresses a meaningful applied physiology question but requires substantial revision before firm applied conclusions can be supported.

 

Abstract

Lines 26–27: Local sweat sodium concentration is converted to WBSSC, yet the abstract does not specify the conversion model or validation context, which is central to interpretation of whole-body sodium loss. Clarification is recommended.

 

Lines 32–35: The conclusion that WBSSC “varies trivially” and “can be predicted using a single sweat sample” appears stronger than supported by the design, particularly given the small sample and absence of external validation. Consider moderating causal certainty and emphasizing context-specific estimation.

 

Introduction

Lines 49–59: The modelling framework for sodium replacement is well summarized; however, thresholds derived from modelling should be clearly framed as model-dependent rather than universally applicable to all endurance contexts.

 

Lines 67–72: Because WBSSC measurements originate from previously collected data not reported in the original trial, the present study represents a secondary analysis, which should be emphasized explicitly due to implications for statistical planning and interpretative scope.

 

Methods

Lines 181–183: WBSSC is derived using a regression equation from regional sweat sampling. Greater transparency is needed regarding:

• validation of the equation under prolonged cycling heat stress
• potential estimation error and uncertainty propagation

 

as these directly influence the certainty of conclusions regarding whole-body sodium loss.

 

Lines 227–231: The a posteriori power calculation based on n = 8 and assumed effect size does not fully address precision or regression stability, particularly for predictive modelling. This limitation should be acknowledged more explicitly.

 

Results

Lines 268–274: Temporal stabilization of WBSSC after 40 min is clearly demonstrated; however, interpretation should clarify that stabilization reflects model-derived WBSSC, not direct whole-body sodium secretion.

 

Lines 279–287: Reported coefficients of variation (~13–17%) are described as physiologically trivial, yet such interpretation requires explicit linkage to sodium-replacement decision thresholds, rather than statistical magnitude alone.

 

Lines 314–327: High R² values for prediction from the 40-min sample are derived from within-session modelling without independent validation, raising the possibility of overfitting and optimistic prediction accuracy in a very small sample. Explicit acknowledgment and/or internal validation would strengthen confidence.

 

Discussion

Lines 365–391: Statements asserting physiologically negligible WBSSC variation and practical sodium-replacement implications appear over-generalized relative to the limited experimental context and small sample. Stronger emphasis on uncertainty and context specificity is recommended.

 

Lines 408–428: Predictive conclusions regarding early sweat sampling should be framed as hypothesis-generating, since predictive validity across separate exercise sessions or field testing remains unverified.

 

Lines 429–442: Limitations are appropriately acknowledged; however, their implications for inferential strength and applied recommendations could be emphasized more explicitly.

 

Conclusions & Perspectives

Lines 444–461: Practical recommendations regarding sodium replacement appear strong relative to the evidential base of a small, homogeneous laboratory sample. Consider moderating language to reflect context-specific preliminary evidence.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

Dear Authors

Thank you for the thorough and thoughtful revisions. All my major concerns from the first review have been adequately addressed, significantly strengthening the manuscript.

Specific Responses to Revisions:
1. Mathematical Part-Whole Bias: Formula clarification confirms no self-prediction issue. Well addressed.
2. Risk of Overfitting (Models 2/3): Excellent decision to remove these models, demonstrating appropriate statistical conservatism.
3. Baker (2009) Equations Extrapolation: Limitation now explicitly discussed (Lines 458-461).
4. Secondary Analysis Justification: Novelty clearly differentiated from primary paper (Lines 77-81).
5. Sample Generalizability: Conclusions appropriately tempered for elite male population (Lines 461-487).
6. Patch Saturation: Quantitative verification (mean mass 0.741g) confirms adequacy.
7. All minor points (missing data, acclimatization, etc.) adequately addressed.

Minor Suggestions (non-mandatory):
1. Consider adding a sentence in the Discussion about potential applications for field-based ultra-endurance events (e.g., Ironman).
2. Table 3/4 formatting is improved, but consider bolding statistically significant p-values for emphasis.

The manuscript is now publication-ready. No further revisions required.

Author Response

"Please see the attachment."

Author Response File: Author Response.pdf