Associations of anthropometric adiposity indexes with hypertension risk

Abstract Background and objective: The association between hypertension and obesity has been confirmed, while no agreement has been reached about which anthropometric adiposity index is the best. This meta-analysis aimed to perform a systematic review and meta-analysis on the associations of hypertension risk with body mass index (BMI), waist circumstance (WC), waist-to-hip ratio (WHR), and waist-to-height ratio (WHtR), and a prospective urban and rural epidemiology study from China (PURE-China) was added into this meta-analysis as an individual study. Methods: Systematic literature searching was conducted to identify relevant articles published up to September 2018 in CNKI, WANFANG Data, Web of Science, SinoMed, PubMed, MEDLINE, EMBASE, Cochrane Library and cross-referencing. Literature reporting the association of hypertension risk with BMI, WC, WHR, and WHtR were defined as eligible. PURE-China data were analyzed and included as 1 eligible study into meta-analyses. Summary odds ratio (OR) and area under receiver operating characteristic curve (AUC) were pooled using meta-analysis methods. Heterogeneity and publication bias were evaluated. Subgroups based on gender, country and study design were conducted as well. Results: Thirty-eight original articles including PURE-China were included into meta-analyses, involving 309,585 subjects. WHtR had the strongest association with hypertension risk (OR, 1.68; 95% confidence interval, [CI]:1.29–2.19) and prediction ability (AUC, 70.9%; 95% CI: 67.8%–74.2%), which were also confirmed in subgroup analyses based on gender and country. However, BMI was found to have the highest prediction ability in adjusted models of PURE-China and followed WC, both of which were superior to WHtR (73.7% and 73.4% vs 73.2%). Conclusions: Our overall meta-analysis further confirmed WHtR as a good indicator at discriminating those individuals at increased risk of hypertension, and in some cases, it is better than BMI, WC, and WHR.


Introduction
Hypertension is not only a common disease itself, but also one of the main causes for risk of cerebrovascular and cardiovascular diseases, such as stroke, metabolic syndromes, and coronary artery diseases. [1][2][3][4][5] According to World Health Organization (WHO) Report in 2013, 1 billion individuals suffered from hypertension worldwide, and 9 million are deceased due to raised blood pressure annually. [6] Moderate numbers of studies provided strong evidence that hypertension contributes markedly to the global burden of diseases. [7][8][9][10][11] Although hypertension diagnosis seemed easier and cheaper than other cardiovascular diseases, no syndromes are reported by a number of people with high blood pressure. Additionally, some population is not engaged in annual physical examinations due to busy working, unlike to hospital, and self-feeling healthy and others. Therefore, the awareness, treatment, and control of hypertension are very low in some countries. [12][13][14][15][16][17][18][19][20] Thus, applying some simple anthropometric adiposity indexes (AAI) in evaluating and predicting the risk groups of hypertension is valuable. Since obesity has a strong association with hypertension, [21][22][23][24] 4 AAI are common to be used as risk evaluation indexes in many epidemiological studies, [25][26][27][28][29][30][31][32][33][34] including body mass index (BMI), waist circumstance (WC), waist-to-hip ratio (WHR) and waist-to-height ratio (WHtR), all of which can be self-measured. Two meta-analytic reviews were published in 2008 and provided more supports for centralized obesity, especially WHtR, while BMI was the poorest discriminator for detecting cardiovascular risk factors in both male and female. [35,36] Additionally, a robust association was observed among Asians compared to non-Asian populations. [36] However, Lee et al [35] only searched MEDLINE database up to 2006, and another study [36] used the original data of 19 crosssectional studies from 10 countries in the Asia-Pacific regions. A number of individual studies were reported in the last decade. [37][38][39][40][41][42][43][44] Thus, we conducted an updated systematic review and metaanalysis and summarize literature evidence of association of hypertension risk with BMI, WC, WHR, and WHtR, as well as further evaluate sex-based and country-based difference for these associations. Our data in a prospective urban and rural epidemiology study in China (PURE-China) was added into meta-analyses as an individual study.

