U-shaped association between BMI and the risk of PAD in Chinese hypertensive population


 Background: High body mass index (BMI) is a well-recognized risk factor of cardiovascular diseases. But its role in peripheral artery disease (PAD) remains perplexing. Our study evaluated the association of BMI with PAD in Chinese hypertensive population. Methods: This is a cross-sectional study with enrollment data from the Chinese H-type Hypertension Registry.10896 hypertensive patients aged ≥18 years were included in the final analysis. Results: The prevalence of PAD diagnosed by ABI in this study was 3.2% (n=351). A U-shaped relationship between BMI and PAD was found. Per SD increment (3.6 kg/m2) on the left side of the BMI threshold (BMI < 25.7 kg/m2) was associated with a 27% decreased in the adjusted risk of PAD [OR, 0.73; 95% confidence interval (CI) 0.60, 0.89; P=0.002]; BMI was significantly positively associated with the risk of PAD (OR, 1.52; 95% CI 1.52, 1.93; P=0.001) in those with BMI ≥25.7 kg/m2. Conclusion: A “U-shaped” relationship between BMI and the risk of PAD in Chinese hypertensive population was found. BMI with the lowest risk of PAD was estimated to be 25.7 kg/m2.


Introduction
Peripheral arterial disease (PAD) is the third leading atherosclerotic disease after coronary heart disease and stroke [1], mainly caused by the accumulation of lipid and brous material between the intima and media of lower limb arteries, resulting in luminal stenosis (focal or diffuse). It is well known for a sharp increase in the prevalence of PAD with advanced age [2,3]. With the aging of the Chinese population, PAD has become an increasingly severe clinical and social problem. Allison et al. also showed ethnic differences were independent factors in the prevalence of PAD [4]. Compared to Whites, Blacks seem to be more vulnerable to PAD, while Asians seem to have a lower prevalence of PAD [5].
The prevalence of PAD was higher in people with underweight, but the association between BMI and PAD was uncertain due to a variety of potential covariates [6,7]. A small prospective cohort study showed that obesity independently predicts severe PAD [8]. However, the recent observational study with more than 3 million sample size has found "J-shaped" relationship between BMI and PAD only in females [9].
Epidemiology of Dementia in Central Africa (EPIDEMCA) study recruited the elderly in the Central African Republic and the Republic of Congo, showed underweight and obsity were all associated with the risk of PAD [10].
Due to the inconsistent and the evidence of relationship between BMI and prevalence of PAD in the Chinese was still lacked. Our study aims to explore the association between BMI and the risk of PAD in Chinese hypertensive patients.

Study Design and Participants
The study population was drawn from the China Hypertension Registry, a real-world observational registry of hypertension designed to investigate the prevalence and treatment of hypertension in China and to assess prognostic risk factors. Details of the inclusion and exclusion criteria for the study have been published [11]. From March 2018 to August 2018, we recruited a total of 14,268 study participants in Wuyuan, Jiangxi Province, China as our study population, and nally analyzed the data of 10802.

Laboratory Biochemical Examination
All subjects were asked to do an overnight fast Venous blood samples were obtained from all study participants and analyzed by Biaojia Biotechnology Laboratory in Shenzhen, China. Lipids (including total cholesterol (TC, mmol/L), triglycerides (TG, mmol/L), high-density lipoprotein-cholesterol (HDL-C, mmol/L)), estimated glomerular ltration rate (eGFR, ml/min/1.73 m 2 ), fasting blood glucose (FBG, mmol/L) and homocysteine (Hcy, µmol/L) were measured using automatic clinical analyzers (Beckman Coulter, USA) and the laboratory staff were blind to the research protocol.

Measurement Of BMI
The height and weight of the subjects were measured by trained staff using standardized equipment in accordance with standard operation procedure. BMI = Weight (kg)/Height (m) 2 .

Measurement Of ABI And De nition Of PAD
The ABI of each lower limb was calculated by dividing the systolic pressure of the ankle artery of the corresponding lower limb by the systolic pressure of the brachial artery. Subjects rested quietly in a warm room for more than 10 minutes and fully exposed their upper limbs and ankles. Trained technicians used the Omron Colin BP-203RPE III device (Omron Health Care, Kyoto, Japan) to simultaneously measure bilateral brachial and ankle arterial systolic pressures in supine subjects. And the software automatically calculates the bilateral ABI data according to the above calculation formula. All measurements were conducted in accordance with strict standard protocols. PAD was de ned as an ABI ≤ 0.9 in either lower limb [12]. Subjects with ABI > 1.4 were excluded because of abnormal elevation of ABI may due to calci cation of the arterial wall [13].

