Predictors of Hypertension in Mauritians with Normotension and Prehypertension at Baseline: A Cohort Study

We have provided more information about the   assessment of physical activity at baseline in the methodology section. The   description reads now as follows:

Physical activity was determined by a self-administered   questionnaire following the recommendations of World Health Organization Multinational Monitoring   of Trends and Determinants of CVD (WHO MONICA) project protocol [21,23]. Both   leisure time and occupational physical activity were categorized on a   four-level scale (sedentary, light, moderate, and heavy) based on usual   patterns over the past year. Interviewers were given guidelines for   allocating activity levels for both leisure time and occupational physical   activity scales. Occupational physical activity was classified as sedentary   for office workers and the unemployed, light for shop assistants and general   housework, moderate for trade workers such as carpenters, and heavy for   building laborers and sugarcane cutters. Leisure   time physical activity was graded as sedentary for those generally housebound   with no regular outside activity; light for regular but relaxed pursuits such   as gardening and walking; moderate for active sports such as jogging,   volleyball, or cycling for >=30 min 1-2 days/week; and heavy for active   sports undertaken >+ 3 days/week.

Following reference has been added:

Dowse, G.K.; Zimmet, P.Z.; Gareeboo, H.; Alberti, G.   M.M.; Tuomilehto, J.; Finch, C.F.; Chitson, P.; Tulsidas, H. Abdominal obesity and physical inactivity as risk factors for NIDDM and   impaired glucose tolerance in Indian, Creole, and Chinese Mauritians. Diabetes Care 1991,   14(4), 271-82.

Page 3, line 122.

Our study is a cohort study following up individuals free   of hypertension at baseline with the outcome of incident hypertension. As we   do not have follow-up time of the study participants, we were not able to use   Cox regression models or present survival curves. All stud participants were   re-examined after 5- and 11 –years at the same time point. Thus, if they had   hypertension at the follow-up examination, we would not know at which time   point they developed hypertension. Using the same follow-up time (5 yrs,   respectively 11 yrs) for survival time would inaccurate survival times for   people who developed hypertension. Furthermore, follow-up times will be the   same and lead to similar estimated as odds ratios. Therefore, we decided to   use logistics regression models, a perfectly valid statistical method to   calculate associations if survival times are not available.

We have addressed this in the limitations of the study   section adding the following text:

“Moreover,   we were not able to analyze the data using survival analysis as information   on the outcomes were only assessed at the 5-year and 11-year examination   visits, and individual times-to-event information was not available for the   study participants who developed the outcome event.”

Page 12, line 298

Our study is a cohort study following up individuals free   of hypertension at baseline with the outcome of incident hypertension. As the   follow-up time of every study participant was identical, Cox regression analysis   (which takes into account variable follow-up times) will not be appropriate   method.. All study participants were re-examined after 5- and 11 –years at   the same time point. Thus, if they had hypertension at the follow-up   examination, we would not know at which time point they developed hypertension.   Using the same follow-up time (5 yrs, respectively 11 yrs) for Kaplan-Meier survival   time would give inaccurate data on survival times for people who developed   hypertension. Furthermore, follow-up times will be the same for each   individual and lead to similar estimated as odds ratios. Therefore, we   decided to use logistics regression models, a perfectly valid statistical   method to calculate associations if survival times are not available.

We have addressed this in the limitations of the study section   adding the following text:

“Moreover, we were not able to analyze the data using   survival analysis as information on the individual times-to-event was not   available for the study participants who developed the outcome event.”

We   did not find any evidence of bias or interaction by participants’ sex.   Therefore, we present data for sexes combined to increase the power of the   study. We have added the following sentence to the statistical methods   section:

“As no evidence of bias or interaction was found by   participants’ sex, the data were analyzed combining men and women.”

We   agree that another research question may be to assess the dose-response   effect of both systolic and diastolic blood pressure on future risk of   hypertension. As for clinical decision making in primary health-care the use   of hypertension categories are more common and recommended by international   guidelines, we prefer to analyze the associations in this manuscript   according to BP categories. However, we will address the issue of   dose-response in a future work of these data. Also, it is not possible to   extend the length of this manuscript and to address this important topic in   depth which would require several tables/figures and much more text for   Introduction, methods, results and discussion sections.

Page 12, line 298

Page 5, line 171

Reviewer   #3

It   is not clear whether some individuals in the pre-hypertensive group were   treated or not , please clarify

Participants with prehypertension at baseline did not   receive antihypertensive treatment. Any person treated with antihypertensive   drugs at baseline were considered as hypertensive, and excluded from the   cohort as indicated in the methods section.

We have added following sentences to the methodology   section:

A systolic blood pressure of 120 to 139 mmHg or a   diastolic blood pressure of between 80 and 89 mmHg and free-of anti-hypertensive medication was considered as   prehypertension.

Page 3, line 104

Is the   presence of diabetes a predictor of hypertension in pre-hypertensive   individuals? or after adjustment the statistical significance of increased   risk is  lost ?

The association between diabetes mellitus and future   hypertension was not statistically significant after adjustments for the   covariates (Table 3). Therefore, it was eliminated from the model. Only   covariates that were significantly associated with the outcome and improved   the model remained in the adjusted logistic regression model. The significance   of diabetes was lost in the multivariable model due to fact that some other   factors were confounders between diabetes at baseline and hypertension   incidence at follow-up.