Utility of Obesity Indicators for Predicting Hypertension among Older Persons in Limpopo Province, South Africa

Round 1

Reviewer 1 Report

The manuscript has improved significantly. Few minor suggestions to be considered-

i) Please include the full forms of BMI, WC, WHR, WHtR in abstract.

ii) Thanks for adding the description about socio-economic disadvantage in the discussion. However it is still unclear how this relates to your current study findings? Consider deleting if not relevant.

Author Response

Dear Reviewer,

Thank you for the comments.

The manuscript has improved significantly.

Response: Thank you

Few minor suggestions to be considered-

1. Please include the full forms of BMI, WC, WHR, WHtR in abstract.

Response: Added in lines 24-26

1. Thanks for adding the description about socio-economic disadvantage in the discussion. However, it is still unclear how this relates to your current study findings? Consider deleting if not relevant.

Response: Information deleted. We only left one summarized sentence with references on similar reports; lines 205-207. Hope this is ok?

Regards, 

Reviewer 2 Report

I have read the article entitled "Utility of Obesity Indicators for Predicting Hypertension among Older Persons in Limpopo Province, South Africa". The authors declared that WC was significantly associated with hypertension in older women, proposing WC as a screening tool for the prediction of hypertension in South African older women. Overall, this article lacks novelty and has important issues that need to be addressed or revised. I have some review comments and hope the comments will be of some use for the authors.

1. The article lacks novelty, and the conclusion is already well-known.
2. The authors seemed to pile up references, for example, in Line 34-35, the statement that the burden of hypertension has been emphasized as a public health issue is well-known, but the authors cited seven references. The same is true elsewhere, which leads to a whopping 89 references in this epidemiological study, comparable to a review.
3. The sample size calculation process was not clearly described. Cross-sectional studies have their own sample size calculation methods. According to the 10EPV rule in logistic regression, it is hard to believe that the final sample size was only 350 cases, particularly in the surveyed communities with large numbers of people. Besides, the authors declared that thay used random and systematic sampling to select participants, so, sampling weights should be considered to reflect overall population characteristics.
4. This study was a cross-sectional study, and only association/correlation could be obtained using logistic regression, but causality couldnot be obtained. Therefore, the use of "predicting", "predictors" in the paper was wrong. In addition, ROC analyses only considered the relationship between a single anthropometric index and SBP/DBP, and did not consider other covariables. Therefore, the main result should be multivariate logistic regression. However, in multivariate logistic regression, the authors did not assess the association between anthropometric indices and HTN across the entire population, but across genders. Therefore, the impact of WC on HTN cannot be extended to the whole population. Why not perform the association between anthropometric indices and HTN in the entire population?
5. In Table 4, the authors obtained the cut-off values for each anthropometric index by ROC. In Table 5, the authors converted these  anthropometric indices into categorical variables and assessed their associations with HTN. How were the boundaries of these variables (BMI≥25 kg/m2, WHR≥0.90, WHtR≥0.50 and WC≥90cm) determined in Table 5?
6. Other representation questions: in Table 2, what did "normal SBP/DBP" mean? If the participants had SBP < 90mmHg, DBP < 60mmHg or even lower, were they also included in this group? The author did not address these issues in the paper.

Author Response

Dear Reviewer, 

Thank you for the comments.

Hopefully, we have addressed them to the best of our abilities. 

I have read the article entitled "Utility of Obesity Indicators for Predicting Hypertension among Older Persons in Limpopo Province, South Africa". The authors declared that WC was significantly associated with hypertension in older women, proposing WC as a screening tool for the prediction of hypertension in South African older women. Overall, this article lacks novelty and has important issues that need to be addressed or revised. I have some review comments and hope the comments will be of some use for the authors.

1. The article lacks novelty, and the conclusion is already well-known.

Response: Thank you for the comment. When we conceptualized the study, we identified paucity of data in relation to elderly people from the rural area in South Africa. We hope this makes the study relevant and to some extent, novel, by adding knowledge in to what is already existing in South Africa.

