Association Study of Fasting Blood Glucose and Salt Sensitivity of Blood Pressure in Community Population: The EpiSS Study

Wenjuan Peng Capital Medical University School of Public Health Yunyi Xie Capital Medical University School of Public Health Han Cao Capital Medical University School of Public Health Han Qi Capital Medical University A liated Anding Hospital Kuo Liu Capital Medical University School of Public Health Juan Xia Capital Medical University of Public Health Zheng Liu Peking University People's Hospital Xiaohui Liu Capital Medical University School of Public Health Bingxiao Li Capital Medical University School of Public Health Fuyuan Wen Capital Medical University School of Public Health Fengxu Zhang Capital Medical University School of Public Health Ling Zhang (  zlily_epi@ccmu.edu.cn ) Capital Medical Univeristy https://orcid.org/0000-0002-6132-4258

mortality independent of BP, and the correlation between cardiovascular diseases with SS is as strong as that with BP [6,7]. Therefore, a better understanding of the risk factors associated with SS may possibly reduce the huge burden of cardiovascular diseases.
Several studies have showed that the FBG level was higher in SS individuals than that in SR groups in different populations, including healthy male (n = 18) [16], obese subjects with mild hypertension (n = 18) [17], hypertensive patients (n = 99) [18] and young normotensive subjects (n = 23) [19]. Chen et al. perform a low and high-sodium diet test in adults and found that the changes of BP were signi cantly greater in participants with metabolic syndrome [20]. According to the statement from the American Heart Association in 2005, the abdominal obesity, raised BP, high triglycerides (TG) concentration, low highdensity lipoprotein cholesterol (HDL-C), or elevated plasma glucose were the main risk factors of metabolic syndrome [21]. This research showed that the risk of salt sensitivity rose with increasing numbers of risk factors for metabolic syndrome [20]. These evidences indicated that FBG might played a more important role in the development of SS. Takashi et al. suggested that hyperglycemia may enhance the reabsorption of sodium through sodium glucose cotransporter 2 in proximal tubule especially in diabetes patients [22], thus might stimulate the occurrence of SS. However, whether the level of FBG could be a risk factor for SS has been largely overlooked in these previous studies due to the small sample size and focusing on the insulin or insulin resistance.
Up to now, few population-based epidemiological studies have focused on the association between blood glucose levels and SS. We hypothesized that FBG might be an independent risk factors, and dosedependent associated with the SSBP. Therefore, based on the study of systemic epidemiology of salt sensitivity (EpiSS), we aimed to analyze the relationship between FBG and BP changes both in acute salt load period and in diuresis shrinkage period, and to compare the differences in SS prevalence and BP changes among subjects with different blood glucose levels.

Participants
Data analyzed in this study was the baseline of the EpiSS study. This study was registered in the Chinese Clinical Trial Registry (No: ChiCTR1900024725, http://www.chictr.org.cn/index.aspx). The protocol for selection of the subjects, sample collection and measurement methods have been described in detail previously [23]. Brie y, participants aged 35 to 70 years were recruited from ve community health centers in Beijing and six community health centers in Liaoning Province during July 2014 and July 2016. Patients with hypertension and diabetes were required to stop the intake of all antihypertensive and antidiabetic drugs for at least 24 hours. In addition, patients with secondary stage and above hypertension [systolic BP (SBP) > 160 mmHg and (or) diastolic BP (DBP) > 100 mmHg], and individuals with clinical diagnosis of cardiovascular disease, kidney disease, liver disease or malignant tumors were excluded.

Determination of salt sensitive
In EpiSS study, the modi ed Sullivan's acute oral saline load and diuresis shrinkage test (MSAOSL-DST) [24][25][26] was performed to evaluate the SSBP. As descripted in the previous study [23], MSAOSL-DST contains the following steps: First, the baseline BP (BP 0 ) was measured twice and mean BP was calculated before test. Then, the subjects were asked to take 1000 mL of 0.9% saline solution orally within 30 minutes, and the second time BP (BP 1 ) was measured after two hours from the time individual nished drinking saline. Third, immediately after the second time BP measurements, the subjects were given a 40 mg furosemide, and the third time BP (BP 2 ) was measured two hours after taking furosemide.

