Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This manuscript examines occupational disparities in lifestyle behaviors and adiposity among working women in Peru using pooled DHS data. The topic is relevant and timely, particularly in the context of noncommunicable diseases and gendered health inequalities in low- and middle-income countries. The large nationally representative sample (n = 40,726) and the attempt to jointly examine behavioral and anthropometric outcomes are notable strengths.

However, the current version remains largely descriptive, and several issues related to theoretical positioning, methodological transparency, and interpretation of findings need to be addressed. Without these revisions, the contribution of the study to the literature remains limited.

The study relies exclusively on crude models without adjustment variables, which severely limits interpretability.While the authors justify this as descriptive, the manuscript frequently discusses mechanisms (e.g., stress, workload, gender burden), which are not tested empirically.
The introduction and discussion imply an interest in mechanisms and inequalities, yet:No mediation/moderation analysis、No stratified or interaction analysis
Although the study pools multiple survey rounds (2014–2024), it is still treated as a repeated cross-sectional design without temporal analysis
Clarify why 2020 data were excluded beyond comparability (add sensitivity note)
Table/Figure referencing could be more precise
Abstract should include more quantitative results (effect sizes)
Consider adding limitations about selection bias (healthy worker effect)

Author Response

We thank the reviewers for their careful evaluation of the manuscript and for the constructive comments. We revised the manuscript and Supplementary Materials accordingly. All textual changes in the manuscript and supplementary document are highlighted in yellow. The main revisions include: (i) correction of the eligible age range to 18–49 years, consistent with the ENDES/DHS women’s questionnaire and the analytic dataset; (ii) addition of secondary adjusted models; (iii) addition of missing-data sensitivity analyses; (iv) addition of exploratory occupation-by-period interaction tests; (v) expanded explanation of the structural role of occupation, central adiposity, private households, data linkage, consent/anonymization, and measurement limitations; and (vi) more cautious interpretation of mechanisms that were not directly measured.

Reviewer 1

Comment 1. The study relies exclusively on crude models without adjustment variables, which severely limits interpretability. While the authors justify this as descriptive, the manuscript frequently discusses mechanisms that are not tested empirically.

Response. We agree that crude models alone limited interpretability. We retained crude estimates as the primary descriptive contrasts, because the objective is to describe observed occupational disparities rather than estimate the causal effect of occupation. However, we added secondary adjusted models including age group, survey year, education, wealth quintile, area of residence, natural region, and marital status. These results are now presented in the revised Results section and Supplementary Tables S8 and S9. We also revised the Discussion to explicitly frame stress, unpaid burden, job control, and other work-related mechanisms as plausible explanations, not empirically tested mechanisms.

Comment 2. The introduction and discussion imply an interest in mechanisms and inequalities, yet no mediation/moderation, stratified, or interaction analysis was conducted.

Response. We did not perform formal mediation analysis because the proposed mediators, such as job stress, working hours, childcare burden, informality, and unpaid domestic work, were not directly measured in ENDES. Estimating mediation without measured mediators would be methodologically inappropriate. To address the concern about heterogeneity, we added exploratory occupation-by-period interaction tests for selected outcomes with sufficient event frequency. These are reported in Supplementary Table S12 and summarized in the Results. We also revised the Discussion to separate structural determinants from behavioral pathways.

Comment 3. Although the study pools multiple survey rounds, it is treated as a repeated cross-sectional design without temporal analysis.

Response. We added an exploratory temporal analysis. Survey years were grouped into broad periods (2014–2016, 2017–2019, and 2021–2024; for WC outcomes, 2018–2019 and 2021–2024) and occupation-by-period interactions were tested for selected outcomes. The analysis suggested temporal heterogeneity for television viewing and obesity, whereas other outcomes showed no strong evidence of changing occupational patterns. These results are presented in Supplementary Table S12 and summarized in Section 3.5.

Comment 4. Clarify why 2020 data were excluded beyond comparability and add a sensitivity note.

Response. We expanded the Study Design section. The revised text explains that 2020 was excluded a priori because the COVID-19 pandemic introduced fieldwork disruptions and a population-level shock to work, mobility, health behaviors, and anthropometric measurement conditions. Therefore, including 2020 would not have represented a neutral sensitivity analysis but would have mixed regular survey years with an atypical pandemic round.

