Review Reports - Evaluation of Five Bioelectrical Impedance Analysis Equations Against Air-Displacement Plethysmography in an Eastern European Population

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The study compares five equations for predicting lean body mass derived from bioimpedance analysis (BIA) with values measured using air displacement plethysmography (ADP) in 101 Caucasian adults from Eastern Europe. The equations tested include the proprietary equation of the Maltron device and four from the literature. The comparison uses bias, limits of agreement, Pearson's correlation coefficient, and concordance correlation coefficient. The Deurenberg formula has the smallest bias and best overall agreement, while the others show more marked overestimations or proportional biases. The authors are requested to provide the necessary additions and responses to the comments below.

1. What practical steps do you envisage to test the generalisability of the Deurenberg formula or any new equations in other centres or regions of Romania, with ethnic, socioeconomic or lifestyle differences?

2. Do you plan to replicate the study with a more distributed sample?

3. Have you considered, or do you plan to explore, the use of machine learning models that integrate raw BIA parameters and anthropometric measurements to create equations that are more adaptable to local subpopulations?

4. The authors should integrate advanced modelling and artificial intelligence methods for monitoring and instrumental validation into the discussion. Such integration, with validation from recent studies, is also useful in the nutritional or body composition context to improve the generalisability of predictive equations.

5. Authors should also integrate the section on clinical and application relevance with the study doi:10.3390/app15084306 because MEMS and IoT technology can facilitate continuous monitoring of mobility and health status, related to changes in body composition. Linking BIA validation to contexts where the technology is integrated into care systems could influence the use of the best equation and the need for adaptations.

6. Did the authors consider analysing impedance signals at multiple frequencies or applying deep learning methods to raw data to reduce bias or improve accuracy? In this regard, I do not expect the authors to develop a model from scratch, but only to integrate the discussion on the potential development of new BIA signal analysis algorithms with the study doi: 10.3390/signals6030038, which, although focused on a different type of analysis, illustrates how image analysis techniques and neural networks can be used to evaluate quality parameters with non-linear or signal-based approaches.

7. How do the authors intend to bridge the age gap, especially for those over 60? Are there plans for a specific study on the elderly, given that body composition and hydration vary with age and may influence the equations?

8. It is stated that hydration remains a source of uncertainty. In this regard, do the authors have data or analyses on intra-day, seasonal or post-exercise variations in the sample? Is it possible to quantify how much these factors influence bias or limits of agreement between equations and ADP, and suggest more standardised protocols?

9. A Maltron multi-frequency device is used. How do the authors assess the transferability of the results to other commercial BIA devices? Is validation necessary for each model, or is it possible to create a general correction factor? Could you propose guidelines on the choice of devices?

10. You use CCC, Pearson, Bland–Altman; robust results. Did the authors explore multivariate regression or calibration models to correct for systematic errors? In particular, given that some equations show proportional or FFM-dependent biases, is it possible to apply a linear or non-linear adjustment?

11. In addition to gender, did the authors evaluate subgroups by BMI or physical activity status? It may emerge that some equations work better in specific ranges of BMI or lean mass; if so, differentiated use can be proposed.

12. Could the authors supplement the paper with a more extensive comparison table with results from other European countries, highlighting not only bias but also limits of agreement, to facilitate the choice of equations in similar contexts?

13. What would be the proposed workflow for clinicians or researchers adopting the Deurenberg formula? Do you recommend applying it directly, or mediating with other parameters? What quality or data control measures do you recommend during daily use?

Author Response

We thank the reviewer for their careful reading of our manuscript and for the valuable comments provided. We have addressed all points raised, as detailed below. Changes in the manuscript are highlighted where appropriate.

What practical steps do you envisage to test the generalisability of the Deurenberg formula or any new equations in other centres or regions of Romania, with ethnic, socioeconomic or lifestyle differences?

Thank you for the useful suggestions.

We added in the Introduction information about particularities in the Romanian population and in the Discussion section the future directions of our research addressing the topic (rows 85- 88 and 372-375).

