Review Reports
- Alexander Thomas Mboma 1,2,
- Saulos Kondwani Nyirenda 3 and
- Tuyen Van Duong 1,*
- et al.
Reviewer 1: Anonymous Reviewer 2: Anonymous Reviewer 3: Anonymous
Round 1
Reviewer 1 Report
Comments and Suggestions for Authors1. Please justify why hypertension-specific anthropometric cut-offs are needed in a population where blood pressure measurement is already part of the routine diabetes clinic visit, since this is central to the clinical relevance of the work.
2. State which FDR method was used and how the analysis group was defined
3. Please shorten the Conclusions to correspond directly to the three stated aims - sex-stratified associations, comparative discrimination, and empirically derived sex-specific cut-offs - remove the duplicated statement, and confine the forward-looking content to a single sentence on the need for external validation and prospective evaluation of incident hypertension.
4. The study design is described inconsistently. The Methods correctly identify a cross-sectional design (line 99), and the limitations acknowledge that temporal sequence cannot be established (lines 549–550), but the Abstract (line 25) and the Conclusions (line 572) refer to a "multi-center cohort". Since the paper reports odds ratios per 1-SD increase and proposes screening cut-offs, the term "cohort" risks being read as evidence of incident risk.
Author Response
Sex-Stratified Associations and Discriminatory Performance of Anthropometric and Body-Shape Indices for Identifying Hypertension in Adults with Type 2 Diabetes in Malawi: A Cross-Sectional Study
|
Response to Reviewer 1 Comments |
||
|
1. Summary |
|
|
|
We thank the three reviewers for their careful and constructive assessments. All the changes are shown in tracked changes in the revised manuscript. Review comments are reproduced and our detailed responses follow, with the location of each change indicated in red font. Two reviewer requests required analyses that were not part of the original submission (Reviewer 2 Comment 5 and Reviewer 3, IPAQ-SF interaction). However, both have been conducted and results reported in the supplementary tables (S8-S10). |
||
|
3. Point-by-point response to Comments and Suggestions for Authors |
||
|
Comment 1: Please justify why hypertension-specific anthropometric cut-offs are needed in a population where blood pressure measurement is already part of the routine diabetes clinic visit, since this is central to the clinical relevance of the work. |
||
|
Response 1: We agree this was a weak point in the original framing. A new paragraph has been added to the Introduction setting out three specific gaps that anthropometric screening addresses in this setting, none of which involves replacing blood pressure measurement: 1. Classification here rests on a single clinic reading taken under routine conditions, which misclassifies patients in both directions; anthropometric status is stable between visits and can flag the patients for whom a repeat or confirmatory measurement is most worthwhile. 2. Where staff time and consumables limit how thoroughly blood pressure can be assessed, an anthropometric threshold provides a triage rule requiring no equipment beyond a tape measure. 3. Unlike blood pressure, adiposity is itself the target of lifestyle counselling, so an index identifying hypertension-relevant adiposity also identifies the patients most likely to benefit from that counselling. The paragraph closes by noting that establishing which index performs best, and at what threshold, is a prerequisite for any of these uses. The corresponding claim in the Discussion has been narrowed to emphasize the adjuncts for prioritizing patients within an existing hypertension monitoring pathway.
Introduction, new paragraph before the aims paragraph; Discussion 4.4. |
||
|
Comment 2: State which FDR method was used and how the analysis group was defined. |
||
|
Response 2: The Statistical Analysis section now specifies the Benjamini–Hochberg procedure at q = 0.05, and defines the family explicitly: the 21 pairwise comparisons among the seven measures constitute one family, and the procedure was applied separately within each of the three analysis groups (overall, men, women), giving three families of 21 tests.
All pairwise comparisons have been re-run with an independent DeLong implementation and Benjamini–Hochberg correction (revised Supplementary Table S6). We also checked whether the family definition matters: correcting across all three groups pooled gives exactly the same set of significant comparisons as correcting within each group, so the choice does not affect any conclusion.
