The NHANES is a series of nationwide cross-sectional surveys on the health and nutrition of the US population conducted by the National Center for Health Statistics (NCHS) in a 2-year cycle. This study was approved by the NCHS Research Ethics Review Board, and all participants provided written informed consent. More detailed information on the NHANES can be found on the official website: www.cdc.gov/nchs/nhanes/.

The current study used data from the 2009 to 2010 NHANES cycles. A total of 5,793 participants aged 20-79 years were recruited. Of these participants, 2,325 had available data on plasma TFAs biomarkers and blood pressure measurements. Participants who were pregnant (n = 26) or had incomplete covariate data (n = 329) were excluded. A total of 1,970 participants were included in this study ( Supplementary Figure S1 ).

The four TFAs were analyzed in plasma using isotope dilution-gas chromatography-negative chemical ionization-mass spectrometry (ID-GC-NCI-MS), as detailed in a previous study (18). These TFAs include trans-9-hexadecenoic acid (palmitelaidic acid, C16:1n-7t), trans-9-octadecenoic acid (elaidic acid, C18:1n-9t), trans-11-octadecenoic acid (vaccenic acid, C18:1n-7t), and trans-9, trans-12-octadecadienoic acid (linolelaidic acid, C18:2n-6t, 9t), and are the most prevalent TFAs in foods, accounting for about 40–60% of total TFAs present in the human body (19).

Blood pressure measurements were performed by trained examiners using a mercury sphygmomanometer, following established protocols. After 5 min of rest in a seated position, the participant’s maximum inflation level was established, and blood pressure was subsequently measured on the same arm three times consecutively. Average systolic and diastolic blood pressures (SBP and DBP, respectively) were calculated.

Based on the 2017 American College of Cardiology (ACC)/American Heart Association (AHA) blood pressure guideline (7), hypertension was defined as the average SBP being ≥130 mm Hg, or the average DBP ≥80 mm Hg. If a participant answered “yes” to the question ‘Are you now taking prescribed medicine for high blood pressure?’ or self-reported hypertension was considered hypertensive.

The potential confounders for this study were selected a priori from previous studies on TFAs and hypertension (20). This study collected data on demographic variables such as age, sex, race/ethnicity, education level, and poverty income ratio (PIR) through household interviews and questionnaires. Educational level was categorized as follows: less than high school, high school graduate or equivalent, and college or above. The PIR, an index of the ratio of family income to poverty, was calculated by dividing family income by poverty guidelines specific to family size (21). The participants’ weights and heights were obtained from examination data. Body mass index (BMI) was calculated as weight in kilograms divided by height in meters squared (< 25 kg/m², 25–30 kg/m², and ≥30 kg/m²). Data on smoking status (smoking at least 100 cigarettes in life or not), alcohol consumption (at least 12 alcohol drinks/1 year or not), and recreational physical activity (vigorous, moderate, or no activity) were acquired from the questionnaire data. Diabetes was defined as meeting one of the following four conditions: a) glycated hemoglobin > 6.5%; b) fasting blood glucose >126 mg/dL; c) 2 h plasma glucose ≥200 mg/dL, and d) a self-report of a diagnosis by a physician or health care professional (22). The estimated glomerular filtration rate (eGFR) was calculated using new creatinine- and cystatin C–based equations (23). We defined CKD as either individuals developing an eGFR ≤60 mL/min/1.73 m² or urine albumin-creatinine ratio ≥30 mg/g. A history of atherosclerotic cardiovascular disease (ASCVD) was defined as at least one diagnosis of coronary heart disease, angina, heart attack, or stroke.

Participants’ demographic characteristics were presented with the mean ± standard deviation (SD) for continuous variables, and counts (percentages) for categorical variables. We used the geometric mean (GM), geometric standard deviations (GSD), and percentiles to characterize the distributions of plasma TFA concentrations. TFA concentrations were natural log (ln)-transformed to approximate a normal distribution. Pearson’s correlation was used to assess the correlation between pairs of ln-transformed TFA concentrations.