Do you plan to replicate the study with a more distributed sample?

Indeed, twe aim to continue our research on a larger sample, allowing us to refine the statistical analysis (rows 369- 375).

Have you considered, or do you plan to explore, the use of machine learning models that integrate raw BIA parameters and anthropometric measurements to create equations that are more adaptable to local subpopulations?

The aim of the present study was to identify the best predictive equation in a Romanian population, as recommended by ESPEN guidelines. Nevertheless, future research may include more advanced techniques and larger samples allowing the use of algorithms and machine learning models.

The authors should integrate advanced modelling and artificial intelligence methods for monitoring and instrumental validation into the discussion. Such integration, with validation from recent studies, is also useful in the nutritional or body composition context to improve the generalisability of predictive equations.

We added a paragraph concerning the subject (rows 376- 381). Thank you for the reference you suggested.

Authors should also integrate the section on clinical and application relevance with the study doi:10.3390/app15084306 because MEMS and IoT technology can facilitate continuous monitoring of mobility and health status, related to changes in body composition. Linking BIA validation to contexts where the technology is integrated into care systems could influence the use of the best equation and the need for adaptations.

The use of IoT technology is an interesting emerging subject in body composition literature. We introduced a paragraph concerning the perspectives on using body composition prediction equations in AI based models (rows 376- 381).

Did the authors consider analysing impedance signals at multiple frequencies or applying deep learning methods to raw data to reduce bias or improve accuracy? In this regard, I do not expect the authors to develop a model from scratch, but only to integrate the discussion on the potential development of new BIA signal analysis algorithms with the study doi: 10.3390/signals6030038, which, although focused on a different type of analysis, illustrates how image analysis techniques and neural networks can be used to evaluate quality parameters with non-linear or signal-based approaches.

Multi-frequency BIA is used to determine several body composition parameters. However, the majority of equations in the literature addressing the foot-to-hand technique use 50 kHz, as at this frequency the current passes through both intra- and extracellular fluid, allowing the prediction of total body water, as well as fat mass and fat-free mass. Multiple frequencies are tipically used in segmental techniques, which were not employed in the present study study. Nevertheless, this topic is of interest and may be addressed in future research.

How do the authors intend to bridge the age gap, especially for those over 60? Are there plans for a specific study on the elderly, given that body composition and hydration vary with age and may influence the equations?

We added future perspectives in the Discussion section (rows 366- 370))

It is stated that hydration remains a source of uncertainty. In this regard, do the authors have data or analyses on intra-day, seasonal or post-exercise variations in the sample? Is it possible to quantify how much these factors influence bias or limits of agreement between equations and ADP, and suggest more standardised protocols?

There are several studies in the literature concerning the topic. The cross-sectional design of the present study does not allow to analyse these variations. However, since these factors influence body composition parameters, there have been created guidelines for clinical, as well as for research purposes. We included a short paragraph mentioning the guidelines used in the present study (rows 153- 154)

A Maltron multi-frequency device is used. How do the authors assess the transferability of the results to other commercial BIA devices? Is validation necessary for each model, or is it possible to create a general correction factor? Could you propose guidelines on the choice of devices?

We added a paragraph to address this issue (rows 359- 363)

You use CCC, Pearson, Bland–Altman; robust results. Did the authors explore multivariate regression or calibration models to correct for systematic errors? In particular, given that some equations show proportional or FFM-dependent biases, is it possible to apply a linear or non-linear adjustment?

The equations used to predict FFM were developed using linear regression models. The aim of the article was to identify the best predictive formula in a Romanian population, in order to use it in clinical settings. The study sample was to small to allow the development of a new formula (since a validation sample would also be necessary).

In addition to gender, did the authors evaluate subgroups by BMI or physical activity status? It may emerge that some equations work better in specific ranges of BMI or lean mass; if so, differentiated use can be proposed.

The size of the study sample and distribution of BMI did not allow a more detailed analysis. Although information about physical activity was collected, the responses of the participants were considered rather subjective. (rows 372- 375)

Could the authors supplement the paper with a more extensive comparison table with results from other European countries, highlighting not only bias but also limits of agreement, to facilitate the choice of equations in similar contexts?