Methods 2.4.4. |
||
|
Comment 3: Please shorten the Conclusions to correspond directly to the three stated aims - sex-stratified associations, comparative discrimination, and empirically derived sex-specific cut-offs - remove the duplicated statement, and confine the forward-looking content to a single sentence on the need for external validation and prospective evaluation of incident hypertension. |
||
|
Response 3: The Conclusions have been rewritten. The original contained the same statement about WHtR, BRI, and WC twice (sentences 2 and 4); this duplication has been removed. The revised section now maps directly onto the three aims in order: associations, comparative discrimination, sex-specific cut-offs and ends with a single forward-looking sentence on external validation and prospective evaluation of incident hypertension.
Section 5 |
||
|
Comments 4: The study design is described inconsistently. The Methods correctly identify a cross-sectional design (line 99), and the limitations acknowledge that temporal sequence cannot be established (lines 549–550), but the Abstract (line 25) and the Conclusions (line 572) refer to a "multi-center cohort". Since the paper reports odds ratios per 1-SD increase and proposes screening cut-offs, the term "cohort" risks being read as evidence of incident risk |
||
|
Response 4: The wording has been corrected throughout. The term ‘Multi-center cohort’ has been removed from the Abstract and the Conclusions; ‘this cohort’ in Discussion 4.1 is now ‘this sample’; ‘large multicenter diabetic cohort’ in the Strengths paragraph is now ‘large multicenter clinical sample’; and the NHANES-derived comparison in 4.5 now reads ‘population’ rather than ‘cohort’. The word ‘cohort’ no longer appears in the body of the manuscript except where it refers to cited studies that were genuinely cohorts.
Title; Abstract; Discussion 4.1, 4.5, 4.6; Section 5; keywords |
||
|
|
||
Reviewer 2 Report
Comments and Suggestions for AuthorsThe manuscript addresses a relevant topic and provides useful data from an underrepresented population. The multicenter design, large sample size, and extensive sex-stratified analyses are important strengths. However, some aspects of the interpretation should be revised, particularly the emphasis on sex-specific differences despite non-significant interaction tests and the modest discriminatory performance of the anthropometric indices. The manuscript would also benefit from a more cautious discussion of the clinical applicability of the proposed cut-offs and of the limitations related to routine blood pressure measurement. Please find attached my comments.
Comments for author File:
Comments.pdf
Well written and understandable, but some sentences would benefit from minor language editing to improve clarity, flow, and consistency of terminology and correct a few stylistic and grammatical issues.
Author Response
|
Response to Reviewer 2 Comments |
|
We thank the three reviewers for their careful and constructive assessments. All the changes are shown in tracked changes in the revised manuscript. Review comments are reproduced and our detailed responses follow, with the location of each change indicated in red font. Two reviewer requests required analyses that were not part of the original submission (Reviewer 2 Comment 5 and Reviewer 3, IPAQ-SF interaction). However, both have been conducted and results reported in the supplementary tables (S8-S10). |
|
Comment 1: The manuscript addresses a relevant topic and provides useful data from an underrepresented population. The multicenter design, large sample size, and extensive sex-stratified analyses are important strengths. |
|
Response 1: Thank you for your generous comment. |
|
Comment 2: The sex-specific message should be interpreted more cautiously. The manuscript is strongly framed around sex-stratified associations and sex-specific cut-offs, but the formal interaction analyses were not significant for any of the anthropometric measures. Different estimates in men and women do not necessarily indicate true effect modification by sex. The authors should distinguish more clearly between descriptive sex-stratified findings and evidence of a genuine sex interaction, especially in the Abstract and Discussion. |
|
Response 2: We agree with the reviewer. Consequently, the following revisions have been made to the write up: 1. Abstract. The sentence claiming different leading measures in men and women has been replaced with: ‘The ranking of measures was similar in men and women, and no measure showed a statistically significant sex interaction.’ The Methods sentence now states that effect modification was tested formally using interaction terms, and the Conclusions state that the sex-stratified estimates are exploratory. 2. Results 3.2. The hedged phrase ‘did not show consistent evidence of effect modification’ has been replaced with a direct statement that no statistically significant sex interaction was detected for any of the seven measures, and that the stratified estimates are descriptive subgroup findings. 3. Discussion 4.5. The paragraph now states explicitly why the apparent differences should not be over-read: the interaction tests were null; the male and female AUC confidence intervals overlap substantially (WHtR 0.65-0.74 in men vs 0.63-0.71 in women); and the non-significant ABSI estimate in men (aOR 1.19, 95% CI 0.97-1.46) is not statistically distinguishable from the significant estimate in women (aOR 1.27, 95% CI 1.08–1.49) - a difference in significance is not a significant difference. The reduced precision of the smaller male stratum (n = 494) is also noted. 4. Introduction. The directional hypothesis about sex-specific cut-offs has also been reworded as an exploratory question.
