The National Health and Nutrition Examination Survey (NHANES) is an ongoing survey approved by the Institutional Review Boards of the National Center for Health Statistics and the Centers for Disease Control and Prevention. NHANES utilizes a complex survey methodology to provide a nationally representative sample of the U.S. civilian population. Linked with the National Death Index (NDI), the database enables prospective mortality follow-up^[15]. Written informed consent was obtained from all participants under protocols #2005-06 (2005-2010) and #2011-17 (2011-2014). The initial population included 6,507 hypertensive adults (aged ≥ 20 years) from the 2005-2014 survey cycles, all of whom underwent urinary nitrate testing. After excluding participants with missing data on BP, comorbidities and follow-up, the final analytical sample comprised 6,130 individuals. Hypertension was defined as meeting at least one of the following criteria: self-reported physician diagnosis, use of ≥ 2 antihypertensive medications (including combination preparations), or a mean systolic blood pressure (SBP) ≥ 140 mmHg or mean diastolic blood pressure (DBP) ≥ 90 mmHg (based on the average of three measurements). This threshold is consistent with the Joint National Committee (JNC) 7 guidelines, thereby aligning the analysis with the clinical standards contemporary to the survey period. Mortality data were obtained through prospective follow-up by linking participant records to the NDI and official death certificates.

Urinary nitrate concentrations were measured in spot urine samples. Urine specimens were properly stored at -20 °C until shipment to the National Center for Environmental Health for testing. Nitrate levels were quantified by ion chromatography with tandem mass spectrometry (IC-MS/MS). Detailed protocols for specimen collection and processing are described in the NHANES Laboratory/Medical Technologists Procedures Manual^[16].

In this study, primary outcomes consisted of BP and cardiovascular mortality. Secondary outcomes were the prevalence of CVD, including angina pectoris, myocardial infarction (MI), coronary heart disease (CHD) and heart failure (HF). Mortality status was determined through linkage with the NDI for the 1999-2014 NHANES cohorts, providing follow-up data through December 31, 2015. Follow-up duration was defined as the interval from survey participation to either the date of death or December 31, 2015. The cause-specific death categories were based on the ICD-10 codes, with cardiovascular death codes including I00-I09, I11, I13, I20-I51 and I60-I69. For a more detailed description of the methodology, please refer to the online source^[17].

Demographic and physical examination data were collected, including sex, age, race (Mexican American, Other Hispanic, Non-Hispanic White, Non-Hispanic Black and Other Race), poverty income ratio (PIR), body mass index (BMI), SBP, and DBP. Smoking and drinking habits were categorized as current, former and never. Dietary factors, such as the intake of processed meat, sodium, and dark-green vegetables, were adjusted for as covariates. The major comorbidities included angina pectoris, MI, CHD, HF, stroke, diabetes and cancer. Diabetes was defined as meeting at least one of the following criteria: a self-reported diagnosis; glycated hemoglobin (HbA1c) ≥ 6.5%; fasting plasma glucose ≥ 7.0 mmol/L; random plasma glucose ≥ 11.1 mmol/L; 2-h oral glucose tolerance test (OGTT) glucose ≥ 11.1 mmol/L; or use of glucose-lowering medications. Self-reported information was used to determine the presence of other medical conditions. Medication history (antihypertensive, glucose-lowering, and lipid-lowering medications) and biochemical parameters [HbA1c, total cholesterol, and estimated glomerular filtration rate (eGFR)] were assessed.