Table S2 has been included in the Supplementary Materials to address this issue.

13. What would be the proposed workflow for clinicians or researchers adopting the Deurenberg formula? Do you recommend applying it directly, or mediating with other parameters? What quality or data control measures do you recommend during daily use?

Taking into account the results of our study, the formula by Deurenberg represents a viable alternative to the inbuilt Maltron formula. Although its use may be considered time-consuming in clinical settings (given the supplementary computation), it provides more accurate prediction. While guidelines (such as ESPEN) recommend testing and selecting the most appropriate equation for a given population, the final choice ultimately rests with the individual researcher/ clinician. (rows 357- 360)

Reviewer 2 Report

Comments and Suggestions for Authors

Dear Authors

Thank you for your submission your manuscript titled “Evaluation of Five Bioelectrical Impedance Analysis Equations Against Air-Displacement Plethysmography in an Eastern European Population” to this journal. Your work is valuable although my comments and my suggestion can improve your study.

Strengths:
Clear and relevant research objective.
Appropriate and well-documented methodology.
Comprehensive statistical analysis, including Bland-Altman plots, CCC, and subgroup analyses.
Well-structured discussion that contextualizes findings within existing literature.
Ethical compliance and transparent reporting.

Major Comments:
Introduction
Consider adding a brief justification for focusing on the Romanian population beyond genetic diversity—e.g., are there known anthropometric, dietary, or lifestyle factors that may influence body composition in this group?

Methods
Please clarify whether the same operators performed both BIA and ADP measurements for all participants. If not, was inter-operator reliability assessed?
In the BIA electrode placement description, referencing a recognized standard protocol (e.g., ESPEN or NIH guidelines) would enhance reproducibility.
Consider mentioning whether participants were measured in a fasting state consistently across all assessments.

Results
In Table 3, please add a brief footnote to define “ΔFFM” clearly for readers.
The Bland-Altman plots (Figure 1) are informative, but the regression equations and R² values within the plots could be more clearly integrated into the figure legend for easier interpretation.
When reporting CCC values, consider adding a brief interpretive phrase (e.g., “very strong concordance”) in the text to aid readability.

Discussion
While you note that the Deurenberg equation performed best, a deeper exploration of why—beyond the shared density-based reference method—would be valuable. For instance, does its inclusion of age and sex improve adaptability across subgroups?
The limitation regarding the underrepresentation of older adults is appropriately noted. Please consider adding a sentence on how this may affect the generalizability of your findings to elderly Romanians.

Thank you again

Comments on the Quality of English Language

Dear Authors

Thank you for your submission of your manuscript titled “Evaluation of Five Bioelectrical Impedance Analysis Equations Against Air-Displacement Plethysmography in an Eastern European Population” to this journal. The English of your work is fine, although my comments and suggestions can improve that.

Edits and Suggestions

Consistency in Terminology:
Use “BIA” consistently after its first introduction (avoid alternating with “bioelectrical impedance analysis” in mid-sentence unless for clarity).
“Fat-free mass” is sometimes abbreviated as FFM and sometimes spelled out; consider using the abbreviation after its first definition.

Sentence Flow:
Some longer sentences could be broken into shorter ones for improved readability.
Example (Page 2):
Original: “BIA determines the impedance (Z) of the body by applying a low intensity alternating current, typically across a range of 1 to 1000 kHz.”
Suggestion: “BIA determines the impedance (Z) of the body by applying a low-intensity alternating current, typically within a frequency range of 1 to 1000 kHz.”

Minor Grammatical/Phrasing Adjustments:
Page 3: “subjects were instructed to void” → consider “subjects were asked to empty their bladders” for clinical clarity.
Page 6: “the biases obtained with all the other four formulas exceeded those observed in the entire study group” → consider rephrasing for clarity: “the biases for the other four formulas were larger in women than in the overall group.”

