This study included adults aged 18 or above in the NHANES database and spanning the years from 1999 to 2020. Initially, a total of 66,568 adult participants were enrolled. Participants were excluded if they lacked information on SII data (n = 6836) or had missing data on stroke diagnosis (n = 2059), smoking status (n = 47), hyperlipidemia (n = 2), and hypertension (n = 24). The final analysis included 57,600 participants, of which 2368 were stroke patients (refer to Fig. 1 for the participant flowchart). The participants were categorized into three groups based on SII tertiles: SII-low (384), SII-median (384–597), and SII-high ($\ge$597). The primary aim of this study was to examine the association between SII and the prevalence of stroke in the entire population.

Fig. 1.

Flowchart of participant selection. SII, systemic immune-inflammation index; NHANES, National Health and Nutrition Examination Survey.

SII was calculated using the formula: (platelet count $\times$ neutrophil count)/lymphocyte count. Lymphocyte, neutrophil, and platelet counts were measured using automated hematology analyzers and expressed as $\times$ 10${}^3$ cells/mL [11].

The Ethics Review Board of the National Center for Health Statistics (NCHS) of the Centers for Disease Control and Prevention approved the database protocols, and all participants provided written informed consent before enrollment.

Participants with incomplete data for SII (n = 6836), stroke (n = 2059), smoking status (n = 47), hyperlipdemia (n = 2) or hypertension (n = 24) were excluded, giving a final total of 57,600 participants from the initial NHANES group (n = 66,568). Participants were categorized into three groups according to SII tertiles: SII-low (384), SII-median (384–597), and SII-high ($\ge$597).

The primary outcome measure in this study was self-reported stroke, as determined by asking participants: “Have you ever been told you had a stroke?”, or “Has a doctor or other health professional ever told you that you had a stroke?”. Diabetes mellitus was defined based on self-reported medical history, use of oral hypoglycemic agents or insulin, fasting glucose level $\ge$126 mg/dL, or hemoglobin A1c level $\ge$6.5% [12]. Hyperlipidemia was defined as serum triglycerides $\ge$150 mg/dL, total cholesterol $\ge$200 mg/dL, low-density lipoprotein cholesterol $\ge$130 mg/dL, high-density lipoprotein cholesterol $\le$40 mg/dL in men or $\le$50 mg/dL in women, or use of medication for hyperlipidemia [13]. The presence of cancer was determined based on the self-reported question: “Have you ever been told you had cancer or malignancy?”. The biochemical parameters mentioned above were measured in a subset of participants who provided blood samples at the Mobile Examination Center (MEC).

In this study, appropriate weights were employed to account for the complex sampling design of NHANES, thereby ensuring a representative sample of the US national population (https://www.cdc.gov/nchs/nhanes/index.htm). Baseline characteristics were reported as weighted means $\pm$ standard error for continuous variables, and frequency (weighted percentages) for categorical variables. Differences between groups were compared using Analysis of Variance (ANOVA) for continuous variables, and chi-square tests for categorical variables. The percentage of missing data for covariates was 5% (body mass index (BMI) [1.72%]), except for the family income to poverty ratio (9.4%). Missing values for the family income to poverty ratio were categorized as “Unknown”. To incorporate all available data for modeling, imputation with the median of each variable was performed.

For the analysis, odds ratios (ORs) and 95% confidence intervals (CIs) were estimated using multivariate logistic regression models to assess the association between SII and the prevalence of stroke. When treating SII as a continuous variable, a log-transformation was applied and the change in stroke prevalence for each one-unit increase in log-transformed SII was determined. Restricted cubic spline (RCS) analysis in the fully adjusted model was employed to evaluate the dose-response relationship between SII and stroke, with nonlinearity assessed using the likelihood ratio test.

