This study aimed to explore the predictive significance of NHHR for all-cause and cardiovascular mortality, attempting to seek possible threshold point. The civilian, non-institutionalized U.S. population's health and nutritional status are evaluated through the NHANES survey program. It consists of physical examinations conducted in mobile examination services as well as interviews conducted in homes (20).

A BMI of 30 kg/m² or higher was the criteria for classifying individuals as obese, with BMI calculated by dividing weight in kilograms by height squared in meters (kg/m²) (3). As demonstrated in Figure 1, participants were included at baseline based on the subsequent standards:(1) At least eighteen years of age; (2) Without cancer diseases or being pregnant at baseline; (3) Body mass index (BMI) ≥30 kg/m²; (4) Having complete follow-up, BMI, and blood lipid profiles data. Besides, the National Center for Health Statistics’ Ethics Committee authorized the protocols and procedures for the study.

The lipid profile of the fasting blood samples from the participants in this study were used to compute NHHR levels. Non-HDL-C levels were calculated by subtracting HDL-C from TC, and NHHR was calculated by dividing Non-HDL-C by HDL-C (21).

The endpoints of this study involved all-cause mortality and cardiovascular mortality. Death records were obtained from the publicly accessible linked mortality files, encompassing mortality-related variables exclusive to adults. The National Death Index (NDI) offered details about the survival condition and causes of death for the surveyed persons, with data recorded up to December 31, 2019.All-cause mortality was defined as death from any cause. Cardiovascular mortality consisted of deaths caused by CVD or cerebrovascular diseases (22).

Utilizing data information derived from questionnaire, laboratory examinations and physical examinations conducted in the NHANES. We gathered information on age, sex, race, poverty-income ratio, education level, smoking and drinking habit through questionnaire interviews. There were five categories for race: Mexican American, Non-Hispanic Black, Non-Hispanic White, Other Hispanic, and Other Race. The Poverty-Income Ratio (PIR) was calculated as the ratio of household income to the poverty threshold. Three categories were used to categorize educational attainment: above high school, below high school, and high school.

Participants were grouped into two categories: smokers and non-smokers, depending on their answer to a question about smoking at least 100 cigarettes in their lifetime. Additionally, participants were classified as drinkers or non-drinkers based on whether they had ever had more than 12 glasses of alcohol annually. Information on serum biochemical indicator [serum uric acid (SUA), fasting blood glucose (FBG), albumin (ALB), TC, TG, HDL-C, LDL-C] were collected through laboratory examinations. Comorbidities [hypertension, coronary heart disease (CHD), myocardial infarction, heart failure, stroke, hyperlipidemia, gout, diabetes and chronic kidney disease (CKD)], and medication usage (hypertensive medications, lipid-lowering, and diabetic medications) were also considered and included at baseline. Measurements of waist circumference and BMI were taken in accordance with a standardized procedure.

The diagnosis of hypertension was established if any of the following conditions were met: (1) SBP ≥ 140 mmHg and/or DBP ≥ 90 mmHg; (2) Currently taking antihypertensive medication; (3) Acknowledgement of a history of hypertension in a questionnaire; (4) Diagnosed by a clinician (23). Similarly, a history of CHD was diagnosed if the questionnaire responses acknowledge any of the following conditions: angina, coronary heart disease, or heart attack, in which case the individual was classified as having CHD. Hyperlipidemia was defined as TC ≥ 200 mg/dl, TG ≥ 150 mg/dl, LDL-C ≥ 130 mg/dl, or HDL-C ≤ 50 mg/dl for women and ≤40 mg/dl for men, or using lipid-lowering drugs (24). Diabetes mellitus was identified based on FBG ≥ 126 mg/dl, self-reported history of diabetes, hemoglobin A1c levels ≥6.5%, or use of taking diabetes pills (25). Calculations of estimated glomerular filtration rate (eGFR) were performed using the chronic kidney disease epidemiology collaboration (CKD-EPI) equation (26). CKD was defined as participants having an eGFR < 60 ml/min/1.73 m². Other medical history information is obtained through questionnaires.

Proper sampling weights were applied to reconstruct the data representative of the US civilian non-institutionalized population in order to reflect the complex survey methodology of NHANES. According to NHHR concentrations, participants were classified into four groups: 0%–25%, 25%–50%, 50%–75% and 75%–100% (Q1–Q4). Continuous variables were assessed for normality using the Kolmogorov–Smirnov test. Normal distributions were represented as the mean ± standard deviation (mean ± SD), while non-normal variables were denoted as the median (25th percentile, 75th percentile), using the Wilcoxon rank-sum test for inter-group comparisons. Categorical variables were represented as frequencies and weighted percentages, and comparisons were made using chi-square tests.

