We searched the NHANES database for 5 periods of data from 2009 to 2018 for research. Of the 49,693 potential participants in the study, 42,708 were excluded for the following reasons: missing UA data (n = 18,154), missing systolic blood pressure (SBP) and diastolic blood pressure (DBP) data (n = 1,309), without hypertension and missing hypertension data (n = 21,049), missing data of medicine for hypertension (n = 1,202), and age = 80 (n = 994). The reason why people (=80 years old) were excluded was because NHANES defined all the people (≥ 80 years old) as 80 years old. In the end, 6,985 people were included in the study, and they were divided into two groups: the hypertension treatment group (n = 5,983) and hypertension non-treatment group (n = 1,002; Figure 1).

In this study, UA was the exposed variable, and BP was the primary outcome including systolic blood pressure (SBP) and diastolic blood pressure (DBP). After the participants sat down with their feet supporting the ground, they rested quietly for at least 5 min, and three blood pressure readings were obtained with a mercury sphygmomanometer. Each reading was separated by 30 s. If one or more previous readings were interrupted, the fourth reading was obtained. SBP and DBP were made up of the average of all available measurement data. Race was divided into four categories: Mexican-American, White, Black, and Other. Age was divided into four groups, which included <18, 18–44, 45–59, and ≥ 60. Body mass index (BMI) was divided into five groups, which were underweight (< 18.5 kg/m²), normal weight (18.5~24.99 kg/m²), overweight (25~29.99 kg/m²), obesity (≥ 30 kg/m²), and not recorded. Alcohol consumption was divided into three groups based on the question: “In the past 12 months, how often did you drink any type of alcoholic beverage?” The groups were composed of drinking, no drinking, and not recorded. Diabetes was divided into four groups based on the question, “Have you ever been told by a doctor or health professional that you have diabetes or sugar diabetes?”, and were yes, no, borderline, and not recorded. Smoking was divided into three groups based on the question, “Do you now smoke cigarettes?”, and were smoking, no smoking, and not recorded. Glomerular filtration rate (GFR) was estimated by the simplified Modification of Diet in Renal Disease (MDRD) equation 186 × SC^−1.154 × Age^−0.203× (0.742 if female) (10). The method of obtaining other variables, such as UA, creatinine (CR), glucose (GLU), hemoglobin (HGB), total cholesterol (TC), low-density lipoprotein (LDL), and high-density lipoprotein (HDL), can be found at www.cdc.gov/nchs/nhanes/.

The associations between UA and BP including SBP and DBP were explored by applying a multiple linear regression model. Mean ± SD (normal distribution) or median (quartile) (skew distribution) was applied to represent continuous variables, including CR, GLU, HGB (missing data n = 18), HDL (missing data n = 1), TC (missing data n = 1), GFR, SBP, DBP, UA, and LDL (missing data n = 3,634). Percentage or frequency was utilized to represent categorical variables, including gender, race, alcohol consumption (missing proportion 19.44%), diabetes (missing proportion 0.06%), smoking (missing proportion 50.15%), age, and BMI (missing proportion 1.43%). The missing data of HGB, HDL, and TC were filled in with the median, and the missing data of LDL, alcohol consumption, diabetes, smoking, and BMI were included in the data analysis separately as a group. Smooth curve fittings were used to explore the non-linear relationship between UA and SBP or DBP. We also conducted a log-likelihood ratio test of the one-line (non-segmented) model with the piecewise regression model to determine whether the threshold exists. Statistical software packages R (http://www.R-project.org) and EmpowerStats (http://www. empowerstats. com, X&Y Solutions, Inc.) were used for the data analyses. Statistical significance was based on the value of p < 0.05.