Data collection and variables
A structured questionnaire was administered by trained staff to obtain sociodemographic information and behavior habits, including age, sex, smoking and alcohol consumption status, dietary, sleep and exercise habits; medical history including coronary artery disease, hypertension, diabetes and stroke, and medication history including the use of antihypertensive drugs including calcium channel blockers, angiotensin converting enzyme inhibitors, angiotensin receptor inhibitors, diuretic, beta-blockers, the compound preparation and Chinese patent medicine, and antidiabetic drugs including biguanides, sulfonylureas, thiazolidinediones, glinides, alpha-glycosidase inhibitors and insulin. BP were measured after at least 15-min rest with an automatic sphygmomanometer (Omron HEM-7118, Japan) [28]. BP measurement was carried out twice and the mean value was calculated for data analysis. Hypertension was de ned as SBP ≥ 140 mmHg and/or DBP ≥ 90 mmHg, or a self-reported physician diagnosed with hypertension, or an individual currently using antihypertensive drugs [29,30]. According to the Working Group on Obesity in China [31,32], normal fasting glucose (FNG) was de ned as 2.80 ≤ FBG < 6.11 mmol/L, impaired fasting glucose (IFG) was de ned as 6.11 ≤ FBG < 7.00 mmol/L, diabetes was de ned as FBG ≥ 7.00 mmol/L. Furthermore, diabetes was also de ned as a self-reported physician diagnosis of diabetes, or taking oral hypoglycemic medication or insulin [33].
The anthropometric examinations included height, weight, waist and hip circumference. Body mass index (BMI) was calculated as a person's weight in kilograms divided by the square of his/her height in meters (kg/m 2 ), and waist-to-hip ratio (WHR) as the ratio of waist circumference to hip circumference (cm). BMI was grouped according to the guidelines for prevention and control of overweight and obesity in Chinese adults [34], normal weight (BMI < 24.0 kg/m 2 ), overweight (24.0 ≤ BMI < 28.0 kg/m 2 ) and obesity (BMI ≥ 28.0 kg/m 2 ).
Fasting venous blood samples of each participant were extracted by venipuncture. The FBG was measured via the hexokinase/glucose-6-phosphate dehydrogenase method. The total cholesterol (TC), TG, low-density lipoprotein cholesterol (LDL-C) and HDL-C concentrations were determined by enzymatic methods.

Statistical analysis
All data analyzed were using the statistical package R (http://www.r-project.org) and SPSS 24.0 for Windows (SPSS, Inc., Chicago, IL, USA). A two-tailed P < 0.05 was statistically signi cant. Baseline characteristics of the participants were displayed as medians (interquartile range, IQR) for continuous variables (the continuous variables were non-normally distributed) and numbers (percentages) for categorical variables, and were further compared between SS and SR groups using Mann-Whitney U-test or Chi-square test.
The dose-response relationships between FBG with SSBP were conducted by restricted cubic spline (RCS), with knots of 10th, 50th and 90th percentiles of the FBG distribution and the 50th percentile of FBG set as the reference. The FBG was divided into quartiles (Q 1 , Q 2 , Q 3 and Q 4 ), and the associations between FBG with SS prevalence or with ΔBP 1,2 (ΔSBP 1,2 , ΔDBP 1,2 and ΔMAP 1,2 ) were evaluated using multivariate logistic regression models (odds ratio [OR] and 95% con dence interval [95%CI]) or multivariate linear regression models (beta coe cient [β] and 95%CI), adjusted for major covariables including age, sex, sleep (as a categorical variable), current smoking, current drinking, TG, LDL-C and MAP 0 . Strati ed analysis was performed by age, sex, BMI, hypertension status, diabetes status, smoking and drinking. In addition, sensitive analysis was performed to examine whether the antihypertensive or antidiabetic medications could in uence the associations between FBG with SSBP.

