Gender-specic Predictive Ability for the Risk of Hypertension Incidence Related to Baseline Level or Trajectories of Adiposity Indices: Cohort Study of Functional Community in Beijing Urban from 2015 to 2019 year

Background: Early prevention of hypertension is important for global cardiovascular disease morbidity and mortality. The current study aims to exploit better predictor for hypertension incidence related to baseline level or trajectories of adiposity indices, as well as the gender-specic effect. Methods: 6085 subjects from a functional community cohort in urban Beijing participated in our study. Restricted cubic splines were used to estimated nonlinear associations of BMI and WHtR as continuous variable with risk of hypertension. Stepwise logistic regression model was performed to estimate the RRs of adiposity indices and metabolic status, adjusted for covariates. Nomogram models and receiver operating characteristic (ROC) curve analysis was used to evaluate the predictive power of BMI trajectory groups and WHtR trajectory groups. Further, all analysis were performed by gender. Results: The risk of hypertension incidence was signicant related to BMI trajectory groups (persistent overweight: RR=1.99, 95%CI: 1.57-2.52; persistent obese: RR=3.07, 95%CI: 2.40-3.93; persistent the highest: RR=4.55, 95%CI: 3.37-6.14) and WHtR trajectory groups (persistent medium: RR=2.90, 95%CI: 2.23-3.77; persistent high: RR=4.48, 95%CI: 3.41-5.89; increasing to higher: RR=7.77, 95%CI:5.48-11.0). In total population, BMI trajectories and WHtR trajectories showed similar ability to predict the risk of hypertension incidence with AUC 0.713 and 0.720, respectively. After stratied by gender, BMI trajectories showed higher power in female than male butsimilar in WHtR trajectories (BMI trajectories: 0.762 vs. 0.662; WHtR trajectories: 0.769 vs. 0.663). Conclusions: BMI and WHtR trajectories have higher predictive power for hypertension incidence compared to baseline data. Women are more vulnerable to obese and abnormal MS than man.


Introduction
Hypertension is one of the leading causes of premature death worldwide that signi cantly increases the risk of stroke, heart attract, kidney and blindness (1)(2)(3). According to the statistics of World Health Organization (WHO), 1.13 billion people were diagnosed as hypertension but less than 1 in 5 has it under control. Hypertension prevalence was reported lower in high-income countries (35%) than other groups at 40%. In China, an increased hypertension incidence has been reported in adult from 14.5% in 1991 to 34.0% in 2012 (4)(5)(6). Clinical and epidemiological researches showed that early prevention of hypertension is important for global cardiovascular disease morbidity and mortality (7).
It's known that obese is a major risk factor for hypertension (8)(9)(10)(11)(12). Weight gain contributes to the risk of hypertension (13)(14)(15)(16)(17), whereas weight loss decreases the risk of hypertension (18)(19)(20). Although BMI is widely used in association studies between adiposity and hypertension (11,14), it is suggested that BMI may result in misclassi cation on an individual level without considering body-fat distribution. Indeed, it is important to consider the varying contributions of bone mass, muscle mass and uid to body weight (21). Compared with BMI, waist-to-height ratio (WHtR) was known as a better predictor of cardiovascular disease risk factors (22,23). WHtR also has the advantage of being age-and sex-independent, and it is easier to use (24). It is con rmed that metabolic health is a transient phenotype. Subgroups of obese individuals with metabolic abnormalities or not may have different risk of cardiovascular disease (CVD) (25,26). On the other hand, metabolically unhealthy individuals are at high risk of CVD irrespective of BMI (27).While the association between obesity and the metabolic status (MS) has been widely studied, it is still proposed merely a dose-response relationship (27)(28)(29). Moreover, the other factors of MS except for blood pressure also showed dramatic correlations with hypertension (30,31). Thus, these problems should be addressed.
The value of BMI or WHtR is not always constant, and the timing addition and reduction of BMI or WHtR may uctuate over time, so that different individuals often have different trajectories. Due to different trajectories may also have different cause, identi cation of different trajectories effects on hypertension would enable better prevention of highly burdening problem (32). Previous studies were more likely to focus on the risk of hypertension incidence related to baseline value of BMI or WHtR, thus, limited knowledge was obtained on the associations between hypertension incidence and BMI or WHtR trajectories.
With the prevalence of overweight, obesity, and abnormal obesity sharply increasing in current China (33,34), it is urgent that accurate prevention strategies for cardiovascular diseases should be advanced. Therefore, it is necessary to establish an accurate prediction model of hypertension. In this study, we aim to (1) investigate the risk of hypertension incidence related to BMI, WHtR, MS and the combination effect of BMI and MS or WHtR and MS, as well as BMI trajectories and WHtR trajectories; (2) exploit better predictor for hypertension incidence among those indexes; (3) gender-speci c effect of adiposity indices and MS for hypertension incidence.