Typographical Corrections:
“Matron” → “Maltron” (page 7, line 3).
“between-de-vice” → “between-device” (page 8, line 307).
“Feenales” → “Females” (Table 1).

Punctuation and Spacing:
Ensure consistent use of spaces around “±” (e.g., “25.37 ± 5.45”).
Use en-dashes for ranges (e.g., “18–71 years”) consistently.

Thank you again

Author Response

We sincerely thank the reviewer for their careful evaluation of our manuscript and for providing constructive and insightful comments. We have addressed all points raised and made revisions where appropriate. Changes in the manuscript are highlighted, and responses to each comment are detailed below.

Introduction
Consider adding a brief justification for focusing on the Romanian population beyond genetic diversity—e.g., are there known anthropometric, dietary, or lifestyle factors that may influence body composition in this group?

Thank you for the useful suggestions.

We added in the Introduction information about obesity incidence (rows 45- 47 ) and dietary and lifestyle factors in Romania (rows 85- 88 ). We aim to continue research with a larger sample, allowing a more in depth analysis (e.g. BMI groups), especially since the formulas by Kanellakis and Heitmann showed a proportional bias.

Methods
Please clarify whether the same operators performed both BIA and ADP measurements for all participants. If not, was inter-operator reliability assessed?

The same two operators performed all measurements, working in a team. We tried to specify this more clearly (rows 142 and 153- 154)

In the BIA electrode placement description, referencing a recognized standard protocol (e.g., ESPEN or NIH guidelines) would enhance reproducibility.

Measurements were performed according to ESPEN guidelines. We added rows 153- 154.

Consider mentioning whether participants were measured in a fasting state consistently across all assessments.

We mentioned under 2.1 Study group, rows 117- 118.

Results
In Table 3, please add a brief footnote to define “ΔFFM” clearly for readers.

ΔFFM is defined as FFMBIA-FFMADP in row 223 (added), as well as in the footnote, table 3, and table S1 supplementary materials.

The Bland-Altman plots (Figure 1) are informative, but the regression equations and R² values within the plots could be more clearly integrated into the figure legend for easier interpretation.

We included the equations in the figure legend.

When reporting CCC values, consider adding a brief interpretive phrase (e.g., “very strong concordance”) in the text to aid readability.

We added interpretation of the CCC values in the Results section.

We added a brief possible interpretation regarding age in Discussions. However, sex is included as a variable in all four equations. Since we observed a proportional bias for two of the formulas, the differences may be more attributable to differences in BMI distributions between the populations used to develop the formulas and our study sample than to age or sex. A brief possible explanation regarding sex was added in the corresponding rows 331- 333.

The limitation regarding the underrepresentation of older adults is appropriately noted. Please consider adding a sentence on how this may affect the generalizability of your findings to elderly Romanians.

We added a brief statement outlining future research directions for this age group (rows 369-375).

Consistency in Terminology:
Use “BIA” consistently after its first introduction (avoid alternating with “bioelectrical impedance analysis” in mid-sentence unless for clarity).

Terminology consistency was verified.

“Fat-free mass” is sometimes abbreviated as FFM and sometimes spelled out; consider using the abbreviation after its first definition.

We checked for terminology consistency.

We modified as per your suggestions.

Minor Grammatical/Phrasing Adjustments:
Page 3: “subjects were instructed to void” → consider “subjects were asked to empty their bladders” for clinical clarity.

We corrected.

Page 6: “the biases obtained with all the other four formulas exceeded those observed in the entire study group” → consider rephrasing for clarity: “the biases for the other four formulas were larger in women than in the overall group.”

We corrected.

Typographical Corrections:
“Matron” → “Maltron” (page 7, line 3).
“between-de-vice” → “between-device” (page 8, line 307).
“Feenales” → “Females” (Table 1).

We corrected.

Punctuation and Spacing:
Ensure consistent use of spaces around “±” (e.g., “25.37 ± 5.45”).
Use en-dashes for ranges (e.g., “18–71 years”) consistently.

We corrected.

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The authors have given good answers to the review comments and have revised the paper well.