Characteristics of the Study Population
The characteristics of the study population and the comparison between SS and SR groups are presented in Table 1. A total of 2051 participants were included in the current study, with a median (IQR) age of 59 (54 ~ 63) years old, and 554 (27.01%) of male. There were 28.33% SS and 15.94% diabetes patients in participants. The median level of FBG and frequency of diabetes were signi cantly higher in SS group than those in SR group (5.55 vs. 5.39 mmol/L, P = 0.003; 19.62% vs. 14.49%, P = 0.005). Signi cant differences were also observed in variables including SBP 0 , DBP 0 , MAP 0 , TG, LDL-C, WHR, sleep, current smoking and drinking status (P < 0.05), and most of these variables were used as covariates for adjustment in multivariate analyses. The distribution of the ΔMAP 1 and ΔMAP 2 were left and right skewed (P for normality <0.001, Fig. 1), respectively. SBP 0 , the baseline systolic blood pressure; DBP 0 , the baseline diastolic blood pressure; MAP 0 , the baseline mean arterial pressure; FBG, fasting blood glucose; TC, total cholesterol; TG, triglycerides; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol; BMI, body mass index; WHR, waist-to-hip ratio; Antihypertensive drugs, includes calcium channel blockers, angiotensin converting enzyme inhibitors, angiotensin receptor inhibitors, diuretic, beta-blockers, the compound preparation and Chinese patent medicine; Antidiabetic drugs, includes biguanides, sulfonylureas, thiazolidinediones, glinides, alpha-glycosidase inhibitors and insulin. Blood pressure changes of the participants during the MSAOSL-DST Table 2 presented the BP changes of subjects during the MSAOSL-DST. The median (IQR) of ΔSBP 1 in total participants during saline load period were larger than zero, but the ΔDBP 1 was not. There were signi cantly different for all ΔBP between subjects with SS and SR (P < 0.001). Linear relationship between glucose and salt sensitivity of blood pressure Multivariable-adjusted RCS analyses suggested that there were signi cant associations of FBG with SS prevalence, ΔMAP 1 , ΔSBP 1 and ΔDBP 1 in total population (all P for overall <0.05, Fig. 2). However, no signi cant relationships between FBG with ΔMAP 2 , ΔSBP 2 and ΔDBP 2 were observed (all P for overall >0.05, Supplementary Fig. 1). We found evidence of linear associations of FBG with the prevalence of SS (P for nonlinear =0.1044), ΔMAP 1 (P for nonlinear =0.0762) and ΔSBP 1 (P for nonlinear =0.5013), but nonlinear relationship between FBG with ΔDBP 1 (P for nonlinear =0.0271).
Association of fasting blood glucose and salt sensitivity of blood pressure  Association of glucose and salt sensitivity of blood pressure in subgroups  Fig. 2).

Sensitive analysis
The associations between FBG with SS, ΔBP 1 , and ΔBP 2 were also analyzed after excluding patients (721 patients) who received antidiabetic drugs or antihypertensive medicines (Table 4 and Supplementary   Table 2). The positive dose-response associations between FBG with SS prevalence, or with ΔBP 1 were all slightly changed, while the results still signi cant. In Table 4, when compared to participants in the rst FBG quartile (Q 1 ), the adjusted ORs for SS prevalence in Q 2 , Q 3 and Q 4 were all elevated slightly (P for trend < 0.001). Meanwhile, when compared to NFG individuals, the adjusted ORs for SS prevalence in patients with IFG and diabetes increased slightly (P for trend < 0.001). Adjusted for age, sex, sleep, current smoking, drinking, triglycerides, low density lipoprotein cholesterol and baseline mean arterial pressure. * , P < 0.05; ** , P < 0.001.