The interaction tests have been re-run and Supplementary Table S3 revised to report the ratio of odds ratios with both Wald and likelihood ratio p-values. The smallest p-value is 0.26 (BMI); for WHtR, the measure the manuscript emphasises, the ratio of odds ratios is 0.98 (95% CI 0.73-1.32, p = 0.91). The Results text now cites the range directly (all p for interaction ≥ 0.26) rather than the weaker previous presentation.
Abstract; Introduction; Results 3.2; Discussion 4.5. |
|
Comment 3: The clinical meaning of the ROC findings deserves a more balanced discussion. WHtR, BRI, and waist circumference performed better than some of the other measures, but the absolute discrimination remains modest, with AUC values around 0.67–0.69. These parameters may help with risk stratification, but they should not be presented as strong stand-alone tools for identifying hypertension. |
|
Response 3: We agree with the reviewer. Discussion 4.4 now states the full observed range (AUC 0.58-0.69 across all seven measures and all three strata), notes that no measure reached the conventional 0.70 threshold for acceptable discrimination in any stratum, and concludes that these measures are adjuncts for prioritizing patients within an existing blood pressure monitoring pathway rather than stand-alone tools for identifying hypertension. The Abstract Conclusions now explicitly state that ‘none is adequate as a stand-alone screening test.’
Abstract; Discussion 4.4. |
|
Comment 4: The added value of BRI over simpler measures is not entirely clear. BRI and WHtR are almost perfectly correlated and show essentially the same ROC performance. This should be discussed more explicitly, because from a clinical perspective the simpler measure may be more attractive if no clear gain is demonstrated with the more complex index. |
|
Response 4: Thank you for raising this point. The relationship between BRI and WHtR is not an empirical near-correlation but an algebraic identity: since WC(m)/(2π) ÷ (0.5 × Height(m)) reduces to WHtR/π, BRI is a strictly increasing function of WHtR alone. The two measures therefore impose the same ordering on participants and must return identical AUC, sensitivity, specificity and Youden index values, with cut-offs that are exact transformations of one another. Actually, this is confirmed in our own data: applying the BRI formula to the Youden-derived WHtR cut-offs of 0.601, 0.529 and 0.605 returns 5.451, 3.919 and 5.541, matching the reported BRI cut-offs of 5.459, 3.917 and 5.538 to within rounding error. We have therefore added: (i) Methods paragraph (2.4.2) stating both algebraic relations and noting that they were taken into account when interpreting the correlation matrix, the pairwise AUC comparisons, and the incremental-value analysis; (ii) Results statement (3.3) that the identical WHtR and BRI values are a property of construction, and a note in the DeLong paragraph that the WHtR-versus-BRI comparison is uninformative because the two generate the same ROC curve; (iii) Discussion passage (4.4) concluding that because BRI requires a non-linear formula and a calculator while offering no discriminatory gain over a ratio clinic staff can compute mentally, WHtR is the measure of choice, with BRI reported only for comparability with the existing literature. We add that the seven measures examined here represent fewer than seven distinct constructs, and that the proliferation of novel indices should not be mistaken for additional information.
Supplementary Table S6 has been regenerated with the WHtR-versus-BRI cell marked ‘not applicable’ rather than tested. In the male stratum the DeLong variance for that pair is exactly zero, which is the numerical signature of the identity; in the overall and female strata it is non-zero only by floating-point noise (ΔAUC = 0.0000).
Methods 2.4.2; Results 3.3; Discussion 4.4; Section 5. |
|
Comment 5: Hypertension ascertainment is an important limitation. Blood pressure was derived from routine clinic measurements, without systematic documentation of repeated readings, cuff size, rest period, or averaging. Since hypertension is the primary outcome, the possible impact of measurement variability and misclassification should be given slightly more emphasis, particularly when discussing the ROC-derived thresholds. |
|
Response 5: We agree. Discussion 4.4 now links the two directly: hypertension status rested on a single routine clinic reading combined with current medication use, and any resulting misclassification would be expected to attenuate the observed AUCs and to widen the uncertainty around the position of the ROC-derived cut-offs. A further limitation has been added noting that antihypertensive treatment lowers measured blood pressure, so treated participants with controlled pressure were classified as hypertensive on the basis of medication use alone; appropriate for estimating prevalence, but weakening the link between adiposity and the measured pressure that the ROC analysis exploits.