Covariates were progressively adjusted in three models for logistic regression analysis. Model 1 included adjustments for age (continuous), sex (male or female), and race/ethnicity (non-Hispanic White, non-Hispanic Black, Mexican American, and others). Model 2 expanded on Model 1 by further adjusting for smoking status (never, former, current), physical activity (sedentary, insufficient, moderate, high), education level (under high school, high school or equivalent, college or higher), family income to poverty ratio ($\le$1.0, 1.0–3.0, $>$3.0, unknown), and BMI (25.0, 25.0–29.9, $\ge$30.0 kg/m${}^2$). Model 3 (fully adjusted model) incorporated all the covariates from Model 2, as well as diabetes, hyperlipidemia, and hypertension. Supplementary Method 1 provides additional details on covariate assessment.

Subgroup analysis was conducted based on age (60 or $\ge$60 years), sex (male or female), race/ethnicity (White or non-White), smoking status (never or former/current), BMI (30.0 or $\ge$30.0 kg/m${}^2$), physical activity (sedentary/insufficient or moderate/high), hyperlipidemia (yes or no), hypertension (yes or no), and diabetes (yes or no). The significance of interaction terms between the stratification variables and SII was evaluated using the Wald test. Sensitivity analyses were performed by excluding non-Hispanic Black participants and those with missing data on BMI in the fully adjusted model.

All statistical analyses were conducted using R version 4.1.3 (R Foundation for Statistical Computing, Vienna, Austria) with the “survey” package. A two-tailed p-value 0.05 was considered statistically significant.

The baseline characteristics of all participants, categorized by SII levels, are presented in Table 1. The overall weighted mean age of the 57,600 participants included in the analysis was 47.48 years, and 51.86% were males. Individuals with SII-high tended to be older, male, non-Hispanic white, nonsmokers, with higher levels of education, family income, BMI, waist circumference, and physical activity, and lower levels of estimated glomerular filtration rate (eGFR). The prevalence of diabetes, hypertension, hyperlipidemia, cancer, and stroke was significantly lower in the SII-low group compared to the SII-high group.

The association between stroke and SII, analyzed as a continuous or categorical variable, is presented in Table 2. The results indicate that higher SII levels are associated with an increased likelihood of stroke. When SII was analyzed as a continuous variable, each unit increase in log-transformed SII was associated with a 34% higher risk of stroke in Model 2 (OR 1.34, 95% CI 1.02, 1.76). A statistically significant association was not observed with Model 3 when SII was analyzed as a continuous variable, however the trend was consistent with Model 2 as p trend 0.05. When SII was examined as a categorical variable, participants in the SII-high group had a significantly higher prevalence of stroke compared to those in the SII-low group (OR 1.18, 95% CI 1.01, 1.42). RCS analysis was also performed to evaluate the relationship between SII levels and stroke risk in the overall participant group. The analysis demonstrated a positive linear correlation between SII level and the prevalence of stroke (p for non-linearity = 0.387), as shown in Fig. 2.

Fig. 2.

Restricted cubic spline model for the association of SII with the prevalence of stroke. RCS analysis demonstrated a positive linear correlation between SII level and the prevalence of stroke (p for non-linearity = 0.387) in the overall participant group. SII, systemic immune-inflammation index; RCS, restricted cubic spline; CI, confidence interval; HR, hazard ratio.

The results of subgroup analyses are presented in Table 3. Similar trends were observed across different subgroups, including age (60 or $\ge$60 years), sex (male or female), race/ethnicity (non-Hispanic White or other), smoking status (never or former/current), BMI (30 or $\ge$30 kg/m${}^2$), physical activity (sedentary/insufficient or moderate/high), dyslipidemia (yes or no), hypertension (yes or no), and diabetes (yes or no). No significant interactions were found between the ORs and these stratifying variables, except for sex. This suggests that male participants tended to have higher SII levels and a higher likelihood of developing stroke.

Table 4 provides a summary of the sensitivity analyses conducted to assess the associations between SII and the prevalence of stroke. The results remained consistent even after excluding non-Hispanic Black participants and participants with missing data on BMI.