Comment 5. Table/Figure referencing could be more precise.

Response. We corrected the internal references to the supplementary tables and figures. In particular, behavioral prevalence now refers to Table S2, crude behavioral models to Table S3, disaggregated smoking/alcohol models to Table S4, WC/AO descriptives to Table S5, crude obesity/BMI models to Table S6, and crude AO/WC models to Table S7. The Supplementary Materials list was updated to include Tables S8–S12.

Comment 6. The abstract should include more quantitative results.

Response. We revised the Abstract to include 95% confidence intervals for key crude estimates, to mention the addition of secondary adjusted models, and to summarize the adjusted obesity findings.

Comment 7. Consider adding limitations about selection bias, including the healthy worker effect.

Response. We expanded the Limitations section to explicitly acknowledge selection into or out of work based on health status, including a potential healthy-worker or survivor effect.

We hope that the revised manuscript addresses the reviewers’ concerns and improves the clarity, methodological transparency, and interpretability of the study.

Sincerely,

The Authors

Reviewer 2 Report

Comments and Suggestions for Authors

Dear Authors,
First of all, I would like to congratulate you on your interesting and valuable work. Please find below my comments and suggestions.
Abstract
This section is clear and well written, and I have no major comments. However, I would suggest adding “disparities” as a keyword.

Introduction

This section is very clear and well structured. I have a few suggestions that may help strengthen it further:
-The authors refer to central adiposity as a marker, but they do not specify how it is operationalized in the study (e.g., waist circumference). Clarifying this would improve conceptual precision.
-It would also be useful to elaborate on how occupation functions as a structural determinant of health. Making this explicit would enhance the theoretical coherence of the section.
Material and Methods

-Study design-This section is clear. Nothing to add
-Data source-This section is clear. Nothing to add
-Study population and eligibility criteria-This section is clear and well presented. I only have one suggestion: it would be helpful to provide a more detailed explanation of how “private households” are defined and how the health module is structured or linked to the main questionnaire.
-Main exposure-This section is clear and well written, and I have no further comments.
-Outcomes- This section is clear and well written, and I have no further comments.
-Covariates-This section is clear and well written, and I have no further comments.
-Data Preparation and Cleaning-This section is clear and well written, and I have no further comments.
-Statistical Analysis- This section is clear and well written, and I have no further comments.
-Ethical Considerations-This section is clear. I would also suggest that the authors provide additional information on the type of informed consent obtained (e.g., oral or written) and offer a more detailed description of the anonymization procedures applied to the data.

Results
-Participants selection-The information is clear; however, in my view, this section would be more appropriately placed in the Methods section.
-Sample Characteristics-This section is clear, and I have no further comments. However, I would suggest presenting Table 1 in a horizontal (landscape) format to improve readability and overall presentation.
-Lifestyle Behaviors-This section is clear and well written, and I have no further comments.
-Adiposity levels- This section is clear and well written, and I have no further comments.

Discusion
The Discussion is well structured, comprehensive, and clearly linked to the study findings and existing literature. The interpretation is generally strong and coherent. However, a few aspects could be further strengthened in terms of conceptual framing and clarity of interpretation.

-The distinction between lifestyle behaviors and adiposity is well presented; however, the mechanisms explaining the mismatch between both could be elaborated further.
-The finding on WC vs BMI is interesting, but the interpretation of divergence between indicators would benefit from a more explicit conceptual explanation.
-The explanation of agricultural work could further distinguish between physical activity at work vs overall energy balance.
-The gender perspective is appropriate and important, but could benefit from a slightly clearer distinction between structural determinants and behavioural pathways

Limitations
This section is clear. I would only suggest explicitly addressing potential measurement bias arising from the self-reported nature of all behavioural variables, ideally in a more unified way within the limitations section.

Conclusions
This section is clear and well written, and I have no further comments.

I hope the suggestions will help improve the manuscript. Good luck!

Author Response

Reviewer 2

Comment 1. Add “disparities” as a keyword.

Response. We added “health disparities” to the Keywords section.