Discussion 4.4; Discussion 4.6. |
|
Comment 6: The covariate selection strategy needs a clearer explanation. The manuscript describes covariates as “pre-specified,” but also indicates that they were selected on the basis of bivariate associations at p<0.20 and biological plausibility. In addition, different covariates were used in the overall, male, and female models. The authors should clarify this process and explain the rationale for using different adjustment sets across strata. |
|
Response 6: We agree, and the reviewer has identified a genuine inconsistency. The wording ‘pre-specified’ has been removed and replaced with an accurate description: candidate covariates were specified a priori from the literature, and those associated with hypertension at p < 0.20 in the corresponding bivariate analysis were retained, which is why the adjustment sets differ across strata. We have also added the explanation for excluding BMI category from the covariate set (BMI is itself one of the exposures under evaluation).
To remove any suggestion that the ranking of measures is an artefact of differential adjustment, a Methods paragraph has been added describing a sensitivity analysis in which all models are refitted using a single common adjustment set across all three strata (age category, marital status, education, household income, glycemic control, night eating, physical activity).
Methods 2.4.4. The sensitivity analysis has been run and is reported in new Supplementary Table S8 and in Results 3.2. Using a single adjustment set across all three strata, no adjusted odds ratio changed by more than 0.01 and the ranking of measures was identical in the overall sample and in both sexes. For example, WHtR was 1.79 (95% CI 1.55-2.07) under both schemes overall, 1.84 (1.44-2.35) in men, and 1.78 (1.48-2.13) in women. Differential adjustment therefore explains none of the reported pattern. We are confident, this is what the reviewer was seeking. |
|
Comment 7: The proposed cut-offs should remain clearly exploratory. They were internally derived in a cross-sectional cohort from tertiary diabetes clinics and changed somewhat according to the ROC method used. These values are potentially useful, but they should not be interpreted as clinically validated thresholds until external validation is available. |
|
Response 7: We agree. Discussion 4.4 now states that the cut-offs remain exploratory, that they were derived internally in a cross-sectional sample from three urban tertiary clinics, that they shifted when the closest-to-(0,1) criterion replaced the Youden index (Supplementary Table S7), and that Youden-derived thresholds are known to be unstable at this sample size. The Abstract Conclusions and Section 5 carry the same qualification.
Abstract; Discussion 4.4; Section 5. |
|
Comment 8: Please standardize the terminology used for the anthropometric measures. Terms such as “novel body-shape indices,” “body-shape parameters,” “central adiposity measures,” and “anthropometric measures” are sometimes used interchangeably. A more consistent terminology would make the manuscript easier to follow. |
|
Response 8: A defining sentence has been added to Methods 2.4.2 fixing three terms and stating that no other collective term is used: ‘traditional anthropometric parameters’ (BMI, WC, WHtR), ‘novel body-shape indices’ (WWI, ABSI, conicity index, BRI), and ‘anthropometric measures’ for all seven collectively. Instances of ‘body-shape parameters’, ‘central adiposity measures’, and ‘waist-derived parameters’ have been brought into line. We have also resolved a related collision where ‘CI’ was used for both conicity index (Table 1 footnote, both figure legends, and several passages of text) and confidence interval (Tables 2 and 3, and throughout the Results). The conicity index is now spelled out everywhere and the abbreviation ‘CI’ is reserved for confidence interval.
Abstract; Introduction; Methods 2.4.2, 2.4.4; Results; Discussion 4.1; Table 1 footnote; Figure 1 and Figure 2 legends. |
|
Comment 9: The rationale for including some covariates, particularly night eating and physical activity, could be stated more clearly. A brief explanation of their role as potential confounders would improve the transparency of the analytical approach. |
|
Response 9: A sentence has been added stating that physical activity was retained because it is associated with both adiposity and blood pressure and is therefore a potential confounder of the association under study, rather than a variable of primary interest. The existing rationale for night eating (circadian eating patterns and cardiometabolic regulation) has been retained.