2. The authors seemed to pile up references, for example, in Line 34-35, the statement that the burden of hypertension has been emphasized as a public health issue is well-known, but the authors cited seven references. The same is true elsewhere, which leads to a whopping 89 references in this epidemiological study, comparable to a review.

Response: We have revised the references with the aim to reduce them. We are now remaining with 75 references, from 90.

3. The sample size calculation process was not clearly described. Cross-sectional studies have their own sample size calculation methods. According to the 10EPV rule in logistic regression, it is hard to believe that the final sample size was only 350 cases, particularly in the surveyed communities with large numbers of people. Besides, the authors declared that thay used random and systematic sampling to select participants, so, sampling weights should be considered to reflect overall population characteristics.

Response: We understand this concern, reviewer, and yes, your point is valid. We have added the paragraph (lines 89 – 99): “During house visits, every 5th house was considered in the selected sections of the communities. Initially, 501 households were selected, and from each house, one elderly person participated in the study. However, at the end of data collection, 360 participated in the study, making the response rate in the study 87%, which is considered good [49]. The 501 households that were initially selected, did not included the houses that did not have elderly people, and houses that had elderly people, but were critically ill. Some houses were empty, as it is a case in most rural areas, due to most children and grandchildren of elderly people taking them to their houses in the cities, as it was explained by the neighbours, during recruitment. Most of the older Black South Africans continue to live in extended household structures with children, grandchildren, and other kin [50]. Rural areas are sparsely populated because many people leave rural areas and settles in the urban areas for more facilities. These society has homogeneity in terms of the culture and socioeconomic status [51]. A sample of 360 was obtained, however, during data analysis, 10 questionnaires had missing information of over 10%, which included important variables, such as SBP, DBP, and weight values.

4. This study was a cross-sectional study, and only association/correlation could be obtained using logistic regression, but causality couldnot be obtained. Therefore, the use of "predicting", "predictors" in the paper was wrong. In addition, ROC analyses only considered the relationship between a single anthropometric index and SBP/DBP, and did not consider other covariables. Therefore, the main result should be multivariate logistic regression. However, in multivariate logistic regression, the authors did not assess the association between anthropometric indices and HTN across the entire population, but across genders. Therefore, the impact of WC on HTN cannot be extended to the whole population. Why not perform the association between anthropometric indices and HTN in the entire population?

Response: In ROC analyses, the predictive capabilities of a variable is commonly summarized by the area under the curve (AUC), which can be found by integrating areas under the line segments (Muschelli, J., 2020. ROC and AUC with a binary predictor: a potentially misleading metric. Journal of classification, 37(3), pp.696-708; Chua, E.Y., Zalilah, M.S., Haemamalar, K., Norhasmah, S. and Geeta, A., 2017. Obesity indices predict hypertension among indigenous adults in Krau Wildlife Reserve, Peninsular Malaysia. Journal of Health, Population and Nutrition, 36(1), pp.1-7). In the study, ROC analyses were used to determine the best predictive capabilities of single anthropometric variables. The terms "predicting" and "predictors" were not used in logistic regression, but they were referred in ROC analysis. The decision to stratify the results by gender in order to assess the relationship between anthropometric indices and HTN was motivated by preliminary analysis in Table 2 and previous studies that revealed gender differences in anthropometric indices.

5. In Table 4, the authors obtained the cut-off values for each anthropometric index by ROC. In Table 5, the authors converted these anthropometric indices into categorical variables and assessed their associations with HTN. How were the boundaries of these variables (BMI≥25 kg/m2, WHR≥0.90, WHtR≥0.50 and WC≥90cm) determined in Table 5?

Response: Only BMI was significant in Table 5 for men, with a cut-off value of 27.50, which is close to the standard cut-off value. As a result, the authors decided to stick with the standard cut-off point of 25 kg/m2. The same is true for the other two significant variables, WHtR and WC for Women.