A total of 6,985 participants were included in our study based on the exclusion criteria. Among them, there were 5,983 people in the hypertension treatment group and 1,002 people in the hypertension non-treatment group (Table 1). In the hypertension treatment group, the percentage of men and women was 47.38 and 52.62%, respectively. With regard to ethnicity, 11.88% were Mexican-American, 37.66% were White, and 29.82% were Black. Overall, the mean (SD) values for CR, GLU, HGB, HDL, TC, GFR, SBP, DBP, UA, and LDL were 1.02 (0.72) mg/dl, 6.44 (2.76) mmol/L, 13.78 (1.52) mg/dl, 1.34 (0.43) mmol/L, 4.88 (1.1) mmol/L, 70.24 (24.73) ml/min/1.73 m², 132.75 (19.15) mmHg, 71.9 (12.62) mmHg, 5.89 (1.56) mg/dl, and 2.8 (0.93) mmol/L, respectively. Among the participants, 38.79% were <60 years old, 13.27% were of normal weight, 31.14% had diabetes, 17.48% were smokers, and 56.11% were alcohol drinkers. In the hypertension non-treatment group, percentage of males and females was 51 and 49%, respectively. With regard to ethnicity, 15.47% were Mexican-American, 36.33% were White, and 26.75% were Black. Overall, the mean (SD) values for CR, GLU, HGB, HDL, TC, GFR, SBP, DBP, UA, and LDL were 0.93 (0.62) mg/dl, 6.06 (2.98) mmol/L, 14.04 (1.75) mg/dl, 1.34 (0.43) mmol/L, 5.16 (1.11) mmol/L, 80.28 (27.19) ml/min/1.73 m², 135.69 (22.12) mmHg, 78.2 (13.88) mmHg, 5.54(1.44) mg/dl, and 3.11 (0.94) mmol/L, respectively. Among the participants, 71.86% were <60 years old, 16.77% were of normal weight, 14.57% had diabetes, 29.44% were smokers, and 65.47% were alcohol drinkers. The Univariate analysis for SBP and DBP is demonstrated in Supplementary Tables 1, 2.

We analyzed the independent effect of UA on SBP and DBP with three models (Table 2). In model I, in the hypertension treatment group, a significantly negative association can be found in SBP (β, −0.49 [95% CI, −0.8, −0.18]) and DBP (β, −0.29 [95% CI, −0.5, −0.09]). In the hypertension non-treatment group, a significantly positive association can be found in SBP (β, 1.6 [95% CI, 0.66, 2.55]) and DBP (β, 1.4 [95% CI, 0.81, 2]). In model II, gender, race, and age were adjusted. In the hypertension treatment group, the relationship still existed in SBP (β, −0.59 [95% CI, −0.9, −0.27]) and DBP(β, −0.48 [95% CI, −0.68, −0.28]). In the hypertension non-treatment group, the relationship also still existed in SBP (β, 1.21 [95% CI, 0.17, 2.25]) and DBP (β, 1.2 [95% CI, 0.55, 1.85]). In model III, in which all the potential confounding factors were adjusted for, the significantly negative association was discovered in SBP (β, −0.36 [95% CI, −0.71, −0.01]) and DBP (β, −0.47 [95% CI, −0.69, −0.26]) in the hypertension treatment group. In the hypertension non-treatment group, the negative relationship was found not to exist in SBP (β, 1.12 [95%CI, −0.02, 2.26]) and DBP (β, 0.52 [95% CI, −0.19, 1.22]).

A smooth curve fitting was used to explore the non-linear relationship between UA and SBP and DBP. From Figures 2, 3, the associations between UA and BP including SBP and DBP which were both a linear relationship basically were found in the hypertension treatment group. As the level of UA increased, SBP and DBP showed a gradual decrease. However, in the hypertension non-treatment group, the associations between UA and BP including SBP and DBP were both an inverted U-shape. So, a two-piece linear regression model was utilized to calculate the inflection point, and the inflection point of SBP and DBP was 7 and 7.5 mg/dl, respectively (Table 3). For SBP, the association was positively significant (β, 3.11 [95% CI, 1.67, 4.56]) before the inflection point of 7 mg/dl. However, after the inflection point of 7 mg/dl, the association was negative (β, −5.44 [95% CI, −8.60, −2.28]). For DBP, the inflection point was 7.5 mg/dl, and the effect size was positive (β, 1.19 [95% CI,0.37, 2.01]) before the inflection point. However, after it, the effect size was negative (β, −3.24 [95% CI, −5.72, −0.76]). We also conducted a smooth curve fitting by gender in the hypertension treatment group and the non-treatment group. From Supplementary Figures 1, 2, the negative association can still be seen in the hypertension treatment group stratified by gender. From Supplementary Figures 3, 4, the inverted U-shaped relationship also existed in the hypertension non-treatment group. With the exception of females, the association between UA and DBP was positive.