Study population
The subjects were derived from a functional community cohort,which was established in 2010 in urban Beijing. All the participants aged 25 to 65 were followed up each year at the Health Management Center of Xuanwu Hospital, Capital Medical University. The cohort was composed of nearly 8671 occupational induvial including employees from governments, schools, hospitals, factories, business, and service institutions in Xicheng district, representing most of the occupational population in urban Beijing. The detail of inclusion and exclusion criteria has been described in previous study (35). Study protocols were approved by Xuanwu Hospital and Capital Medical University, and all participants were informed at enrollment.
In this study, 7482 individuals aged from 30 to 70 years participating in 2015 year were chosen for our analysis. Among those, individuals were excluded if they were missing simultaneously during follow up from 2016 to 2019 [n = 207, (BMI: n = 54; WHtR: n = 87; Blood pressure: n = 66)] and diagnosed hypertension (n = 1190), leaving 6085 subjects for our analysis (Fig. 1).

Data Collection And Clinical Measurements
Data on socio-demographic information such as age, gender, occupation and nationality were collected by standard questionnaires.

Assessment Of Adiposity Indices And Metabolic Status
BMI and WHtR were obtained to estimate adiposity in this study. Weight, height, and WC were collected in light clothing and barefoot, using standard measurement by trained healthcare workers. Height and body weight were accurate to 0.5 cm and 0.1 kg, respectively. With the subjects standing naturally upright, waist circumference (WC) was measured halfway between the lower costal border and iliac crest, while hip circumference was measured horizontally circling the front of public symphysis and highest point of large gluteus maximus. BMI was calculated by dividing weight in kilograms by the square of the height in meters (kg/m2). WHtR was calculated as the WC divided by height. BMI were further categorized at the

Assessment Of Hypertension
Three blood pressure (BP) measurements were taken at 5-min interval in sitting position. The average of the second and third measurements was used in the analysis. Hypertension was de ned as systolic blood pressure (SBP)/diastolic blood pressure (DBP) ≥ 140/90 mm Hg, the use of hypertensive medications, or a self-reported diagnosis. Participants were considered as hypertension at whichever waves during follow-up.

Statistical Analysis
Chi-square, t test and analysis of variance (ANOVA) were employed to examine characteristics of the study population by baseline levels of BMI_MS or WHtR_MS, as well as hypertension incidence. The associations of BMI and WHtR as continuous variable with risk of hypertension were estimated with the use of restricted cubic splines. The Akaike information criteria (AIC) was used to choose the optimal knots (n = 4).
Group-based trajectory model (GBTM) was employed to describe the trajectories of BMI and WHtR from 2015 to 2019. To get the optimal trajectory model, we considered the number of trajectory groups and shape of polynomial curves as described in previous literatures (38). Firstly, the shape of one group was exploited in order of cubic, quadratic and linear, until the components were signi cant at P < 0.05 did we add another trajectory group. Then, we add another group with the similar procedure. The Bayesian Information Criterion (BIC) index was employed to select the best number of groups for tting model in this study. Univariate logistic regression was used to estimate crude risk ratio (RR) of the hypertension incidence in relation to BMI and WHtR trajectory groups.
Stepwise logistic regression model was performed to estimate the RR of hypertension incidence in relation to BMI, WHtR, MS, BMI_MS, WHtR_MS, BMI trajectory groups and WHtR trajectory groups, adjusted for covariates. Since stepwise logistic regression analysis reported higher RRs of BMI trajectory groups and WHtR trajectory groups, nomogram models were employed to describe the impacts of those two variables, respectively. Receiver operating characteristic (ROC) curve analysis was used to evaluate the predictive power of BMI trajectory groups and WHtR trajectory groups for risk of hypertension incidence based on the area under curve (AUC). Further, all analysis were performed by gender to investigate gender-speci c ability of adiposity indices and metabolic status to predict the risk of hypertension incidence.
Croup-based trajectory models were conducted in Stata using plugin STATA TRAJ. All the other analysis was performed in R project. Table 1 shows average levels of study variables in 4-category BMI_MS according to levels of baseline of BMI and MS. All variables were found signi cant difference among four groups. Participants with nonoverweight and normal Mets were younger than the other three groups (P < 0.001), moreover, man were more likely to be in overweight or abnormal MS than women (P < 0.001). Those with non-overweight and normal MS showed the lowest levels of WC, LDL, TG, TC, FBG, creatinine, urea, uric acid, ALT and AST but the highest level of HDL compared to the other groups at baseline (P < 0.001). Similar characteristic of study variables in groups by levels of baseline WHtR and Mets were reported in Table 2. Table 3 shows prevalence per 100 persons, and gender-speci c characteristics of the study population of hypertension at baseline. At baseline, 3496 man (48.1%) and 3779 (51.9%) women participated in this study. The average age of this study was 48.6. Among men, age, WC, LDL, TC, creatinine, uricacid, ALT, AST, TG, FBG, MS, BMI, WHtR, BMI_MS and WHtR_MS were found signi cant difference with hypertension prevalence, but not urea and HDL. Among women, hypertension prevalence was signi cantly associated with age, WC, LDL, TC, uricacid, ALT, AST, HDL, TG, FBG, MS, BMI, WHtR, BMI_MS and WHtR_MS, while no signi cant associations were reported between hypertension prevalence and creatinine or urea. Table 4 showed baseline characteristics of hypertension-free population and new cases of hypertension developed during 5-year follow-up by gender. The cumulative incidence of hypertension was 18.4% in the total population, and the males had a signi cant higher incidence 23.3% than females 14.2% (P < 0.001).