Searching strategies
All procedures of this study followed the guidelines of the preferred reporting items for systematic reviews and metaanalyses (PRISMA) statement. [45] A systematic searching was conducted to identify the related articles in the following literature databases up to September 2018, including Cochrane Library (CENTRAL), PubMed, MEDLINE, EMBASE, Web of Science, WANFANG Data, China National Knowledge Infrastructure (CNKI), and SinoMed, and using the combinations of the following terms: ("body mass index" or "BMI") and ("waist" or "waist circumference" or "WC") and ("waist to hip ratio" or "waist-hip ratio" or "WHR" or "WHPR" or "waist; hip ratio") and ("waist to height ratio" or "waist-height ratio" or "waist: height ratio" or "waist to stature ratio" or "waist-stature ratio" or "WHtR" or "WHTR" or "WSR" or "WHeiR") and ("blood pressure" or "hypertension"). Corresponding Chinese terms with above-mentioned terms were used for searching in Chinese literature databases, such as CNKI, WANFANG Data, and SinoMed. All the bibliographical references found in target literature databases were imported into Endnote X8 for verifying eligibility checking. Each title and/or abstract was screened to evaluate its possible relevance after excluding duplicates. Fulltext articles were downloaded for further review and eligibility determination if both titles and abstracts were not enough to make decision. All article-selecting were completed by 2 researchers (Deng GJ and Liu WD) independently, the senior researcher (Yin L) made final decision when any discrepancies were shown. Personal email contacts with authors were used to obtain data when needed data were not explicitly reported or not derived from data in the articles. Cross-referencing was also conducted to improve the study identification process.

Inclusive criteria
The inclusive criteria of article selection were described as follows: (1) only original articles were considered, and editorials, comments or reviews were excluded; (2) hypertension risk was evaluated in epidemiological studies; (3) only adults were included (age≥18-year-old), but studies with older adults (age≥60-year-old) were excluded; (4) odds ratio (OR) for the associations of hypertension risk with BMI, WC, WHR, and WHtR, and/or area under receiver operating characteristic curve (AUC) for prediction abilities of hypertension risk had to be reported in 1 study. Studies with lack of any one of the indexes above-mentioned were excluded.

Data extraction
If articles were regarded as eligible, at least 2 co-authors extracted the following data independently in a standardized manner and any disagreement was discussed and resolved in our research group, including author's name, publication year, country of study, study duration, study design, recruited participants (age, number, gender, BMI, WC, WHR, and WHtR), OR, and AUC with their 95% respective confidence interval (CI) for hypertension risk related to BMI, WC, WHR, and WHtR.

Literature quality assessment
The assessment for the quality and potential bias of the included articles were executed by 2 researchers independently using forms from Agency for Healthcare Research and Quality (AHRQ), [46] which consists of 11 items scored 0 or 1. One score was counted if any item was answered "Yes", while the score was 0 when any item was answered "No" or "Unclear". The total score was calculated by adding all the scores of 11 items, and the quality level was determined as low if the total score3, medium if the score ranged from 4 to 7, and high if total score≥8.

General information of PURE-China
Details of PURE-China have been reported elsewhere. [47,48] Based on 46 Guided by 2010 Chinese guidelines of hypertension management, [49] hypertension is defined if 1 of the following 3 criteria is fulfilled: (1) taking antihypertensive drugs regularly; (2) history of hypertension diagnosis; (3) SBP≥140 mmHg and/or DBP≥90 mmHg. BMI was calculated as weight (kg) divided by height square (m 2 ), WHR computed using WC (cm) divided by HC (cm), and WHtR using WC (cm) divided by height (cm).

Statistical analyses
Stata 12.0 was used for the meta-analyses. OR and AUC with their respective 95% CI for hypertension risk with 4 AAI (BMI, WC, WHR, WHtR) was defined as effect sizes. Heterogeneity was present if P value of Q test was typically 0.10. I 2 statistic was used to evaluate the heterogeneity across all included studies. If studies were homogeneity, the pooled OR and pooled AUC were calculated by using a random effects model with DerSimonian and Laird method. If not, the fixed effect models on the Mantel-Haenszel method were applied. [50][51][52] P <.05 with 2-sided will be considered as statistical significance regarding the pooled results of all outcomes. Subgroup analyses based on gender were performed to compare potential variations among females and males. The potential publication bias was examined by constructing a "funnel plot", and the Egger linear regression test was applied to test for asymmetry of funnel plots at 0.05 level for significance. [53] In order to test for the robustness of the results, sensitivity analyses were conducted by deleting 1 study each time, which was considered as having little influence on the overall effect size if the point estimate of its "deleted" analysis always lay inside the 95% CI of the pooled statistic. Metaregressions were used to examine the impact of moderator variables (including gender and country) on study effect sizes using regression-based techniques. [54] The Statistical Analysis System (SAS 9.4 for Windows; SAS Institute Inc., Cary, NC) software was used for the statistical analyses of PURE-China. Only baseline data were used for analyses. Continuous variables were shown as the mean ± standard deviation (SD), and categorical variables as numbers (n) and percentages (%). The OR with 95% CI and AUC with 95% CI for hypertension risk in relation to BMI, WC, WHR, WHtR were computed using multivariate logistic regressions adjusted for age, sex (not for subgroup analyses by gender), education levels, alcohol use, smoking status, living location, levels of physical activities, as well as taking anti-diabetics drugs and lipidlowering drugs. Subgroup analyses stratified by gender country and study design also were conducted.

