The NHANES is a cross-sectional survey conducted by the National Center for Health Statistics (NCHS) in the United States. The NCHS and the Centers for Disease Control and Prevention (CDC) conducted the survey. The NCHS’ Research Ethics Review Board evaluated and approved the NHANES study protocol. The study protocol was approved by the NCHS Ethics Review Board, and all participants provided informed consent. For further confirmation, please refer to the link to the NCHS ethics approval document for the NHANES data ¹ . Briefly, the NHANES uses a stratified and multistage probability approach, which surveys approximately 5,000 participants annually. The data collected includes demographic data, questionnaire data, laboratory data, examination data and limited access data.

There were a total of 101,316 participants included in the NHANES from 1999 to 2018. Of those, we excluded 62,568 participants because their age was younger than 30 or older than 74; 3,676 subjects were excluded due to imponderable total cholesterol (TC); 2 individuals were excluded due to imponderable high-density lipoprotein (HDL); 17,935 were excluded due to missing glucose; 655 participants were excluded due to imponderable blood pressure; and 58 individuals were excluded due to imponderable SII. Of 1,692 participants with CVD,448 were excluded for using NSAIDs and statins; 4,381 subjects with arthritis and thyroid disease were excluded. Finally, 9901 participants were included in this study (Figure 1).

The SII is based on a complete blood count. Standardized protocols for measuring blood counts (platelets, lymphocytes, and neutrophils) were provided by the NHANES Laboratory Procedures Manual. More details can be found at ² . In the present study, the following equation was utilized to calculate the SII (SII = platelet count × neutrophil count ⁄ lymphocyte count) [17]. The SII data were unevenly distributed and skewed to the right. Natural logarithmic transformation converts absolute differences into proportional changes. For clinical interpretation, Ln-transformed SII (Ln(SII)) values can be back-transformed to the original scale. A one-unit increase in Ln (SII) corresponds to an e-fold (≈2.72-fold) multiplicative change in the original SII values. Based on this, the SII data need to be Ln-transformed. Supplementary Figure S1 shows the distributional characteristics and normality assessment of SII values. (A) The histogram of the raw SII values showed extreme right-skewness. (B) The quantile-quantile plot confirmed severe non-normality (Anderson-Darling A = 617.20, P < 0.001). (C) The histogram of Ln (SII) showed substantial improvement, although with residual skewness. (D) The quantile-quantile plot of the transformed values demonstrated a persistent deviation from normality (A = 6.62, P < 0.001). The red lines represent theoretical normal distributions.

The FRS is a general 10-year risk estimate for CVD, which was developed using a Cox model based on the Framingham Heart Study [4]. In the present study, the FRS was calculated using sex, age, TC, HDL, systolic blood pressure (SBP), treatment for hypertension, smoking and diabetes. The FRS was Ln-transformed to reduce the effects of non-normality. To eliminate the potential contribution of neighborhood clustering by age and sex on neighborhood-level variance, the outcome variable used in the analysis was the normalized residual of the Ln-transformed FRS regressed on age and sex. Low 10-year CVD risk was defined as FRS <10%; FRS ≥10% was considered intermediate or high 10-year CVD risk.

The baseline characteristics included demographic factors (age, sex (men or women), education (high school or below and college graduate or above), marital status (married/Living with a partner, never married and widowed/divorced/separated), poverty income ratio (PIR) (2, 2-4 and 4) and ethnicity (Mexican American, Other Hispanic, Non-Hispanic White, Non-Hispanic Black, and Other Ethnicity), behavioral factors (smoking, alcohol consumption, physical activity (PA), and energy intake), and low-density lipoprotein (LDL). The metabolic equivalent (MET) represents the rate of oxygen uptake required to maintain the body’s basic metabolic processes while at complete rest. In line with World Health Organization (WHO) recommendations, we used MET values of 3.3 for walking, 4 for moderate activities, and 8 for vigorous activities [27]. PA was valued by total MET-minutes/week. Vigorous activity was defined as accumulating at least 3,000 MET-minutes/week, while moderate activity required a minimum of 600 MET-minutes/week. Individuals who met neither criterion were classified into the light activity category. Energy intake was estimated using the dietary intake data from the 24-h period before the interview. Body mass index (BMI) was calculated by dividing weight in kilograms by the square of height in meters. Participants were categorized into four BMI groups according to WHO criteria: underweight (BMI <18.5 kg/m²), normal weight (BMI 18.5–24.9 kg/m²), overweight (BMI 25.0–29.9 kg/m²), and obese (BMI ≥30.0 kg/m²). The diagnosis of CVD was determined by self-reported physician diagnoses obtained during interviews using a standardized questionnaire. subjects were asked, “Has a doctor or other health expert ever informed you that you have angina/congestive heart failure/coronary heart disease/heart attack (myocardial infarction)/stroke?” If the answer was “yes” to any of the above questions, the subjects were considered to have CVD.

Participants were stratified into four groups according to Ln (SII) quartiles. Continuous and categorical baseline characteristics were presented as mean ± standard deviation (SD) and number (percentage), respectively. Differences in the distribution of variables were compared using a weighted one-way analysis of variance (ANOVA) test for continuous data and a weighted chi-square test for categorical data, respectively.