6. Other representation questions: in Table 2, what did "normal SBP/DBP" mean? If the participants had SBP < 90mmHg, DBP < 60mmHg or even lower, were they also included in this group? The author did not address these issues in the paper.

In lines; 108-110, we have indicated that the average of the last two readings was used as a participant’s blood pressure [55], and hypertension was defined at SBP ≥140mmHg and/or DBP ≥90mmHg for hypertension [25], high SBP was ≥140mmHg, and high SBP was ≥90mmHg.

Kind regards, 

Reviewer 3 Report

This is a nice written paper that covers a public health burden through modifiable factor(s) and it is true that the use of obesity indicator type remains controversial and it should be evaluated in various populations and age groups.     

Line 39: please state numbers in consistent format(% or millions)

A 90%CI, adds a large variation in the study and decreases power. please comment in discussion that results need to be duplicated at the least in order to be more conclusive. 

Overall the sample size is the mai limiting factor of this study. 350 individuals marginally reaches adequacy for such and analysis. By using further stratification he results can only be seen as preliminary findings. No actual conclusions can be made. 

This needs to be clearly stated. 

Please place SD after mean in Tables where appropriate.

Results are very nicely described.  

Author Response

Dear Reviewer,

Thank you for the comments.

This is a nice written paper that covers a public health burden through modifiable factor(s) and it is true that the use of obesity indicator type remains controversial and it should be evaluated in various populations and age groups.     

Response: Thank you

Line 39: please state numbers in consistent format(% or millions)
Response:

Response: millions format used, lines have changed to 42-43

A 90%CI, adds a large variation in the study and decreases power. please comment in discussion that results need to be duplicated at the least in order to be more conclusive. Overall the sample size is the main limiting factor of this study. 350 individuals marginally reach adequacy for such and analysis. By using further stratification, the results can only be seen as preliminary findings. No actual conclusions can be made. This needs to be clearly stated. 

Response: Thank you for this advice. We saw this advice befitting to be added in the limitations; Lines 267-270, hope this is ok? We have considered most of your wording captured clear.

Please place SD after mean in Tables where appropriate.

Response: SD added in the abstract; line 22 - mean age, lines, results; line 138, line 143 – table 1, line 155 – table 2

Results are very nicely described.  

Response: Thank you

Regards, 

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

The article presents the association between obesity indicators and hypertension in a rural setting, South Africa. There are few areas that could have been improved for further clarification. Please see details below. The article also requires some English editing and a proof reading as some spelling errors are obvious e.g. simultatneous, propolation. The authors have also moved from hypertension and cardiovascular risk here and there in the article, which need to be addressed.   

Abstract and Introduction: Abstract needs some rephrasing and rewriting. Lines 16-17, community-based study is redundant. Lines 22-23, please rephrase this sentence. e.g. Mean age "of study participants". Line 23, obesity indicator is redundant. Lines 23-25. Please specify older men and women for each result.

On several occasions throughout the introduction, the association between obesity and hypertension texts appear to be repetitive. It would be better to tidy them up for better flow of information. Lines 60-62, the statement was unclear regarding the association between hypertension and obesity. Are the authors referring high prevalence of both conditions as an association?

Methods: What is the total population of Vhembe district? Did you include all the villages? How did you select at village level? What was your estimated sample size?

Line 99- Perhaps =>25-29 applies to overweight. Please mention obesity definition separately. Please also mention adjusted variables list under data analysis. What criteria was used to select the variables in the adjusted model?

Results: It will be better to present the demographic factors by male/female. What was the proportion of male/female in the study? Percentage for receive old grant (table 1) is missing. Please check percentage of family H/O hypertension and be consistent with Yes/No order for all variables. What is my HH headship others? HH income- please indicate (South African Dollar).

Lines 125-126, Government grant - this sentence is redundant. Please provide the information once.

Line 126- please rephrase sentence about smoking, nearly 50% were smokers

Table 2: what were the mean (sd) of SBP, DBP and obesity indicators?