The study population was stratified by urinary nitrate tertiles to describe baseline characteristics. Continuous variables are presented as weighted means (standard errors), while categorical variables are reported as unweighted counts (weighted percentages). Generalized additive models (GAMs) were employed to visualize the dose-response relationships between urinary nitrate and blood pressure. Linear association was analyzed using standard linear regression, whereas potential non-linearity was explored using two-piecewise linear regression. The optimal turning point was identified through a grid search to maximize the likelihood function. A log-likelihood ratio test was subsequently performed to compare the linear model with the piecewise model, with P < 0.05 indicating the presence of a threshold effect. Logistic regression and Cox proportional hazards models were utilized to evaluate the associations of urinary nitrate with CVD prevalence and cardiovascular mortality, respectively. Furthermore, subgroup and mediation analyses were conducted to explore potential interactions and mediating pathways. Mediation analysis was performed within a causal mediation framework to evaluate the mediating effects of mediators on the association between urinary nitrate and cardiovascular outcomes. The “mediation” package in R was employed. All analyses followed NHANES data analysis guidelines, incorporating sample weights. Given that urinary nitrate was measured in a specific laboratory sub-sample, the corresponding sub-sample weights were applied to ensure national representativeness. Missing values were addressed via multiple imputation. All statistical analyses were performed using R software (version 4.5.2; R Foundation for Statistical Computing, Vienna, Austria) and EmpowerStats (X&Y Solutions, Inc., Boston, MA). A two-sided P < 0.05 was considered statistically significant.

A total of 6,130 participants were included in the final analysis [Figure 1]. The baseline characteristics are summarized in Table 1. The mean (standard deviation, SD) age was 56.85 (15.12) years, and 50.5% were men. Non-Hispanic Whites predominated, accounting for 71.3% of the population. The mean (SD) urinary nitrate concentration was 5.09 (4.34) mg/dL. Participants were stratified into tertiles based on urinary nitrate levels (T1, ≤ 2.82 mg/dL; T2, 2.82-5.52 mg/dL; T3, > 5.52 mg/dL). The T1 group had the lowest proportion of males. As urinary nitrate levels increased, upward trends were observed in the prevalence of alcohol consumption and tobacco use, alongside elevations in BMI and eGFR. Conversely, the mean age, SBP, and the proportion of participants with HF or stroke, as well as those using antihypertensive or lipid-lowering agents, decreased across the tertiles. The T3 group exhibited the lowest prevalence of stroke and cancer. No significant differences were observed in the proportions of CHD, MI, or diabetes among the three groups.

As illustrated in Figure 2A, a nonlinear association between urinary nitrate and SBP was observed. After multivariable adjustment, a saturation effect was identified, with an optimal turning point of 4.40 mg/dL [95% confidence interval (95%CI), 3.68-5.25; log-likelihood ratio test P = 0.029]. Within the range of 0-4.40 mg/dL, each 1 mg/dL increase in urinary nitrate was associated with a 0.60 mmHg decrease in SBP [95%CI, (-0.99)-(-0.21); P = 0.003; Table 2]. However, this association was no longer significant when urinary nitrate levels exceeded 4.40 mg/dL. In contrast, while Figure 2B suggested a linear trend between urinary nitrate and DBP, the fully adjusted model showed no significant association (β = -0.02, 95%CI, -0.10-0.05; P = 0.507; Table 3).

As shown in Table 1, among the 6,130 participants, the prevalence of angina pectoris, MI, CHD, and HF was 4.3%, 6.7%, 6.7%, and 4.9%, respectively. Multivariable logistic regression analysis [Table 4] revealed that urinary nitrate levels were significantly and inversely associated with HF prevalence. The unadjusted, partially adjusted and fully adjusted odds ratios (ORs) with 95%CIs were 0.880 (95%CI, 0.846-0.916; P < 0.001), 0.902 (95%CI, 0.866-0.940; P < 0.001) and 0.942 (95%CI, 0.905-0.981; P = 0.004), respectively. A similar downward trend in the prevalence of angina pectoris was observed with increasing urinary nitrate, although it did not reach statistical significance (fully adjusted OR, 0.989; 95%CI, 0.939-1.041; P = 0.672). No significant associations were found between urinary nitrate and the prevalence of MI (OR, 1.017; 95%CI, 0.984-1.051; P = 0.315) or CHD (OR, 1.016; 95%CI, 0.982-1.051; P = 0.363).

During a weighted median follow-up of 65.0 months (interquartile range: 37.0-97.0 months), 214 participants (weighted proportion: 2.0%) died from cardiovascular disease. The Kaplan-Meier curves [Figure 3] demonstrated that elevated urinary nitrate was significantly associated with superior cardiovascular-specific survival (Log-rank test, P < 0.001). Consistently, multivariable-adjusted Cox proportional hazards models showed that each unit increase in urinary nitrate was associated with a 7.9% reduction in cardiovascular mortality risk [hazard ratio (HR), 0.921; 95%CI, 0.867-0.977; P = 0.006; Table 5].