Characteristic of All Subjects
Among man, hypertension cases were more likely to be older at baseline and showed signi cant higher levels of WC, creatinine, urea, uricacid, AST, TG and FBG, as well as they were more likely in high levels of abnormal MS (60.9% vs. 46.3%), BMI (73.4% vs. 64.1%), WHtR (66.9% vs. 52.2%), BMI_MS (48.2% VS. 36.3) and WHtR_MS (46.1% vs. 32.0%). Consistent results were also found among women. Signi cant associations of hypertension with LDL, TC and ALT were found in woman but not in man.

The association between hypertension incidence and BMI, WHtR and MS
The risk of hypertension incidence associated with BMI and WHtR as continuous variable were evaluated by restricted cubic splines. As showed in Fig. 2, nonlinear dose-response associations of hypertension incidence were signi cant related to BMI (P < 0.001) and WHtR (P < 0.001), respectively. The risk of hypertension incidence increased nonlinearly with continuous change of BMI or WHtR. Strati ed analysis by gender found that the ideal BMI and WHtR of female was less than male. Thus, female is more likely to have higher risk of hypertension incidence compared to male. We also found that female is more sensitive to BMI and WHtR than male with lower cutoff point.
Gender-speci c effects of adiposity indices and metabolic status on hypertension incidence with the use of stepwise logistic regression model were showed in Table 3 The association between hypertension incidence and BMI trajectories and WHtR trajectories Both BMI and WHtR trajectory models were obtained as four liner trajectory groups based on BIC (Fig. 3). The four trajectories were labelled in terms of the baseline value and dynamic during the 4-year follow-up. BMI trajectories model include the following groups: persistent normal (n = 1376, 23.0%), persistent overweight (n = 2526, 40.7%), persistent obese (n = 1706, 28.3%), persistent the highest (n = 477, 8.0%). WHtR trajectories model include the following groups: persistent low (n = 1469, 24.4%), persistent medium (n = 2560, 41.4%), persistent high (n = 1736, 28.7%), increasing to higher (n = 320, 5.6%).

The Predictive Power Of Adiposity Indices Trajectories
Consequently, a nal model including age, TG, FBG and adiposity indices trajectories was built as a nomogram (Fig. 5). The nomogram illustrated that age is a strong risk factor to predict hypertension incidence. It seems that WHtR trajectory groups have greater contribution than the BMI trajectory groups. Each category within these variables was assigned a score on the point scale. By adding up the total score and locating it on the total point scale, probability of recurrence could easily be estimated. Strati ed analysis showed that both WHtR trajectory groups and BMI trajectory groups have higher contribution in male than female. Figure 6 reported the predictive power of adiposity indices trajectories for risk of hypertension incidence using ROC curve analysis. In total population, BMI trajectories and WHtR trajectories showed similar ability to predict the risk of hypertension incidence with AUC 0.713 (P < 0.001) and 0.720 (P < 0.001). After strati ed by gender, BMI trajectories and WHtR trajectories showed higher power in female than male (BMI trajectories: 0.762 vs. 0.662; WHtR trajectories: 0.769 vs. 0.663).