Table 3: WHtR - percentage dose not add up to 100?

Lines 153-156: Needs some rephrasing to be consistent with the the figure 1. What cut-offs were used for elevated SBP and DBP?

Line 173, mentioned about unadjusted logistic regression analysis. But Table 5 did not include any unadjusted analysis. Perhaps delete it. What was your outcome variable for this analysis? Logistic regression is appropriate for binary outcome? Did you asses elevated SBP vs normal SBP, or hypertension? Please clarify and update the table heading accordingly.

Table 5: Model 4, please provide correct confidence interval for aOR 4.22.

All the figures could have been improved with clear title reflecting ROC curve for men and women presenting the relationship between obesity indicators and BP.

Discussion: It was not clear how the authors refer to poor living conditions of the study participants while there were no such results presented. Please clarify multigenerational households in this study context and relevance to your study findings. 

Lines 238-239, are the authors indicating the inconsistency of results between ethnic groups in this study? However no such results were presented. Line 240, what did you mean by risk of cardiovascular outcomes? Line 247, the previous sentence says similar to WHO cut-off. What did you mean by "deviation to WHO" in this sentence?

Limitations, perhaps shall also mention about other comorbidities (e.g. diabetes).

Author Response

Dear Reviewer,

Thank you for the comments. Please find our responses.

The article presents the association between obesity indicators and hypertension in a rural setting, South Africa. There are few areas that could have been improved for further clarification. Please see details below. The article also requires some English editing and a proof reading as some spelling errors are obvious e.g. simultatneous, propolation. The authors have also moved from hypertension and cardiovascular risk here and there in the article, which need to be addressed.   

Abstract and Introduction: Abstract needs some rephrasing and rewriting. Lines 16-17, community-based study is redundant. Lines 22-23, please rephrase this sentence. e.g. Mean age "of study participants". Line 23, obesity indicator is redundant. Lines 23-25. Please specify older men and women for each result.

Responses: We have rephrased the abstract as advised. For Lines 23-25, we have added the results for the participants. Although we are not sure if we understood the comment. Analysis was done by gender only among all the participants. We did not divide the results by age, since all the participants are older persons aged 60 years and above.

On several occasions throughout the introduction, the association between obesity and hypertension texts appear to be repetitive. It would be better to tidy them up for better flow of information. Lines 60-62, the statement was unclear regarding the association between hypertension and obesity. Are the authors referring high prevalence of both conditions as an association?

Responses: Introduction has been revised; lines 33 – 69.

Methods: What is the total population of Vhembe district? Did you include all the villages? How did you select at village level? What was your estimated sample size?

Responses: Total population, selection and estimated sample size have been improved on; lines 73 – 92.

Line 99- Perhaps =>25-29 applies to overweight. Please mention obesity definition separately. Please also mention adjusted variables list under data analysis. What criteria was used to select the variables in the adjusted model?

Responses: Overweight and obesity defined (lines 104-105. Adjusted variables and cofounders addressed in lines 116 – 117.

Results: It will be better to present the demographic factors by male/female (done in lines 127 – 135). What was the proportion of male/female in the study (Added in lines 127 – 128). Percentage for receive old grant (table 1) is missing (added). Please check percentage of family H/O hypertension and be consistent with Yes/No order for all variables (Revised). What is my HH headship others (other categories have been added)? HH income- please indicate (South African Dollar) (indicated in SA Rands first, then dollars in brackets)

Lines 125-126, Government grant - this sentence is redundant. Please provide the information once. (Addressed)

Line 126- please rephrase sentence about smoking, nearly 50% were smokers (narrative of table 1 has changed when we changed the table to divide the results in men and women)

Table 2: what were the mean (sd) of SBP, DBP and obesity indicators? (mean ±SD have been added in table 2)

Table 3: WHtR - percentage dose not add up to 100? (have been corrected)

Lines 153-156: Needs some rephrasing to be consistent with the the figure 1. What cut-offs were used for elevated SBP and DBP? (lines 150 -151 and lines 156 – 159, lines 163-166 have been rephrased. Cut offs for SBP ≥140mmHg and for DBP ≥90mmHg – added under figures).