Comment 2. Clarify how central adiposity is operationalized in the study.

Response. We revised the Introduction to state that central adiposity was operationalized through waist circumference and abdominal obesity, defined as waist circumference ≥88 cm in women. The Outcomes section continues to provide the formal definition.

Comment 3. Elaborate on how occupation functions as a structural determinant of health.

Response. We added a paragraph to the Introduction explaining occupation as a structural determinant that organizes exposure to material resources, time constraints, physical workload, job control, psychosocial demands, social protection, and unpaid caregiving responsibilities among women.

Comment 4. Provide a more detailed explanation of how “private households” are defined and how the health module is structured or linked to the main questionnaire.

Response. We expanded the Study Population and Eligibility Criteria section. The revised text explains that private households refer to non-institutional dwellings selected in urban and rural areas and clarifies the person-level linkage between the women’s individual questionnaire and the health module using household identifier and respondent line number.

Comment 5. Provide additional information on informed consent and anonymization procedures.

Response. We revised the Ethical Considerations section and the Informed Consent Statement. The revised manuscript states that the original ENDES/DHS data collection used standardized procedures for consent/refusal recording, and that the present analysis used only public-use de-identified microdata without direct personal identifiers.

Comment 6. The participant selection section would be more appropriately placed in Methods.

Response. We added more detail on eligibility and sample derivation in the Methods section and renamed the Results subsection as “Analytic Sample Derivation.” We retained a concise derivation summary in Results because reporting the final numbers at each analytic stage is consistent with STROBE reporting practice and links directly to Figure S1.

Comment 7. Present Table 1 in landscape format.

Response. Table 1 was reformatted in a landscape section to improve readability.

Comment 8. Elaborate on the mismatch between lifestyle behaviors and adiposity, the WC vs BMI divergence, agricultural work, and the distinction between structural determinants and behavioral pathways.

Response. We revised the Discussion substantially. The agriculture paragraph now distinguishes occupational physical activity from overall energy balance and notes that the lower crude obesity estimate in agriculture disappeared after adjustment. The WC/BMI paragraph now explains why BMI and WC are not interchangeable. The gender-related implications paragraph now explicitly distinguishes structural determinants from behavioral pathways and states that unmeasured mechanisms should be interpreted as plausible explanations rather than tested mechanisms.

Comment 9. Explicitly address measurement bias due to self-reported behavioral variables.

Response. We expanded the Limitations section to discuss recall and social desirability bias, the limited validity of television viewing as a proxy for total sedentary time, and the limited sensitivity of the fruit-and-vegetable threshold as a dietary measure.

We hope that the revised manuscript addresses the reviewers’ concerns and improves the clarity, methodological transparency, and interpretability of the study.

Sincerely,

The Authors

Reviewer 3 Report

Comments and Suggestions for Authors

Review Report: Healthcare

The abstract examines occupational differences in lifestyle behaviors and adiposity among working women in Peru using pooled national survey data across ten repeated cross-sectional rounds. The authors compare multiple occupational groups on smoking, alcohol use, sedentary behavior, diet, BMI, obesity, waist circumference, and abdominal obesity to identify health inequalities.

Although the paper addresses an important public health issue and uses a large nationally representative dataset, I have the following questions and suggestions.