Other Variables
Variables included age (years), sex, systolic blood pressure (SBP, mmHg) and diastolic blood pressure (DBP, mmHg) measured by electronic sphygmomanometers after the subjects had rested for 10 minutes.
Quali ed researchers were trained to collect information by using standardized questionnaires, including smoking status (never, former, current), alcohol consumption (never, former, current), antihypertensive drugs (yes or no), the history of comorbid diseases including diabetes mellitus (yes or no), stroke (yes or no), and coronary heart disease (yes or no).

Statistical Analysis
Continuous variables were expressed as mean ± standard deviation (SD), and categorical variables as percentage (%). Population characteristics were described according to BMI classify. Smoothing curve (penalized spline method) was used to show the relationship between BMI and the prevalence of PAD.
Threshold effect analysis was used for in ection points of BMI by using piecewise model tting data. Multivariate logistic regression was used to analyze the relationship between BMI and the risk of PAD around threshold value. P value for interaction was used to compare whether there was a signi cant difference in the correlation between BMI and the risk of PAD before and after in ection point. In addition, possible modi cations of the association between BMI and PAD were assessed for variables including sex, age, blood pressure controlled, pulse rate, Hcy, lipids pro le, smoking status, history of diabetes mellitus and stroke.
All analyses in this study with P values < 0.05 (two-tailed) were considered statistically signi cant. All analyses were statistically analyzed by EnpowerStats (www.empowerstats.com; X&Y Solutions, Inc., Boston, MA) and R statistical software (http://www.r-project.org).

Baseline characteristics of participants
As shown in Table 1, a total of 10896 hypertensive patients with a mean age of 63.9 ± 9.3 years were included in this study. The prevalence of PAD was 3.2%, the mean BMI was 23.6 ± 3.6 kg/m 2 , and 47.1% were male. BMI was strati ed to four groups: underweight (BMI < 18.5 kg/m 2 ), normal (BMI ≥ 18.5, < 25 kg/m 2 ), overweight (BMI ≥ 25, > 30 kg/m 2 ) and obesity (BMI ≥ 30 kg/m 2 ) to describe demographic characteristics. The underweight of participants accounted for 6.3% of the total population, and obesity was only 4.2%. The prevalence of PAD in underweight was the highest (6.7%) and followed by obesity (4.4%), while overweight was only 2.3%. Compared with the other three groups, underweight participants were older, with higher tHcy, HDL-C, current smoking rate, and lower TC, TG, eGFR, the prevalence of diabetes mellitus and the use of the antihypertensive drug. Values are N (%) or mean ± SD. BMI = Body mass index, SBP = systolic blood pressure; DBP = diastolic blood pressure; PAD = Peripheral vascular disease; HDL-C = high-density lipid cholesterol; FBG = Fasting blood glucose; tHcy = total Homocysteine; eGFR = estimated glomerular ltration rate; CHD = Coronary heart disease.
Association Between BMI and PAD As shown in Fig. 1, the relationship between BMI and the prevalence of PAD showed a U-shaped curve, and threshold saturation effect analysis showed that BMI value with the lowest risk of PAD was estimated to be 25.7 kg/m 2 . We strati ed BMI by 25.7 kg/m 2 and used logistic regression analysis models (   whether in the hypertensive population with BMI < 25.7 kg/m 2 or BMI ≥ 25.7 kg/m 2 (All strati ed Pinteractions were > 0.05) (Fig. 2).