The Pearson test was used to analyze the correlation between Ln (SII) and the FRS. Two models were developed to assess associations of Ln (SII) and the FRS or 10-year CVD risk level by using weighted GLM: the crude model was not adjusted; the adjusted model was adjusted for education, marital status, PIR, ethnicity, alcohol consumption, PA, energy intake, BMI, and LDL. To rigorously assess the robustness of our results, we performed a subgroup analysis to investigate potential modifying effects based on ethnicity.

Furthermore, weighted RCS were used to explore non-linear relationships between Ln (SII) and FRS or 10-year CVD risk level. Knot placement and model specification were implemented using Harrell’s rms package. Interior knots were positioned at clinically relevant percentiles of the Ln (SII) distribution following established conventions: the 10th, 50th, and 90th percentiles were used for the optimal 3-knot configuration, which was determined by comparing Akaike Information Criterion (AIC) across 3 to 6 knot models. Boundary knots were automatically anchored at the observed minimum Ln (SII)value of 2.47.

All statistical analyses were performed using R version 4.2.2, with a significance threshold of P < 0.05, and all statistical tests were two-sided.

A total of 9901 subjects were included in the study, among whom 53.27% were men, and 46.73% were women. The number of participants who were judged as having an intermediate or high 10-year CVD risk was 1,438 (14.52%). The demographic characteristics of the subjects by Ln (SII) quartiles are shown in Table 1. The results suggest statistically significant differences in age, gender, marital status, ethnicity, smoking status, treatment for hypertension, BMI, PA, TC, FRS, and 10-year CVD risk level (all P < 0.05). The demographic characteristics of the subjects after weighting are shown in Supplementary Table S1.

A Pearson correlation analysis was performed to examine the correlation between Ln (SII) and the FRS, along with other continuous variables. There was a positive relationship found between Ln (SII) and the FRS, with a corresponding correlation coefficient of 0.09 (P < 0.001). Except for energy intake and PA, there were positive relationships between the other factors and the FRS. The results of the Pearson correlation coefficient are shown in Figure 2.

Table 2 suggests the results of the association between Ln (SII) and Ln (SII) quartiles and the FRS in the crude model and adjusted model. In the crude model, Ln (SII) was analyzed as a continuous variable. The odds ratio (OR) for FRS was 2.18 [95% confidence interval (CI): 1.57–3.02] per unit increment in Ln (SII). Compared to the lowest quartile (Q1) of Ln (SII), the OR (95% CI) for Q2, Q3, and Q4 were 1.22 (0.76-1.96), 1.86 (1.15–3.01), and 3.06 (1.90–4.93), respectively (P for trend <0.001). In the adjusted model, the OR (95% CI) for FRS was 1.52 (1.12–2.06) for a per-unit increment in Ln (SII). Compared to the Q1 of Ln (SII), the OR (95% CI) for Q2, Q3, and Q4 were 1.19 (0.73–1.92), 1.49 (0.95–2.35), and 1.89 (1.20–2.98), respectively (P for trend = 0.004).

A non-linear relationship was explored between Ln (SII) and the FRS by RCS. Figure 3 shows that a linear association was found between Ln (SII) and the FRS (P for non-linearity = 0.972). As Ln (SII) increased, so did the value of FRS (P for overall trend <0.001).

Supplementary Table S2 shows the results of the association between Ln (SII) and Ln (SII) quartiles and 10-year CVD risk in the crude model and adjusted model. After adjusting for the potential confounders, all associations became non-significant (OR range: 1.03–1.13, P > 0.05 for all comparisons). Supplementary Figure S2 shows a linear association between Ln (SII) and 10-year CVD risk (P for non-linearity = 0.541). With the Ln (SII) increasing, 10-year CVD risk gradually increased (P for overall trend <0.001).

Supplementary Tables S3-S12 present ethnicity-stratified analyses of the association between Ln (SII) (as both continuous and quartile variables) and the FRS and 10-year CVD risk level. After fully adjusting for education, marital status, PIR, alcohol consumption, physical activity, energy intake, BMI, and LDL, we observed distinct ethnicity-associated patterns: No significant associations were found among Mexican American patients (OR range: 0.84–1.75, all P > 0.05), Other Hispanic subjects (OR range: 0.34-1.39, all P > 0.05), or Non-Hispanic Black subjects (OR range: 0.98–2.34, all P > 0.05). In the Non-Hispanic White group, higher Ln (SII) was associated with an increased FRS (per-unit OR = 1.72, 95% CI: 1.13–2.60; Q4 vs. Q1 OR = 2.30, 95% CI: 1.27–4.16; P for trend = 0.007), but it was not associated with 10-year CVD risk (OR range: 1.05–1.18). Conversely, the Other Ethnicity group exhibited an inverse 10-year CVD risk association (per-unit OR = 0.46, 95% CI: 0.27–0.79; Q3 vs. Q1 OR = 0.41, 95% CI: 0.19–0.90), whereas no significant association was observed for the FRS (OR range: 0.48–0.86).