Methods 2.4.1. |
|
Comment 9: A final language and formatting revision is recommended. There are several minor typographical and stylistic issues throughout the manuscript that do not affect the scientific content but should be corrected. |
|
Response 9: Corrections made to the entire manuscript. |
Reviewer 3 Report
Comments and Suggestions for AuthorsThe manuscript is based on a solid sample (n=1,356, three centers, covering all administrative regions of Malawi) and is methodologically careful. However, the scientific novelty is weakly articulated, and one of the potentially most interesting instruments used in the study (IPAQ-SF) is applied only in a fragmentary way.
Main concern: comparing traditional anthropometric parameters (BMI, WC, WHtR) with "novel" body-shape indices (WWI, ABSI, CI, BRI) in the context of hypertension in T2DM is a design that has already been replicated many times.
The most interesting finding of the study — a sex-based inversion of the cut-off pattern opposite to what would be expected, contradicting most of the non-African literature — is not sufficiently foregrounded.
Few substantive concerns with the methodology, but a few important caveats need to be either addressed or explicitly stated. Single blood pressure measurement and the use of JNC 7 (2003) as the hypertension definition — an outdated standard compared to ACC/AHA 2017 or ESC/ESH 2023. The choice needs to be explicitly justified.
Title - Too long and overloaded with double enumerations.
Physical Activity Test (IPAQ-SF). The authors use this test only as a categorical covariate. An additional analysis could be conducted that would constitute a genuinely new contribution to the study. An interaction test could be performed — i.e., adding a test for the interaction between IPAQ-SF results and anthropometric indices.
The manuscript is methodologically sound and suitable for publication.
Author Response
|
Response to Reviewer 3 Comments |
|
We thank the three reviewers for their careful and constructive assessments. All the changes are shown in tracked changes in the revised manuscript. Review comments are reproduced and our detailed responses follow, with the location of each change indicated in red font. Two reviewer requests required analyses that were not part of the original submission (Reviewer 2 Comment 5 and Reviewer 3, IPAQ-SF interaction). However, both have been conducted and results reported in the supplementary tables (S8-S10). |
|
Comment 1: Comparing traditional anthropometric parameters (BMI, WC, WHtR) with "novel" body-shape indices (WWI, ABSI, CI, BRI) in the context of hypertension in T2DM is a design that has already been replicated many times. |
|
Response 1: We acknowledge the critic. The Introduction now states what is specific to this study rather than to the design class: comparative evaluations of this kind are numerous in Asian and European populations but scarce in sub-Saharan Africa, and we are not aware of a previous study evaluating all seven measures in an African T2DM clinic population with empirically derived sex-specific thresholds and formal pairwise comparison of discriminatory performance.
We would add that the response to Reviewer 2’s third comment has produced a second contribution that was implicit in the original data but not stated: the seven measures represent fewer than seven distinct constructs, since BRI is an algebraic transformation of WHtR and the conicity index is a scaled version of WWI. That the ‘novel index’ literature routinely reports these as independent competitors is, in our view, a finding worth stating plainly.
Introduction; Methods 2.4.2; Discussion 4.4. |
|
Comment 2: The most interesting finding of the study; a sex-based inversion of the cut-off pattern opposite to what would be expected, contradicting most of the non-African literature, is not sufficiently foregrounded. |
|
Response 2: We have foregrounded this observation while respecting Reviewer 2’s request for restraint on sex-specific claims, which pulls in the opposite direction. Discussion 4.5 now identifies the inversion as the least anticipated observation in the study and the one most clearly at odds with the non-African literature, while stating that it rests on cut-offs estimated separately within each sex rather than on a demonstrated sex interaction. It is presented as a hypothesis for testing in larger and preferably prospective African cohorts, not as a basis for adopting different thresholds in practice. We hope this satisfies both reviewers: the finding is named and given prominence, but the strength of the claim is matched to the evidence supporting it.