Mediation analysis was performed to explore whether the association between urinary nitrate and cardiovascular mortality was mediated by HF or SBP. As shown in Table 6, the average causal mediation effect (ACME) for HF was 1.925 (95%CI, -14.947-20.318; P = 0.836), accounting for only 0.6% of the total effect. Similarly, the ACME for SBP was 2.582 (95%CI, -1.173-9.927; P = 0.244), with a mediated proportion of 1.1%. In contrast, the average direct effect (ADE) remained significant in both models (P < 0.05), suggesting that the favorable cardiovascular outcomes associated with higher urinary nitrate were independent of HF and SBP.

Results of subgroup analyses for the association of urinary nitrate with cardiovascular mortality are shown in Figure 4. A significant interaction was observed for history of CHD [P-interaction < 0.001; false discovery rate (FDR)-adjusted P < 0.001]. The protective association was more pronounced in participants without CHD (HR, 0.849; 95%CI, 0.802-0.898), while no such association was observed in those with CHD (HR, 1.063; 95%CI, 0.991-1.140). Additionally, while a nominal interaction was noted for smoking status (P-interaction = 0.045), it did not retain statistical significance after FDR adjustment (P = 0.202). The inverse association remained largely consistent across most subgroups, including age, sex, BMI, history of diabetes or stroke, history of HF, and antihypertensive medication (all P-interaction > 0.05). It is important to note that some subgroups were characterized by small sample sizes and imbalanced distributions. These results should be interpreted with caution due to the potentially limited statistical power in these specific strata.

This study has several limitations that warrant consideration. First, the cross-sectional design of the NHANES database precludes the establishment of definitive causal relationships between urinary nitrate levels, blood pressure, and cardiovascular outcomes. While robust associations were observed, the lack of temporal sequencing between exposure and outcome leaves the possibility of reverse causality unresolved. Therefore, these findings should be further validated through large-scale prospective cohort studies.

Second, the reliance on a single spot urine sample to assess nitrate exposure represents an inherent limitation. Urinary nitrate concentrations are sensitive to recent dietary intake and exhibit significant diurnal fluctuations. Thus, a single measurement may not accurately reflect chronic exposure, potentially leading to non-differential misclassification of the exposure status. From a statistical perspective, such random measurement error typically attenuates effect estimates (bias toward the null). Therefore, the inverse associations between urinary nitrate and cardiovascular mortality observed in this study may be underestimated. Future research employing repeated measurements or 24 h urine collections would provide a more robust assessment.

Third, the ascertainment of CVD history relied primarily on self-reported questionnaires rather than independent clinical adjudication or comprehensive medical record reviews. This may introduce recall bias or social desirability bias, which could affect the accuracy of our prevalence estimates.

Fourth, although our mediation analysis suggested that urinary nitrate may influence cardiovascular mortality through pathways independent of blood pressure and heart failure, the underlying biological mechanisms require further elucidation. The “direct effect” observed in our models may be partially attributed to unmeasured physiological pathways or complex interactions between nitrates and other metabolic factors.

Fifth, several population-specific limitations must be acknowledged. The study utilized NHANES data, which is representative of the U.S. civilian population but may lack generalizability to other ethnic groups or global regions with distinct environmental and lifestyle exposure profiles.

Finally, the possibility of residual confounding cannot be entirely excluded. Our analysis lacked detailed information on the duration of hypertension. Although we adjusted for current blood pressure levels and antihypertensive medication use as proxies for severity, the absence of long-term blood pressure control data may lead to residual confounding. The duration of disease could independently influence cardiovascular prognosis, and it should be explored in future longitudinal studies with clinical cohorts. Additionally, physical activity was not included due to inconsistent survey methodologies across the 2005-2006 cycle and subsequent cycles. NHANES also lacks data on medication adherence, which may affect the management of hypertension and subsequent cardiovascular risk.