Discussion
Our study found that a signi cant positive linear dose-response association between FBG and SSBP in 2051 Chinese adults from the EpiSS study. The associations of FBG with SS prevalence or all BP changes were signi cantly in saline loading period but non-signi cantly in diuretic shrinkage period. For every IQR increase in FBG, the SS prevalence was increased by 14%, and ΔMAP 1 , ΔSBP 1 and ΔDBP 1 increased by 0.421 mmHg, 0.589 mmHg and 0.340 mmHg, respectively. Meanwhile, we detected that the ΔMAP 1 increased by 0.973 mmHg and 1.449 mmHg in IFG and diabetic patients compare to NFG individuals. Furthermore, in strati ed analyses, we found that the above associations were stronger in youngers (age < 60 years old), females, hypertensives, non-diabetics, non-current smokers and non-current drinkers than those in the corresponding subgroups. Our results supported that blood glucose could be an independent risk factor of SSBP.
Although the exact mechanisms underlying the relationship of the blood glucose with SSBP is unclear, several researches suggested it may be linked to insulin resistance. Insulin resistance is associated with hyperinsulinemia and hyperglycemia, under insulin resistance, the target cells fail to respond to ordinary levels of circulating insulin thus higher concentrations of insulin are required for a normal response [35]. Meanwhile, the impairment of blood glucose uptake in muscle and an increased gluconeogenesis by the liver resulting in hyperglycemia [36]. Hyperglycemia stimulates the reabsorption of sodium. In general, kidneys reabsorb the same amount of blood glucose as they lter each day as to prevent valuable energy from being lost in the urine. Most of the capacity for renal glucose reabsorption is provided by sodium glucose cotransporter (SGLT) 2 in proximal tubule. However, hyperglycemia could enhance the glucose ltration and increase capacity of glucose/sodium reabsorption [22]. SGLT2 inhibitors, a kind of hypoglycemic drugs, could suppress the cotransport of glucose coupled with sodium and signi cantly attenuated the high salt-induced elevation of BP [37]. Therefore, many studies performed a series of analyses to uncover the relationships between blood glucose and/or insulin resistance with SSBP.
Studies of the associations between salt sensitivity and insulin resistance have yielded contradictory results. Maaten et al. [38] and Dengel et al. [39] supported that insulin resistance was negatively correlated with salt sensitivity, but Bigazzi et al. [40] and Giner et al. [41] observed contrary results. The reasons for these apparent discrepancies are not exactly known but may be related to differences in study populations and study methods. Previous studies focusing on the association between FBG and SSBP obtained consistent results. In animal study, high sucrose diets could increase BP of SS rats [42], and the moderate fructose-enriched diet also stimulates salt-sensitive hypertension in rats [43]. Somova et al. observed Dahl salt-sensitive rats signi cantly decreased blood glucose utilization and clearance [44]. Ilhami et al. clari ed that basal blood glucose level was signi cantly higher in SS than in SR rats [45]. In human, Sharma et al. [16,19], Egan et al. [17] and Galletti et al. [18] uncovered that the FBG was higher in SS individuals than in SR group. This study included adequate samples (n = 2051) and found that the FBG level in SS patients was signi cantly higher than that in SR individuals. Our analysis supported the previous ndings and provided clues to the positive correlation between blood glucose and SSBP.
SSBP is also reported to be elevated in patients with diabetes [13]. The current study observed the prevalence of diabetes in SS patients signi cantly higher than that in SR individuals. However, we considered that using blood glucose as a risk factor of SSBP and as the basic of preventive strategies for SS is of greater clinical signi cance than diabetes, since hyperglycemia plays key role in the genesis of SS in patients with type 2 diabetes [22]. And, we found evidence of signi cant positive associations between FBG with prevalence of SS (OR = 1.140) or with ΔBP 1 (β for ΔMAP 1 , ΔSBP 1 and ΔDBP 1 = 0.421, 0.589 and 0.340). Our results are consistent with previous reports [40,46] and clari ed that blood glucose level could be an independent risk factor for SS.
The present study further demonstrated that there were positive dose-response associations between blood glucose with SS prevalence or with ΔBP 1 in 2051 participants. For every IQR increase in FBG, the SS prevalence and ΔBP 1 signi cantly increased for a trend. It was worth noting that compared to participants with FBG < 5.00 mmol/L (Q 1 ), both the SS prevalence and the ΔBP 1 showed signi cantly elevated in the subjects with 5.44 ≤ FBG < 6.19 mmol/L (Q 3 ) and FBG > 6.19 mmol/L(Q 4 ). And the value of 5.44 mmol/L is even slightly below the diagnostic criteria for IFG of 6.11 mmol/L. Consistently, the ΔBP 1 of IFG participants also signi cantly increased when compared with NFG individuals. These results suggested that relatively higher glucose, though not diagnosed as diabetic, could also increase the SSBP.
We highlighted that controlling the elevation of blood glucose in the early stage might be much more important for preventing SS. Our results need to be validated in more larger population association studies.
We further analyzed the associations between FBG with SSBP after strati ed participants according to variables including sex, age, obesity, hypertension, diabetes, smoking and drinking to determine the sensitive population. Our results suggested that the effects of blood glucose on SSBP were a little different in population with different characters. The associations of FBG with SSBP in youngers (age < 60 years old), females, hypertensives, non-diabetics, non-current smokers and non-current drinkers were more signi cant than the corresponding subgroups, which suggested that these sensitive population should pay more attention to the effect of blood glucose on SS.
Some strengths and limitations of the current study should be acknowledged. This is the rst epidemiologic study based on general population to focus on the associations between blood glucose with the SS or with the BP changes during acute salt load period and diuresis shrinkage period. This study included a large sample size and meticulously controlled conditions, which made the result of statistical analysis more persuasive. Our results uncovered the positive dose-response association between blood glucose and SSBP in population and highlighted that controlling the elevation of blood glucose in the early stage might be much more important for preventing SS. Furthermore, we performed strati ed analysis and found the role of blood glucose as an independent risk factor for SS, especially in youngers, females, hypertensives, non-diabetics, non-current smokers and non-current drinkers. Some scholars claimed that acute salt loading has adverse cardiovascular effects [47], therefore we developed a set of strict inclusion criteria for study subjects, and there was no side effect occurred during saline loading. The limitations are as follows, although the methods for determining SSBP are not uniform at present, the dietary intervention methods are more accurate than the acute saline load test methods; this is a cross-sectional study and lack of the ability for causal inference analyses like prospective cohort studies, so the causal association between blood glucose and salt sensitivity is not yet available; blood glucose could affect by renal function and insulin resistance, but we didn't take them into concern due to lack of data; participants were all from two cities in northern China, which may affect the extrapolation of results.

