Analysis of Dose-response Relationship between BMI and Hypertension in Northeastern China Using Restricted Cubic Spline Functions

High body mass index (BMI) was significantly associated with hypertension. The purpose of this study is to investigate the association between BMI and hypertension in people in northeast China. Our study was a cross-sectional study conducted from June to August 2012. According to multistage, stratified cluster sampling, a total of 21435 inhabitants aged between 18 and 79 years in Jilin Province were selected randomly. The prevalence of hypertension was 35.66% overall. After adjusting for potential confounders, the multivariable-adjusted odds ratios for the BMI- hypertension association for overweight and obesity were 2.503 (95% confidence interval = 1.912–2.204) and 4.259 (95% confidence interval = 3.883–4.671). The results of multivariable restricted cubic spline regression analysis showed that there was a non-linear relationship between the continuous change of BMI and hypertension (P < 0.001) after adjusting the confounding factors of different genders and age groups, which indicated that there was an adjusted dose-response association between continuous BMI and hypertension.

higher than that in the general population 14 . A prospective analysis also showed that the high prevalence of hypertension in obese patients(>60%) account for 78% of incident of hypertension in men and 64% in women 15 .
Although the classification system of obesity is not exactly identical, body mass index(BMI) is the most commonly used measure, dividing obesity into overweight and various grades 16,17 . Obesity has been shown to be associated with a variety of diseases, such as cancer 18 , hypertension, dyslipidemia, diabetes and cardiovascular disease (CVD) [19][20][21] . There are few studies to quantify the relationship between BMI levels and hypertension, especially using some intuitive methods such as restricted cubic splines (RCS) 22 . Moreover, nobody has quantified the relationship between BMI levels and hypertension based on the population of Northeast China 15 .
The purpose of this study is to analyse the relationship between BMI and hypertension in people in northeast China. In our study, we used RCS function in dose-response analysis to adjust potential confounding factors to quantify the association between BMI and hypertension. In this way, our study can provide information for people's health promotion.

Methods
Study design. The study which was carried out in Jilin Province was a community-based, cross-sectional study. It starting from June 2012 and ended in August 2012. Research object of the study was the people aged between 18 and 79 years old, and the people must have lived in Jilin Province for more than 6 months. Multistage stratified cluster sampling was used to select research object from cities representative of the distribution of the Jilin population. The detailed sampling procedure had described previously 23,24 . The Ethics Committee of Jilin University School of Public Health approved the study (Reference Number: 2012-R-011). Before participating in this survey, all participants had provided informed consent.
Definitions. BMI was defined as the ratio of body weight (kilograms) to the square of height (meters). Based on the Chinese criteria of weight for adults, BMI < 18.5, 18.5 ≤ BMI < 24.0, 24.0 ≤ BMI < 28.0,BMI ≥ 28.0 were respectively defined as underweight, normal, overweight and obese 22 . Hypertension was defined as systolic ≥ 140 mmHg and/or diastolic ≥ 90 mmHg following the Chinese Hypertension Prevention Guide (2010 Revised Edition).According to WHO criteria(1999): Through glucoseoxidase method, FPG 7.0 mmol/L and/or 2hPG11.1 mmol/L were regarded as diabetes. Abnormal blood lipids (TC ≥ 5.18 mmol/L or TG ≥ 1.70 mmol/L or HDL-C < 1.04 mmol/L or LDL-C ≥ 3.37 mmol/L) was diagnosed as hyperlipidemia based on the criteria of the "Chinese Guidelines on Prevention and Treatment of Dyslipidemia in Adults".
Occupations were classified into 3 groups: mental labor, manual labor, and retirees or unemployed. Mental labor was defined as the person who was managers, administrators, or technicians. The people who worked in the manufacturing industry, agriculture, or service industry were classified into manual labor. According to average monthly household, household incomes were classified into6 categories: below 500 yuan, 500 to <1000 yuan, 1000 to <2000 yuan, 2000 to <3000 yuan, 3000 to <5000 yuan, ≥2000 yuan. Smoking was defined as never, ever (smoking at least 100 cigarettes in their lifetime but did not smoke at all), or current. Drinking habits provided two options-"0" and "1". "0" was defined as no smoking, and "1" was defined as smoking. Exercise habits were classified into 3 categories based on exercise frequency. The exercise frequency of ≥3times a week was defined as "often exercise", once or twice a week was defined as "sometimes exercise", the others were defined as "never or rare exercise". Statistical analysis. Demographics are presented as numbers and frequency distributions for categorical variables as appropriate. Univariate logistic regression analyses were used to determined significant differences between hypertension and nohypertension individuals. The association between BMI and hypertension was investigated by using unconditional multivariable logistic regression models and models which adjusted for age, gender, education level, and other variables, and to evaluate the potential confounding effects among risk factor variables. Multivariable-adjusted odds ratios(ORs) and their 95% confidence intervals (CIs) in 3 different logistic regression models were calculated by the SPSS statistical package, version 19.0 (IBM Corp, Armonk, NY) independently. RCS were used to detect the possible nonlinear dependency of the relationship between the risk of hypertension and BMI levels, using 4 knots at prespecified locations according to the percentiles of the distribution of BMI, 18 kg/m 2 , 22 kg/m 2 , 25 kg/m 2 , 27 kg/m 222 . Stata 12.0 (StataCorp, College Station, TX) was used to carry out the above-mentioned dose-response analyses. A significance level of P < 0.05 (2-tailed tests) was used.