We initially used the restricted cubic spline (RCS) model with three knots at the 10^th, 50^th, and 90^th percentiles to evaluate the non-linear associations between plasma TFA concentrations and SBP and DBP. We used multivariate linear regression models to examine the association between ln-transformed TFA concentrations (continuous variables) and SBP and DBP. Given that a few associations were non-linear (P-value in the RCS < 0.10, as shown in Supplementary Figure S2 ), we categorized plasma TFA concentrations by quartiles to evaluate the nonmonotonic associations of plasma TFA concentrations with SBP and DBP in multivariate linear regression models. Hypertension has a high prevalence in U.S. adults (approximately 50%) (5); therefore, we used a modified Poisson regression model with robust variance estimates to evaluate the association between TFA exposure and the risk of hypertension (24).

We used the BKMR model to evaluate the single and overall effects of exposure to the four TFAs on blood pressure and hypertension. This approach combines Bayesian and statistical learning methods to iteratively regress the response variable on the non-parametric terms of multiple exposure mixture components, allowing for potential non-linear and non-additive effects (25). The base BKMR model was used to evaluate the associations of the four TFAs with the continuous outcomes of blood pressure measurements, whereas the BKMR-P model with a probit link function was used to examine the effect of the four TFAs on hypertension defined by clinical threshold values. All BKMR models were constructed using a Markov Chain Monte Carlo (MCMC) with 10,000 iterations. The posterior inclusion probabilities (PIPs) were calculated to indicate the highest TFA within the mixture. A PIP value of ≥ 0.5 was considered the threshold value (26).

The overall associations of the TFAs mixture with SBP and DBP and the risk of hypertension were visualized by depicting the differences (95% credible intervals, CrI) in SBP, DBP and hypertension, holding the plasma concentrations of all TFAs at the same percentiles (30^th, 35^th, 40^th, 45^th, 50^th, 55^th, 60^th, 65^th, 70^th, and 75^th percentiles) compared to those at the 25^th percentile. The single effects of TFAs on SBP, DBP, and hypertension were presented as the estimated changes (95% CrI) in these outcomes for the changes in each TFA concentration between the 25^th and 75^th, while simultaneously fixing the other TFAs at the 25^th, 50^th, and 75^th percentiles.

We performed several sensitivity analyses. Considering the complexity of the sampling design, we applied survey-weighted linear regression analyses to evaluate the associations of TFAs with SBP and DBP to create nationally representative estimates. Secondly, it is reported that plasma TFA concentrations decreased by 54% from 1999-2000 to 2009-2010 (16). We repeated the main analyses among the population in 1999-2000, which facilitated to understand some patterns of the effects of TFA exposure on blood pressure in the light of different levels. We additionally adjusted the associations for ASCVD history to test the robustness of the results.

SAS 9.4 (SAS Institute Inc., Cary, NC, USA) and R 4.2.2 (R Development Core Team) were used for statistical analyses. The BKMR was implemented with the R packages “bkmr” and “ggplot2”. A P-value of < 0.05 is considered statistically significant.

The demographic characteristics of the 1,970 US adults are presented in Table 1 . In the total population, the prevalence of hypertension was 48.68%, with an average age of 47.94 ± 16.37 years. Approximately half of the participants were males (50.02%), non-Hispanic white (49.34%), had a PIR greater than 1.85 (48.58%), and reported no recreational activities (51.52%). Most adults were overweight or obese (73.05%) and had some college education or higher level of education (73.71%). Among the participants, 45.84% were smokers, 74.47% were alcohol drinkers, and ≥ 10% had diabetes (11.32%) and CKD (12.49%). Except for PIR, significant differences were noted between the hypertensive and non-hypertensive adults in terms of age, sex, race/ethnicity, BMI, education level, recreational activities, smoking, drinking, diabetes, and CKD (all P-value < 0.05).

Table 2 presents the distributions of the plasma TFA concentrations. All four TFAs were detected in plasma samples of US adults. Vaccenic acid had the highest median (interquartile range) concentration (18.10 [13.30–24.90] µmol/L), followed by elaidic, palmitelaidic, and linolelaidic acids. Plasma concentrations of four TFAs were highly correlated with each other (r = 0.60–0.86, P-value < 0.001, Supplementary Figure S3 ).