Conclusions
In conclusion, our ndings suggest that elevated blood glucose is an independent, dose-dependent risk factor for salt sensitivity of blood pressure, especially in youngers (age<60 years old), females, hypertensives, non-diabetics, non-current smokers and non-current drinkers. In addition, relatively higher blood glucose, though not diagnosed as diabetic, might contributed to the salt sensitivity of blood pressure, which highlighted that controlling the elevation of blood glucose in the early stage might be much more important for preventing salt sensitive.

Declarations
Ethics approval and consent to participate: This study was conducted in accordance with the Declaration of Helsinki and was approved by the Ethics Committee of Capital Medical University, Beijing, China.
Written informed consent was obtained from all subjects before their enrollment in this study.
Consent for publication: Not applicable.
Availability of data and materials: The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.
Competing Interest: The authors have no con icts of interest to declare.

Figure 2
The dose-response associations between fasting blood glucose and salt sensitivity of blood pressure. (A), fasting blood glucose and SS; (B), fasting blood glucose and ΔMAP1; (C), fasting blood glucose and ΔSBP1; (D), fasting blood glucose and ΔDBP1. The spline regression model was adjusted by age, sex, sleep, smoking, drinking, triglycerides, low density lipoprotein cholesterol and baseline mean arterial pressure. SS, salt sensitive; ΔMAP1, change of mean arterial pressure during saline loading; ΔSBP1, the change of systolic blood pressure during saline loading; ΔDBP1, the change of diastolic blood pressure during saline loading.

Figure 3
The association between FBG (per IQR increase) with SS or ΔBP1 in different subgroups. (A), fasting blood glucose and SS; (B), fasting blood glucose and ΔMAP1; (C), fasting blood glucose and ΔSBP1; (D), fasting blood glucose and ΔDBP1. FBG, fasting blood glucose; SS, salt sensitive; ΔBP1, change of blood pressure due to saline loading; ΔMAP1, change of mean arterial pressure due to saline loading; ΔSBP1, change of systolic blood pressure due to saline loading; ΔDBP1, change of diastolic blood pressure due to saline loading. Statistical analysis by multiple logistic regression analyses (A) and multivariable linear regression analyses (B, C and D). The model was adjusted by age, sex, sleep, smoking, drinking,