Line 173, mentioned about unadjusted logistic regression analysis. But Table 5 did not include any unadjusted analysis. Perhaps delete it (deleted). What was your outcome variable for this analysis. Logistic regression is appropriate for binary outcome? Did you asses elevated SBP vs normal SBP, or hypertension? Please clarify and update the table heading accordingly. (Responses: Hypertension was the outcome variable – lines 180 and 186)?

Table 5: Model 4, please provide correct confidence interval for aOR 4.22. Response: The odds ratio has been corrected from 4.22 to 1.22. The CI is correct.

All the figures could have been improved with clear title reflecting ROC curve for men and women presenting the relationship between obesity indicators and BP. Response: Figures have been improved.

Discussion: It was not clear how the authors refer to poor living conditions of the study participants while there were no such results presented. Please clarify multigenerational households in this study context and relevance to your study findings. (Response: Addressed; lines 196 – 201).

Lines 238-239, are the authors indicating the inconsistency of results between ethnic groups in this study? (Response: deleted. It was a mistake) However no such results were presented. Line 240, what did you mean by risk of cardiovascular outcomes? (Response: Deleted because the sentence was confusing). Line 247, the previous sentence says similar to WHO cut-off. What did you mean by "deviation to WHO" in this sentence? (Response: deleted)

Limitations, perhaps shall also mention about other comorbidities (e.g. diabetes).

Response: Added; lines 263 - 265

Thank you,

Prof Modjadji - on behalf of authors

Reviewer 2 Report

The introduction and methods are well written and understandable. 

I would suggest some modifications:

1. in line 98, it is better to use only overweight, without obesity. "overweight/obesity was defined at BMI≥25–29.9kg/m2", since obesity definition is BMI ≥ 30.
2. In the following sentence (line 108), it is important to clarify if you have used continuous or dichotomous data. " Multivariate logistic 
   regression analysis to determine the relationship of hypertension with overweight/obesity indicators by gender and adjusted odds ratios (AOR) indicated the strength of the relationship."
3. I would suggest changing the title of table 1 and sending this table to supplementary material. 
4. Table 3, I would also suggest using it as supplementary material. 
5. In the results section, please, be attentive to do not discuss the results in the following sentence. "It was observed that BMI had the highest AUC value, suggesting that BMI is likely to be a better predictor..."
6. The main results are in Tables 4 and 5. Then, I would suggest describing deeply these results. 

Author Response

Dear Reviewer,

Thank you for the comment. Please find the responses.

1. in line 98, it is better to use only overweight, without obesity. "overweight/obesity was defined at BMI≥25–29.9kg/m2", since obesity definition is BMI ≥ 30.

Response: We have corrected that. (lines 104-105)

1. In the following sentence (line 108), it is important to clarify if you have used continuous or dichotomous data. " Multivariate logistic 
   regression analysis to determine the relationship of hypertension with overweight/obesity indicators by gender and adjusted odds ratios (AOR) indicated the strength of the relationship."

Response: We used dichotomous data – added in line 115.

1. I would suggest changing the title of table 1 and sending this table to supplementary material. 
2. Table 3, I would also suggest using it as supplementary material. 
3. In the results section, please, be attentive to do not discuss the results in the following sentence. "It was observed that BMI had the highest AUC value, suggesting that BMI is likely to be a better predictor..."

Response: We have left the tables within the main paper in reconciling your comments and that of the other reviewer, who want us to improve on them. We have used track changes to delete all “discussion” in the results section.

1. The main results are in Tables 4 and 5. Then, I would suggest describing deeply these results. 

Response: The results for table 4 and 5 have been expanded. Lines 168-174, and 179 to 183

Regards,

Prof Modjadji - on behalf of authors

Round 2

Reviewer 2 Report

The manuscript has been improving. However, the results and discussion sections need to be deeply modified.