Discussion 4.5 |
|
Comment 3: Single blood pressure measurement and the use of JNC 7 (2003) as the hypertension definition; an outdated standard compared to ACC/AHA 2017 or ESC/ESH 2023. The choice needs to be explicitly justified. |
|
Response 3: A justification has been added to Methods 2.4.3, giving three reasons for retaining ≥140/90 mmHg rather than the 2017 ACC/AHA threshold of ≥130/80 mmHg: it remains the diagnostic threshold in the 2021 WHO guideline, the 2023 ESH guideline, and Malawian national NCD guidance, and is therefore the threshold under which the participating clinics operate; classification rested on a single unrepeated clinic reading, for which a lower threshold would substantially increase false-positive classification; and it preserves comparability with the African diabetes literature against which our prevalence estimate is benchmarked. We also corrected a misplaced citation: reference [30] (JNC 7) was previously attached to the expanded record-based sensitivity definition rather than to the ≥140/90 mmHg threshold it actually supports. It now sits with the primary definition.
Methods 2.4.3. We have also run the sensitivity analysis the comment implies, and report it in new Supplementary Table S10 and in Results 3.3. Under the ≥130/80 mmHg threshold, 1,120 participants (82.6%) are classified as hypertensive. Adjusted associations and AUC values are essentially unchanged (WHtR aOR 1.81, 95% CI 1.51-2.17, AUC 0.694, versus 1.79, 95% CI 1.55-2.07, AUC 0.688 under the primary definition), and the ranking of measures is identical.
This analysis also strengthens the reviewer’s earlier point about the sex inversion: the Youden-derived WHtR cut-off remains lower in men (0.529) than in women (0.605) under the ACC/AHA definition, so the inversion is not an artefact of the blood pressure threshold chosen. This has been added to Section 4.5 |
|
Comment 5: Title - Too long and overloaded with double enumerations. |
|
Response 5: The title has been revised to: Sex-Stratified Associations and Discriminatory Performance of Anthropometric and Body-Shape Indices for Identifying Hypertension in Adults with Type 2 Diabetes in Malawi: A Cross-Sectional Study
Title |
|
Comment 6: Physical Activity Test (IPAQ-SF). The authors use this test only as a categorical covariate. An additional analysis could be conducted that would constitute a genuinely new contribution to the study. An interaction test could be performed i.e., adding a test for the interaction between IPAQ-SF results and anthropometric indices. |
|
Response 6: We agree this is a worthwhile addition and have added the corresponding Methods text: product terms between meeting the WHO physical activity recommendation and each standardized anthropometric measure, tested in the same way as the sex interactions. The analysis has been run and is reported in new Supplementary Table S9, with results in Section 3.2 and interpretation in Section 4.4. No interaction reached statistical significance for any of the seven measures; the smallest p-value was 0.065, for waist circumference. Adjusted associations were numerically weaker among participants not meeting the WHO recommendation than among those meeting it; for WHtR, aOR 1.49 (95% CI 1.00-2.22) versus 1.87 (95% CI 1.60-2.18), but only 190 participants fall into the inactive group, and the interaction tests do not support a genuine difference. We report this null result plainly rather than framing it as a positive finding.
Results Section 3.2 and 4.4
Optional further use of the IPAQ-SF data. Total physical activity could also be reported as continuous MET-min/week in Table 1 (sample median 2,079). We have not added this because it changes a published table rather than a text passage, but we agree with the reviewer’s underlying point that dichotomizing the IPAQ-SF discards most of what it measures, and are happy to consider this in future similar studies. |
|
Comment 7: The manuscript is methodologically sound and suitable for publication. |
|
Response 7: Thank you. |
Round 2
Reviewer 1 Report
Comments and Suggestions for AuthorsI have no further comments. I recommend acceptance of the manuscript in its present form.
Author Response
Sex-Stratified Associations and Discriminatory Performance of Anthropometric and Body-Shape Indices for Identifying Hypertension in Adults with Type 2 Diabetes in Malawi: A Cross-Sectional Study
|
Response to Reviewer 1 Comments |
||
|
1. Summary |
|
|
|
We thank the editor and the three reviewers for their further review of our manuscript. All revisions are shown with tracked changes in the revised manuscript. Below are our detailed point-by-point responses. |
||
|
3. Point-by-point response to Comments and Suggestions for Authors |
||
|
Comment 1: I have no further comments. I recommend acceptance of the manuscript in its present form. |
||
|
Response 1: We thank the reviewer for the positive assessment and the recommendation for acceptance. |
||
Author Response File:
Author Response.docx
Reviewer 2 Report
Comments and Suggestions for AuthorsThe revised manuscript has been substantially improved, and the authors have adequately addressed the main methodological and interpretative concerns. The additional methodological details, formal sex-interaction tests, adjusted DeLong comparisons, sensitivity analyses, and more cautious interpretation of the derived cut-offs have strengthened the study.