Discussion
In the occupational population study in urban Beijing, our study reported the hypertension incidence was 18.4% in total population. The risk hypertension incidence was signi cantly associated with continuous changes in BMI or WHtR. The combination effect of BMI and MS or WHtR and MS showed the highest risk ratio compared to the other counterparts, furthermore, signi cant interaction between adiposity indices and MS were reported simultaneously. We also found signi cant dose dependent effect of BMI trajectories and WHtR trajectories related to hypertension incidence. Among those factors, BMI trajectories or WHtR trajectories have higher ability to predict the risk of hypertension incidence than the others. Moreover, gender-speci c analysis showed female was more likely in higher risk of hypertension incidence than male.
Few studies have been conducted on the relationships between hypertension incidence and adiposity indices (BMI and WHtR) as continues variable. In this study, the risk of hypertension incidence increased nonlinearly with continuous change of BMI or WHtR. Similar results were reported that signi cant nonlinear association between the continuous change of BMI and hypertension after adjusting the covariates strati ed by gender and age groups indicating a dose-response relationship between adiposity indices and hypertension (39). Moreover, the other studies support our results hold that weight loss could lower blood pressure in both hypertension and non-hypertension population (40,41).
Most of research found that the risk of hypertension increased with adiposity indices (22,23,(42)(43)(44)(45)(46). Moreover, a recent review conducted on the impact of adiposity on hypertension, included 46 studies of BMI and 12 of WC, found that obese males in lean populations were 45% more likely to be hypertensive compared to obese males in not-lean populations (47). Among those researches, some hold BMI has the highest risk ratio compared to other adiposity indices (46), while some thought WHtR was a better risk factor than the others for hypertension (22,23). In line with our results, one suggested those adiposity indices shared similar power for the risk of hypertension (45). In addition, the current study found signi cant interaction between adiposity indices and MS, furthermore, the combination effect of BMI and MS or WHtR and MS showed the highest risk ratio compared to the other counterparts. The results could be explained that metabolism also played an important role in hypertension (30,31), obese individuals exposed abnormal MS might plus the risk of hypertension.
Four trajectory groups of BMI and WHtR t the data best in the model. The majority of our participants was in consistent status or slightly increased over time. Compared to the lowest group of BMI or WHtR, the other three groups showed higher risk of hypertension, even after adjustments for demographic confounders. The results were generally in line with previous research showing long-term BMI gain during adulthood associated with adverse health outcomes including hypertension (48,49), and weight loss might a critical lifestyle modi cation for obese people to control hypertension (16,36,50). We also found signi cant dose dependent effect of BMI trajectories and WHtR trajectories related to hypertension incidence. The higher level of BMI trajectory groups or WHtR trajectory groups the participants are the higher risk of hypertension they would have. Moreover, WHtR trajectory groups were more likely to have higher risk of hypertension compared to BMI trajectory groups consistent with previous studies to a certain degree (51,52).
Most of studies were conducted to exploit the optimal adiposity index for hypertension but remains controversial and inconclusive. Although group-based trajectory model has been widely used to t the variance of considerable diseases over time, few studies were conducted to exploit trajectories of BMI or WHtR. In this current study, we found BMI trajectories and WHtR trajectories have higher predictive power than baseline value of BMI and WHtR. Several reasons could be considered, the data of trajectories was collected every time during follow-up, so that it could describe more variability of adiposity indices over time in line with expectations. Furthermore, the description of the trajectory is closer to the actual situation in real life due to variability, and we could not ignore the fact that the nature of adiposity indices is complex and di cult to estimate.
Regarding to gender-speci c effect of adiposity indices for hypertension, we found that female were more likely in higher risk of hypertension incidence than male in line with previous studies (25,26). Those results could be explained by the following points. First, female with high level of adiposity indices seem to be more susceptible to the adverse health outcomes caused by in ammation and lipid metabolism than male. Second, the difference in the role of gender hormones between female and male might be another potential mechanism. Third, the distribution of body fat is strongly affected by gender hormones and the visceral fat gain is a well-known risk factor for hypertension, in addition, gender hormones could indirectly affect human behavior, for example dietary preference or physical activity level, which may also increase the risk of hypertension.
Several limitations should be pointed. First, hypertension was estimated only at follow-up time, information bias could not be ignored when any information regarding hypertension was ascertained inaccurately due to participants might occur hypertension during the interval between the last and next test. Second, the relationships between adiposity indices and hypertension should be interpreted with caution due to the wide ranges of age in this study. Finally, the data were collected from Chinese population in a single province; thus, the generalizability of the results to other populations remained to be veri ed.
These limitations are balanced by several strengths. First, the current study exploit the risk of hypertension related to not only the baseline data of exposures but also the trajectories of exposures across time. Second, gender-speci c effect was minimized as the strati ed analyses were performed. Final, the study covering most kinds of work was representative for occupational population in Beijing.

Conclusion
In conclusion, obese and abnormal MS increase the risk of hypertension incidence, in addition, BMI and WHtR trajectories has higher predictive power for hypertension incidence compared to baseline data. Female is more vulnerable to obese and abnormal MS than male.

Author Contributions
Ya-ke LU, Jing DONG and Yu-xiang YAN designed and conducted the study. Ya-ke LU cleared the data with the help of Li-kun HU and Yu-hong LIU. Ya-ke LU and Jing DONG analyzed the data and wrote the manuscript with support from Yu-xiang YAN and Yue SUN. All authors reviewed the manuscript.
Availability of data and materials Not applicable.   Table 4 Gender-speci c effects of body fat and metabolic status on hypertension incidence with the use of stepwise logistic regression model.