The study uses only crude models without adjustment for education, wealth, urban/rural residence, and age, even though these factors differ substantially across occupational groups. If the purpose is to describe disparities, that is understandable, but why did the authors not also provide adjusted models as a secondary analysis to determine whether occupational differences persist beyond socioeconomic composition?
Occupational categories appear very broad, especially groups such as services, sales, and skilled manual work. How much heterogeneity exists within these categories? For example, could very different job demands within one category dilute or distort the associations?
Why was professional/technical/managerial chosen as the sole reference group? While this is practical, why not also consider pairwise comparisons or alternative references to better understand disparities across all occupational groups?
The exclusion of non-working women is explained, but why did the authors not perform a supplementary analysis comparing working versus non-working women? This could provide useful context for occupational insertion itself.
The use of television viewing as a proxy for sedentary behavior may be limited, especially because occupational sitting time may differ greatly by job type. Why did the authors not discuss more explicitly how this measure may misclassify sedentary burden across occupations?
Fruit and vegetable intake is measured only as fewer than five servings daily, and over 90% had insufficient intake. If this variable shows minimal variation, why retain it as a central lifestyle indicator? Why not explore more sensitive dietary measures if available?
Waist circumference was available only from 2018 onward, creating a shorter analytic period for abdominal obesity. How comparable are these later waves to earlier rounds, especially given social and economic changes over time?
The authors pooled survey years and treated year categorically, but why was no interaction tested between occupation and year? Occupational health disparities may have changed over the decade.
Missing data were handled by complete-case analysis. Why was there no sensitivity analysis to assess whether excluded women differed systematically from included women, especially for anthropometric outcomes?
Since the study emphasizes women specifically, why were reproductive factors such as parity, pregnancy history, or childcare burden not discussed more deeply as possible contributors to adiposity differences?
The discussion frequently interprets occupational patterns through mechanisms such as stress, unpaid burden, or work structure, but these variables were not directly measured. How do the authors justify these interpretations without overextending beyond the data?
Agricultural workers showed healthier behavioral profiles but not proportionally better adiposity outcomes. Why did the authors not explore whether food insecurity, poverty, or undernutrition-to-obesity transition could explain this discordance?
Why was informal employment status not considered, especially in Peru where labor informality may strongly shape both occupation and health?
The study focuses on women aged 18–49 years. Why were older working women excluded, and could this age restriction limit understanding of occupational adiposity patterns across the broader female workforce?
The paper presents many outcomes separately, which is informative, but why did the authors not consider clustering or profile analysis to identify occupational lifestyle-adiposity patterns more holistically?

Suggestions

The paper would be strengthened by adding adjusted analyses alongside crude descriptive models. This would help readers distinguish whether occupational disparities are primarily occupational or largely socioeconomic.

The occupational classification should be discussed more critically, especially regarding within-group heterogeneity. A sensitivity analysis using broader or alternative classifications may improve confidence.

The discussion should be more cautious when proposing mechanisms like stress or double burden unless directly measured. These explanations are plausible, but they should be framed more clearly as hypotheses.

A temporal analysis would add value. Since the data span ten survey rounds, examining whether occupational inequalities widened or narrowed over time could substantially strengthen the manuscript.

The limitations section is already thoughtful, but more discussion of measurement limitations for sedentary behavior and diet would improve transparency.

The manuscript is generally well written, but some sections of the discussion are repetitive and could be shortened. Greater focus on the most policy-relevant findings would improve readability.

Overall, this is a relevant and potentially valuable population-level study with strong data sources.

Author Response

Reviewer 3

Comment 1. Why did the authors not also provide adjusted models as a secondary analysis?

Response. We added secondary adjusted models for behavioral and adiposity outcomes. These models adjust for age group, survey year, education, wealth quintile, area of residence, natural region, and marital status. The adjusted results are presented in Supplementary Tables S8 and S9 and summarized in the revised Results. We explicitly interpret them as conditional descriptive contrasts, not causal direct effects.

Comment 2. Occupational categories are broad and may contain within-category heterogeneity.

Response. We agree. We expanded the Limitations section to state that the standardized occupational classification does not distinguish within-category differences in physical demands, hours worked, informality, night work, job stability, job control, psychosocial stress, commuting time, or unpaid domestic burden.

Comment 3. Why was professional/technical/managerial chosen as the sole reference group? Why not pairwise comparisons or alternative references?

Response. We clarified the rationale in the Main Exposure section. The professional/technical/managerial group was chosen a priori because it represents higher-skilled nonmanual occupations, has sufficient sample size, and provides an interpretable contrast. We did not add an exhaustive pairwise comparison matrix because it would multiply contrasts and shift the manuscript away from its main descriptive objective. This rationale is now explicitly stated.

Comment 4. Why not compare working versus non-working women?

Response. We did not perform this analysis because “not working” combines substantively different situations, including unemployment, unpaid domestic work, studying, temporary incapacity, and other states. Such a contrast would address labor-force participation rather than occupational-group disparities among workers. We added this rationale to the Study Population and Eligibility Criteria section.