Before acceptance, however, the manuscript requires careful proofreading. Several duplicated, overlapping, or incomplete passages remain, particularly in the Abstract and Conclusions, with additional errors in the Methods, Discussion, and Supplementary Materials statement. Examples include “TThis was,” “wereincluded,” “interaction terms.used,” “sand none,” and the incomplete sentence beginning “howed the strongest…”. Supplementary Table S7 is also listed twice.
The authors should remove all duplicated or superseded text, correct the remaining grammatical and typographical errors, and ensure that the Abstract and Conclusions present a single coherent version. They should also maintain the cautious interpretation of the sex-stratified findings, given the absence of statistically significant sex interactions.
Following this editorial revision, the manuscript should be suitable for publication.
Comments on the Quality of English LanguageThe scientific content is understandable, but the revised manuscript contains numerous grammatical errors, duplicated passages, incomplete words, and residual editing artifacts. Careful professional proofreading and removal of all superseded text are required before publication.
Author Response
Sex-Stratified Associations and Discriminatory Performance of Anthropometric and Body-Shape Indices for Identifying Hypertension in Adults with Type 2 Diabetes in Malawi: A Cross-Sectional Study
|
Response to Reviewer 2 Comments |
|
We thank the editor and the three reviewers for their further review of our manuscript. All revisions are shown with tracked changes in the revised manuscript. Below is our detailed point-by-point responses. |
|
Comment 1: Before acceptance, the manuscript requires careful proofreading. Several duplicated, overlapping, or incomplete passages remain, particularly in the Abstract and Conclusions, with additional errors in the Methods, Discussion, and Supplementary Materials statement. (e.g., “TThis was,” “wereincluded,” “interaction terms.used,” “sand none,” the incomplete sentence beginning “howed the strongest…”). Supplementary Table S7 is also listed twice. The authors should remove all duplicated or superseded text, correct the remaining grammatical and typographical errors, and ensure that the Abstract and Conclusions present a single coherent version. |
|
Response 1: We thank the reviewer and apologize for these editing artifacts, which arose from merging tracked versions during the previous revision. All examples cited by the reviewer have been corrected, and the Abstract and Conclusions now each contain a single version. Supplementary Table S7 appears once in the Supplementary Materials statement. We then further proofread the full manuscript and made further corrections. |
|
Comment 2: They should also maintain the cautious interpretation of the sex-stratified findings, given the absence of statistically significant sex interactions. |
|
Response 2: We appreciate this comment. The sex-stratified estimates are presented as descriptive subgroup findings throughout. In this round we further revised the remaining statements that could imply sex differences: the statement in Section 4.1 that body-composition differences ‘support the need for sex-stratified analyses and caution against applying uniform thresholds’ now states only that these differences motivated presenting results by sex and that the associations did not differ significantly by sex; Section 4.5 has been retitled ‘Sex-Stratified Findings,’ and the mechanistic paragraph is now explicitly conditional on the numerical differences being confirmed in larger samples. These changes also respond to Reviewer 3, Comment 2. |
Reviewer 3 Report
Comments and Suggestions for AuthorsThe authors’ response does not fully address the comment. The question concerns not the role of physical activity as a confounder, but the method used to process the IPAQ-SF data. It is necessary to explain why the data from the validated questionnaire were reduced to a binary variable instead of using the standard IPAQ categories or MET-min/week measures, and to discuss the possible impact of this simplification on the study results.
The authors report no statistically significant interaction between gender and anthropometric measures. In light of this, please clarify the rationale for maintaining such a strong emphasis on sex-stratified analysis in the manuscript. It appears that differences between men and women are not the main finding of the study; therefore, this emphasis—including the title of the article—should be reconsidered.