Multivariate Cox proportional hazards regression models have been applied to explore the linear relationship between NHHR concentration and all-cause mortality and cardiovascular mortality. To investigate any possible nonlinear link between the endpoints and NHHR, Restricted cubic spline (RCS) analysis was used for flexible modeling and identifying a threshold point of NHHR for mortality. Based on these identified thresholds, we performed segmented multivariate Cox regression and subgroup analysis to further examine the relationship between NHHR concentration and mortality risk. Model 1 adjusted for age, gender; Model 2 adjusted for age, sex, race, education level, poverty-income ratio, BMI and waist circumference; Model 3 further adjusted the history of diseases (diabetes, hypertension, gout, CHD, stroke, myocardial infarction, heart failure, CKD) and individual medication history (lipid-lowering drugs, antidiabetic drugs and antihypertensive drugs) based on model 2. We further conducted stratified analyses by age, sex, smoking, drinking, hypertension,diabetes and BMI.Moreover, multiple sensitivity analyses were carried out to evaluate how reliable the findings were. Missing covariates were imputed using the random forest method, which effectively handles missing data by identifying variable types and accounting for collinearity among predictors (27). The “missForest” package was employed for this imputation process. For all statistical analyses, R version 4.3.1 (R Foundation for Statistical Computing, Vienna, Austria) was utilized, and statistical significance was established applying a cut-off of P < 0.05.

Table 1 presented the baseline characteristics of 7,504 subjects, 4,260 of whom were female, with an average age of 46 years. According to NHHR concentrations (Q1: 0.28–2.23 mmol/L, Q2: 2.24–2.95 mmol/L, Q3: 2.96–3.85 mmol/L and Q4: 3.86–25.81 mmol/L), individuals in the higher NHHR group tended to be younger and predominantly male. Additionally, those in the highest group exhibited increased FBG, SUA, waist circumference, TG, TC, LDL-C, Non-HDL-C, eGFR, albumin and higher rates of smoking, drinking and hyperlipidemia (p < 0.05). There were 318 (4.24%) cardiovascular deaths and 866 (11.54%) all-cause deaths over the course of the median follow-up period of 119 months.

In the entire obese population, NHHR, when included as a continuous variable in the three regression model, failed to demonstrate a statistically significant correlation with either cardiovascular or all-cause mortality (Table 2).

A U-shaped association between NHHR and the risk of all-cause (P for nonlinear <0.001, Figure 2A) and cardiovascular mortality (P for nonlinear = 0.006, Figure 2B) was shown by the restricted cubic spline (RCS) curves. A cut-off point for NHHR was observed in our study. The lowest risk of all-cause and cardiovascular death was linked to NHHR concentrations of 3.369 mmol/L and 3.409 mmol/L, respectively (Figure 2).

As shown in Table 3, when NHHR was incorporated into the final fully adjusted model as a continuous variable, we observed that at NHHR concentrations below the cut-off value, an increase in each unit of NHHR level was significantly negatively correlated with all-cause mortality (HR = 0.82, 95% CI: 0.67–1.00) and cardiovascular mortality (HR = 0.68, 95% CI: 0.49–0.94); conversely, at NHHR concentrations greater than the cut-off value, relatively higher levels of NHHR were significantly positively correlated with all-cause mortality (HR = 1.13, 95% CI: 1.01–1.28) and cardiovascular mortality (HR = 1.18, 95% CI: 1.02–1.36).

Supplementary Figure S1 presented the results of the segmented subgroup analysis and interaction tests based on the cut-off points between the NHHR and mortality. No significant interactions were found between the various subgroups.

When NHHR levels were below the threshold, significant negative associations with mortality were observed among people who were sixty years of age or older, males, and smokers. Individuals with a habit of alcohol consumption (HR = 0.75, 95% CI: 0.62–0.91) and with a history of hypertension (HR = 0.81, 95% CI: 0.68–0.96) also showed a significant negative association with all-cause mortality.

In contrast, when NHHR levels were exceeded the threshold, individuals with no smoking history (HR = 1.25, 95% CI: 1.04–1.51) and no drinking habits (HR = 1.21, 95% CI: 1.00–1.45) significantly increase the chance of dying from all causes; those under the age of 60 (HR = 1.23, 95% CI: 1.03–1.47) significantly raise the chance of dying from cardiovascular disease.

To confirm the credibility of the U-shaped connection between NHHR and mortality, we conducted several sensitivity analyses. First, we excluded individuals over the age of 65 and those who experienced events within 1 year of follow-up to minimize the impact of severe acute illnesses on the outcomes. Second, considering the impact of lipid-lowering medication on blood lipid levels, those taking medications to decrease cholesterol at baseline were not included. Third, adjusted for age, gender, race, education levels, BMI, smoking, alcohol use, waist circumference, diabetes, hypertension, gout, CHD, stroke, myocardial infarction, heart failure, CKD, lipid-lowering drugs, antidiabetic drugs and antihypertensive drugs to assess the connection between NHHR levels and mortality. We found that the results were similar to previous studies (Supplementary Figure S2).