Characteristics of the study population.
A total of 21,435 samples participated in our study, screened for data and eventually included in the study for 20,839 by deleting the missing height/weight values. Among the 20,839 participants, the prevalence of hypertension was35.66% (7,431/20,839), with a number of 3,836 for male and 3,595 for female. The mean age was (47.27 ± 13.34) years. Compared with hypertension participants, the no hypertension individuals had a lower BMI (23.46 ± 3.55 kg/m 2 vs 25.58 ± 3.65 kg/m 2 , P < 0.001).
The key variables for hypertension and no hypertension considered in our study are listed in Table 1. Significant differences were found in variables gender,age, residential areas, education, marital status, occupation, household income, smoker, drinking, and exercise (P < 0.001). Table 2 showed the results of univariate and multivariate logistic regression analyses. With the BMI classification as the only covariate, the ORs applied by unadjusted univariate logistic regression model were 0.40(95%CI = 0.33-0.49) for underweight, 2.17(95%CI = 2.0-2.32) for overweight, 3.79(95%CI = 3.48-4.13) for obesity (P value < 0.001). After adjustment for age and gender, the adjusted odds ratios showed a consistent association between BMI classification and hypertension across the model. In a fully adjusted model, the impact of BMI was slightly diminished when adjusting for residential areas, education, (2019) 9:18208 | https://doi.org/10.1038/s41598-019-54827-2 www.nature.com/scientificreports www.nature.com/scientificreports/ marital status, occupation, household income, smoking and drinking status, exercise. Although the association of BMI classification with hypertension prevalence was reduced with sequential adjustments, BMI showed a significant and clear gradient from the lowest underweight to the highest obesity level.

Dose-response relationship between BMI and hypertension.
Based on the stratification of gender and age, we used RCS model with 4knots to simulate the relationship between BMI and the risk for hypertension. Significant nonlinear dose-response association was showed in the relationship between BMI and the risk of  Figures 1 and 2 showed the nonlinear dose-response association, with confounders being adjusted.
Similar relationships between hypertension risk and BMI were found in different age groups when using 4 knots (all P value of nonlinear < 0.001; Fig. 3). In young adults (18-

Discussion
In the population-based study of Jilin Province in Northeastern China, our study reported the prevalence of hypertension was 35.66% including previously diagnosed hypertension and newly diagnosed hypertension. Regardless of gender or age, after adjusting for confounding factors, continuous changes in BMI are significantly associated with the risk of hypertension. And dose-response relationship analysis showed that with the continuous change of BMI, the association strength of hypertension increased nonlinearly.
On a national scale in China, findings based on different backgrounds' survey in the past 30 years basically showed an upward trend in the prevalence of hypertension 25 . China's chronic disease monitoring in 2010 reported that the national prevalence of hypertension was 33.5%, which was lower than the rate of 35.66% reported by our study 26 . Our rate was higher than the rate of 17.2% displayed by the monitoring results of Kunshan city in the same year 22 , but not reached the rate of 38.6% of Jiangsu Province 27 .
As early as 1996, study 19 showed that the increasing of BMIs can make blood pressure increase, which indicated that there was not only a strong relationship between BMI and hypertension, but also was an association between the continuous variables of BMI and blood pressure. Our study and the study 22 conducted in kunshan city confirmed that there was nonlinear dose-response relationship between BMI and the risk of hypertension.
Although the mechanism of how obesity caused hypertension was not yet clear, epidemiological evidence had highlighted that there was a consistent correlation between obesity and hypertension, and obesity predisposes hypertension 28 . The pathophysiological explanation of obesity predisposing hypertensionis elevated cardiac output, which perhaps attribute to excess intravascular volume and reduced cardiac contractility 29 . There was evidence suggested that in obesity individual, nutritional status alteration, gut microbiota, sunlight exposure and physical activity increase played an important role in the hypertension or not 30 . Further research investigating dose-response association between BMI and hypertension would be a significant step to decrease the social burden of hypertension.

conclusion
Our outcomes demonstrated that dose-response relationship exists between BMI and hypertension, and increased BMI is an independent and adjusted dose-dependent risk factor for hypertension among overweight and obesity participants in China.

Figure 3.
Association between BMI and the risk of hypertension in female, allowing for nonlinear effects, with 95%CI. (A) Young adults. Using 4 knots restricted cubic spline for BMI model adjusted gender, residential areas, occupation, smoker, drinking exercise, marital status, and education. (B) Middle-aged adults. Using 4 knots restricted cubic spline for BMI model adjusted gender, smoker, drinking, and exercise. (C) Older adults. Using 4 knots restricted cubic spline for BMI model adjusted gender, smoker, and exercise.BMI, body mass index; CI, confidence interval; OR, odds ratio.