Comment 5. Television viewing may misclassify sedentary burden across occupations.

Response. We expanded the Limitations section to state that television viewing is a proxy for leisure-time screen exposure rather than total sedentary time and does not capture occupational sitting time. We also adjusted the language to avoid interpreting this variable as total sedentary behavior.

Comment 6. Fruit and vegetable intake shows minimal variation; why retain it and why not explore more sensitive dietary measures?

Response. We revised the Discussion to clarify that the five-servings threshold is useful for surveillance but produced a near-ceiling pattern in this sample. We therefore do not interpret it as evidence that diet is unrelated to occupation. Rather, it indicates that the available measure was too coarse to capture differences in diet quality, meal timing, ultra-processed food consumption, or food access. More sensitive dietary measures were not available consistently in the analytic file.

Comment 7. Waist circumference was available only from 2018 onward; how comparable are later waves?

Response. We expanded the Methods and Limitations sections to state that WC and abdominal obesity are subanalyses with more limited temporal coverage than BMI. The Results now also indicate that WC-based analyses used six rounds only. This limitation is explicitly acknowledged when interpreting central adiposity.

Comment 8. Why was no interaction tested between occupation and year?

Response. We added exploratory occupation-by-period interaction tests using broad periods to avoid unstable year-by-year estimates. The results are provided in Supplementary Table S12 and summarized in Section 3.5.

Comment 9. Missing data were handled by complete-case analysis; why no sensitivity analysis?

Response. We added missing-data sensitivity analyses comparing included versus excluded women for the BMI and WC analytic samples. These results are reported in Supplementary Tables S10 and S11 and summarized in Section 3.5. We did not use imputation because missing anthropometric data primarily represented measurements not obtained in the field rather than partially missing information recoverable with high credibility from covariates.

Comment 10. Why were reproductive factors such as parity, pregnancy history, or childcare burden not discussed more deeply?

Response. We verified and clarified that pregnant women were excluded from anthropometric analyses because BMI and waist circumference are not directly interpretable as adiposity indicators during pregnancy. We also expanded the Limitations section to acknowledge residual confounding by reproductive history and childcare burden, which were not incorporated into the main models.

Comment 11. The discussion frequently interprets occupational patterns through unmeasured mechanisms.

Response. We revised the Discussion to avoid overextension. The gender-related paragraph now states explicitly that job stress, working hours, childcare burden, and unpaid work were not directly measured and should be interpreted as plausible mechanisms supported by prior literature, not as mechanisms tested in this analysis.

Comment 12. Agricultural workers showed healthier behavioral profiles but not proportionally better adiposity outcomes.

Response. We revised the agriculture discussion to distinguish physical activity at work from overall energy balance and to mention poverty, food insecurity, irregular meals, limited dietary diversity, early-life deprivation, and reduced recovery as plausible explanations. The revised text also notes that the lower crude obesity estimate in agriculture disappeared after adjustment, reinforcing the need for caution.

Comment 13. Why was informal employment status not considered?

Response. A direct and harmonized measure of informal employment was not available in the analytic classification used for all rounds. We therefore did not force an unstable construct into the models. We added informality to the Limitations section and to the surveillance/public health implications as an occupational condition that future surveys should capture more precisely.

Comment 14. The study focuses on women aged 18–49 years. Why were older working women excluded?

Response. This is consistent with the ENDES/DHS women’s questionnaire and the analytic dataset. We added a limitation stating that the study does not describe occupational adiposity patterns among women older than 49 years.

Comment 15. Why not consider clustering or profile analysis to identify occupational lifestyle-adiposity patterns?

Response. We did not add clustering or latent profile analysis because it would shift the manuscript from occupational disparities to phenotype discovery and would require more granular dietary, physical activity, and occupational-condition measures than those available here. We added a future-research statement indicating that clustering or latent-profile methods could be useful when richer measures are available.

We hope that the revised manuscript addresses the reviewers’ concerns and improves the clarity, methodological transparency, and interpretability of the study.

Sincerely,

The Authors

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The authors have made substantial revisions, and I have no further concerns.

Review Reports

Reviewer 1

Reviewer 2

Reviewer 3