ln-transformed TFA concentrations were associated with increased SBP and DBP ( Table 3 ). Specifically, for per unit increase in ln-transformed plasma concentrations of palmitelaidic, elaidic, vaccenic, and linolelaidic acids, SBP increased by 2.94 (95% confidence interval (CI): 1.27, 4.60), 3.60 (95% CI: 2.21, 4.99), 2.46 (95% CI: 1.07, 3.84), and 4.78 (95% CI: 3.10, 6.46) mm Hg, respectively, and DBP increased by 2.77 (95% CI: 1.53, 4.01), 2.35 (95% CI: 1.32, 3.38), 2.03 (95% CI: 1.00, 3.06), and 3.70 (95% CI: 2.45, 4.95) mm Hg, respectively. Estimates for TFAs, modeled as quartiles, showed similar patterns of positive associations. Higher exposures to palmitelaidic, elaidic, vaccenic, and linolelaidic acids were associated with increased SBP and DBP. Moreover, the highest quartiles carried the highest risk with estimates ranging 2.39–6.01 ( Table 3 ).

Figures 1 and 2 show the overall associations of the four TFAs with SBP and DBP, respectively. The BKMR model indicated that higher plasma concentrations of the TFA mixtures were associated with increased SBP and DBP. Both SBP and DBP increased when all TFAs were fixed at their 30^th percentile or above, compared to the 25^th percentile of the TFAs mixtures. For example, in comparison to the 25^th percentile, the 75^th percentile of the TFAs mixtures was associated with a 3.00 (95% CrI: 1.94, 4.06) mm Hg increase in SBP, and a 2.49 (95% CrI: 1.70, 3.29) mm Hg increase in DBP.

Regarding the single effects of the four TFAs on SBP and DBP, the PIP results for each TFA on SBP and DBP are shown in Supplementary Table S1 . Linolelaidic acid had the highest PIP (PIP_SBP: 0.96, PIP_DBP: 1.00), indicating that linolelaidic acid contributed the most to the increased SBP and DBP ( Figures 1 , 2 ). An interquartile range (IQR) increase in linolelaidic acid concentration was associated with an increase of 2.15 (95% CrI: 0.69, 3.61), 2.23 (95% CrI: 0.81, 3.65), and 2.31 (95% CrI: 0.87, 3.75) in SBP, and an increase of 2.20 (95% CrI: 1.10, 3.29), 2.24 (95% CrI: 1.18, 3.30), and 2.29 (95% CrI: 1.21, 3.37) in DBP when the other TFA concentrations were simultaneously fixed at their 25^th, 50^th, or 75^th percentile, respectively. We did not observe significant single effects for the other three TFAs because of their lower PIP values ( Supplementary Table S1 ).

We found that the slopes of the dose-response curves of certain TFAs were similar at different percentiles of other TFAs, with others fixed at their middle levels ( Supplementary Figure S4 , S5 ), indicating there were no interactions among the four TFAs.

After adjusting for covariates, higher plasma TFA concentrations were associated with a higher risk of hypertension, although statistically significant associations were only observed for vaccenic and linolelaidic acids with hypertension (vaccenic acid: RR = 1.16, 95% CI: 1.07, 1.25; linolelaidic acid: RR = 1.23, 95% CI: 1.12, 1.36, Table 4 ). Additionally, the highest quartiles of elaidic and linolelaidic acid concentrations were associated with the highest risk ( Table 4 ). Adults with plasma concentrations of elaidic acid and linolelaidic acid in the highest quartile had 1.23-fold (95% CI:1.08, 1.39) and 1.34-fold (95% CI:1.18, 1.53) higher risks of hypertension than those in the first quartile.

The BKMR model showed a significant overall association of the TFAs mixture with an increased risk of hypertension when all TFAs were at or above the 30^th percentile compared to their 25^th percentile ( Figure 3 ). Furthermore, consistent with the results of the continuous blood pressure measurement, the BKMR model also identified linolelaidic acid (PIP=0.93) as the most important contributor to an increased risk of hypertension ( Figure 3 ).

Similar positive associations of the four TFAs with SBP and DBP were also observed when we used a survey-weighted linear regression model ( Supplementary Table S2 ). When we repeated the analyses among the population in 1999-2000, the pattern of association between TFAs and blood pressure did not differ by cycles ( Supplementary Figure S6 - S8 ). Moreover, linolelaidic acid also contributed most to the observed associations. Similar results were observed for the analyses when we additionally adjusted for ASCVD history ( Supplementary Tables S3 , S4 ).