Author Response
Sex-Stratified Associations and Discriminatory Performance of Anthropometric and Body-Shape Indices for Identifying Hypertension in Adults with Type 2 Diabetes in Malawi: A Cross-Sectional Study
|
Response to Reviewer 3 Comments |
|
We thank the editor and the three reviewers for their further review of our manuscript. All revisions are shown with tracked changes in the revised manuscript. Below are our detailed point-by-point responses. |
|
Comment 1: The question concerns not the role of physical activity as a confounder, but the method used to process the IPAQ-SF data. It is necessary to explain why the data from the validated questionnaire were reduced to a binary variable instead of using the standard IPAQ categories or MET-min/week measures, and to discuss the possible impact of this simplification on the study results. |
|
Response 1: We thank the reviewer for clarifying the comment. The IPAQ-SF was scored according to the standard protocol, and total MET-min/week was calculated for every participant (Section 2.4.1). We then used a binary indicator of meeting the WHO recommendation (≥600 MET-min/week) for three reasons. First, physical activity entered the analysis only as an adjustment variable, not as an exposure, so a clinically interpretable and guideline-based indicator was preferred. Second, the 600 MET-min/week threshold approximates the boundary between the IPAQ ‘low’ and ‘moderate’ categories, so the binary variable separates participants in the low category from those in the moderate or high categories. Third, self-reported IPAQ-SF volumes show only modest criterion validity against objective measures (Craig et al., 2003; reference 23), which limits the precision that continuous MET-min/week values would add.
Regarding the impact of this choice: dichotomization may leave residual confounding by activity volume. Because higher physical activity is associated with both lower adiposity and lower blood pressure, residual confounding would be expected to bias the adjusted odds ratios upward rather than downward. The simplification has no bearing on the ROC analyses, AUC values, DeLong comparisons, or cut-offs in Table 3 and Figures 1-2, which were estimated for each anthropometric measure without covariate adjustment. Because the same physical activity variable was used in every model, it is also unlikely to have changed the ranking of the anthropometric measures, which is the main comparative finding.
To address and resolve this comment fully, we have specifically added the rationale for the binary coding to Section 2.4.1, and the potential for residual confounding and its likely direction to the Limitations (Section 4.6). |
|
Comment 2: The authors report no statistically significant interaction between gender and anthropometric measures. In light of this, please clarify the rationale for maintaining such a strong emphasis on sex-stratified analysis in the manuscript. It appears that differences between men and women are not the main finding of the study; therefore, this emphasis including the title of the article should be reconsidered. |
|
Response 2: We agree that differences between men and women are not the main finding, and we have reduced the emphasis on them throughout the manuscript. However, we would propose to retain the title, for three reasons. (i) Sex stratification was a pre-specified design feature rather than a post hoc subgroup exploration: the study hypothesized that optimal cut-offs would differ by sex, and the analysis plan accordingly specified stratified estimation with formal interaction testing. The title describes what the study did, not what it found. (ii) A null interaction is a result, not an absence of one. Reporting that the same waist-based measures ranked highest in both sexes, and that the apparent cut-off differences were not statistically supported, is informative precisely because sex-specific anthropometric thresholds are widely used in practice; a title that omitted the sex-stratified design would make this finding harder to locate for readers examining that question. (iii) Removing ‘sex-stratified’ from the title while the manuscript continues to report sex-stratified estimates, sex-specific cut-offs, and interaction tests in the Abstract, Table 2, Table 3, Figure 1, Section 4.5, and Supplementary Tables S1A, S2, S3, S5-S8 and S10 would understate the scope of the analyses rather than describe them accurately. The title states only that associations and discriminatory performance were evaluated in a sex-stratified manner; it makes no claim that they differ by sex, and the Abstract states explicitly that no measure showed a significant sex interaction and that the sex-stratified estimates are exploratory. We would of course defer to the editor should a change really be preferred.
We have made the following changes to the manuscript to reduce the emphasis on sex-stratification: (i) The study aim now reads ‘the associations and comparative discriminatory performance … overall and by sex’ (ii) The heading of Section 3.2 now reads ‘…Overall and by Sex,’ and a sentence in Section 2.4.4 explains why stratification was performed and how the interaction tests were used (iii) The statement in Section 4.1 that body-composition differences support the need for sex-stratified analyses and caution against uniform thresholds has been replaced with a statement that these differences motivated presenting results by sex, and that the associations did not in fact differ significantly by sex (iv) Section 4.5 has been retitled ‘Sex-Stratified Findings,’ a sentence on the practical implication of the null interaction has been added, and the mechanistic discussion is now explicitly conditional on the numerical differences being confirmed in larger samples (v) The Strengths paragraph now refers to ‘sex-stratified analyses supported by formal tests of effect modification.’ |