Systematic searching and article selection
The details of search strategy and included procedure were shown in Figure 1. Total of 1417 records was obtained from 8 abovementioned literature databases and cross-referencing. PURE-China data were analyzed as an individual study. 505 duplicates were excluded. 912 titles and abstracts were screened for potential eligibility, among which 575 were deleted as irrelevant records with our topic, 14 were deleted as they were conference abstracts, and 9 were deleted as they were reviews. Furthermore, full-text reviewing of 314 records was performed, of which 216 were further excluded due to the following reasons: no hypertension risk reported (n = 172), adolescent studies (n = 60), at least 1 index not reported (n = 41), only older adults included (n = 2), only those with BMI <25 included (n = 1). Finally, a total of 309,585 individuals from 38 articles were included in this meta-analysis, including our PURE-China data.

Discussions
Together with PURE-China study, 38 articles involving 309,585 participants were identified to evaluate the associations of hypertension risk with 4 AAI, including BMI, WC, WHR, and WHtR using systematic review and meta-analysis strategies. Our results further confirmed the positive associations between hypertension risk and these AAI. Among the 4 AAI, WHtR has the strongest prediction ability for hypertension risk, irrespective of the gender, though large heterogeneity and publication bias were observed across the included studies. Further sensitivity analyses and trim and fill analyses did not alter the respective prediction abilities.
Similar to previous studies, [35][36][37] significant heterogeneity among females and males was observed when discriminating hypertension risk, and higher combined AUCs were found among females than males, which indicated that the hypertension risk was estimated rather precisely in women. Furthermore, except for WC, the association of hypertension risk was stronger in men than women, although this correlation variation was not confirmed in meta-regression with respect to sex. Additionally, the difference in discrimination abilities for hypertension risk in China and other countries are notable. According to OR, WHtR is the best predictor for both Chinese population and other ethnic groups. When considering about AUC, while the best predictors are BMI and WHtR for China and non-China countries respectively. And current evidence indicated that the strength of the association between the anthropometric measures with hypertension risk is higher in other countries than China, irrespective of indexes. Central adiposity has been emphasized by a number of studies, particularly for Asian populations who may have a 'normal' BMI along with disproportionately large WC. [36,37] However, BMI showed the strongest prediction abilities in adjusted models in our PURE-China study, in either females or males, or both sexes.
Our study has specific strengths and limitations. A major strength is the application of systematic review strategies and comprehensive evaluation of the associations between adiposity measures and hypertension risk from available data, despite large heterogeneity and publication bias were observed. First, major limitations are related to limitations of the data provided by the individual studies. As a result, the risk estimation may be less accurate if individual-level data were not been available. Some studies were excluded due to no complete data used for metaanalyses, even if we contacted with authors via emails. [88][89][90][91] Second, most of studies included in our meta-analyses were observational studies, which have potential methodological limitations to detect causality between exposure and outcome. Third, 3 studies including our PURE-China were defined as outliers when assessing the stability of effect sizes of BMI and WC. Additionally, potential publication bias was detected using Egger tests, though Begg and Mazumdar rank correlation test not. Finally, although 8 databases were searched for the reviews and extensive checks for completeness by cross-referencing were employed, we cannot promise that a relevant study might be missed.

Conclusions
Despite these limitations, our systematic review and metaanalyses summarize the available studies so far and provide a comprehensive picture for the associations between hypertension risk and 4 anthropometric measures. The magnitude of these association was partly similar among Chinese and non-Chinese populations. WHtR was confirmed as a good indicator at discriminating those individuals at increased risk of hypertension.

Acknowledgments
Besides co-authors listed in this study, we would like to thank Ononamadu, CJ from Department of Biochemistry and Forensic science, Nigerian police academy, who share the data we need with us. In addition, we would like to thank those who supported  97: 48 www.md-journal.com