Discussion
In our analysis of this community-based hypertension registry study in China, we noted a "U-shaped" relationship between BMI and risk of PAD. The BMI value with lowest risk of PAD was estimated to be 25.7 kg/m 2 .
A number of studies have reported the relationship between BMI and the risk of PAD. However, the association between BMI and PAD risk was not consistent. Epidemiological studies more than two decades ago reported a positive association between BMI and intermittent claudication in middle-aged males in Israel [14]. However, many population studies after adjusting for the relevant covariates fail to support the signi cant association between BMI and the prevalence of PAD [4,15]. In addition, the San Diego study reported an independent and signi cantly inverse association between BMI and prevalence of PAD (OR: .88) in multi-ethnic population [16]. Studies on the diabetic population in Taiwan showed that compared with diabetic patients without PAD, the BMI of patients with PAD was lower (23.5 ± 3.2 vs.24.8 ± 3.5 kg/m 2 , P < .005). Heffron et al. who gathered data from more than 20 000 sites (n = 3 250 350) in the United States from 2003 to 2008, recently reported BMI and the prevalence of PAD in females showed a "J-shaped" nonlinear relationship; a signi cant positive correlation between obesity and PAD in females, while only a slight positive correlation between obesity (BMI ≥ 40 kg/m 2 ) and PAD in males (OR = 2.98 vs. 1.37) [9]. Stepwise logistic regression analysis showed that the association between BMI and PAD was inverse [17].
To our knowledge, the "U-shaped" relationship between BMI and the risk of PAD shown in our study was the rst reported in Chinese population. Different from the very large sample population studies [9] in the United States, where participants were nearly 30% obese and 3.4% underweight, as well as study of the prevalence of PAD in African [10], where obesity was only 4.5%, 34.1% underweight, we were 6.3%(691) underweight and only 4.2%(454) obesity, nearly 90% of the population was normal BMI and overweight. Over a third of the study population was underweight. A "U-shaped" relationship between BMI and the risk of PAD was observed. Compare to the subjects with normal BMI, underweight and obesity were statistically signi cant association with the risk of PAD (OR, 2.09; 95%CI 1.35, 3.22; p = .0009; OR,1.90; 95% CI 1.04, 3.23; p = .0336), but not overweight (OR, 1.56; 95% CI 0.70, 2.51; p = .7342) [10]. However, Heffron et al. found a "J-shaped" relationship between BMI and PAD only in females, not in males, which may be due to the height and weight data used in this study for self-reporting of participants. Selfreported data may lead to personal BMI classi cation appear serious mistakes [18], di cult to correct the mistakes [19], especially in the strati ed analysis according to gender [20]. Thus, self-report bias may have contributed to the fact that this study found a "J-shaped" relationship between BMI and PAD risk only in females, and not in males.
At present, few studies have elaborated on the possible mechanism of the correlation between BMI and PAD. A cross-sectional study of hemodialysis patients reported a lower prevalence of atherosclerosis and lower levels of in ammation (CRP) in patients with normal BMI and overweight compared with those with underweight and obesity [21]. Lower levels of in ammation and atherosclerosis may be associated with the lowest risk of PAD in this population (normal BMI and overweight).
Not only that, there have been also many reports on the "U-shaped" relationship between BMI and cardiovascular disease and death. A meta-analysis of 97 studies showed that obesity (all grades) and grades 2 and 3 obesity were signi cantly associated with all-cause mortality relative to normal BMI.
However, overweight was associated with a signi cant reduction in all-cause mortality [20]. Among more than 1 million East Asian populations in the Asia Cohort Consortium BMI Project, including Chinese, Japanese, and Korean, the Cox proportional hazard regression model was used to analyze the relationship between BMI and mortality risk, which showed that the population with BMI between 22.6 and 27.5 had the lowest mortality risk [22]. Based on this, we speculate that the "U-shaped" relationship between BMI and peripheral atherosclerosis may, on one hand, explain the causes of the lowest cardiovascular disease risk and all-cause mortality in normal BMI/overweight.

Limitations And Future Directions
Nonetheless, these results must be interpreted with caution, and a number of limitations should be borne in mind. First, subjects in our analysis were middle-aged and elderly patients with hypertension. The "Ushaped" relationship between BMI and the risk of PAD was not necessarily applicable to the general population, but as an independent risk factor for PAD, exploring the relationship between BMI and the risk of PAD in the hypertensive population can serve the high-risk population more precisely. In addition, the association between BMI and the risk of PAD was still controversial. By design, our study was a crosssectional study and cannot study the chronology of BMI and PAD. There might be a reverse causal relationship. The weight change caused by the disease may distort the relationship between BMI and PAD. In the future, large prospective cohort studies on PAD were urgently needed. Final, the obesity rate in our study was low. It has no enough power to assess the relationship between different degrees of obesity or morbid obesity and the risk of PAD. However, our study re ects the real situation of hypertension population in Chinese hypertention, and the results obtained were more suitable for the application of hypertension in middle-aged and elderly people in China.

Conclusions
Our study reported the prevalence of PAD was 3.2%. The "U-shaped" relationship between BMI and the risk of PAD was found in Chinese middle-aged and elderly patients with hypertension. BMI with the lowest risk of PAD was estimated to be 25.7 kg/m 2 in our study.