This cross-sectional study utilized data from the National Health and Nutrition Examination Survey (NHANES) conducted between 1999 and 2018. NHANES is a survey of the civilian non-institutionalized population in the United States that uses a complex stratified multistage probability design (15). The survey includes interviews, physical examinations [at home or a mobile examination center (MEC)], and laboratory tests and is conducted every 2 years. During the period 1999–2018, NHANES conducted 10 separate cycles. Each cycle included entirely new participants, and each participant had a unique identification number, ensuring the uniqueness of the data. Therefore, we could determine that there was no overlap of participants between cycles. In addition, when participants were included in a particular cycle, their baseline clinical characteristics were interviewed or collected during that cycle. NHANES was conducted by the National Center for Health Statistics of the Centers for Disease Control and Prevention (CDC) and approved by the NHANES Institutional Review Board, with written informed consent obtained from all participants.

A total of 23,480 adults (age >18 years), who were classified as obese, were included in the NHANES survey between 1999 and 2018. Of these, 3,997 participants were excluded due to the inability to calculate SII or SIRI data. Of the remaining 19,483 participants, 542 pregnant women and 1,680 cancer patients were excluded, leaving 17,261 obese adults for the analysis.

The Beckman Coulter MAXM instrument at the MEC produced complete blood cell counts on blood specimens. Detailed specimen collection and processing instructions are available in the Laboratory/Medical Technologists Procedures Manual (16).

SII and SIRI were calculated using the following formulas (17, 18):

SII = (platelet count × neutrophil count)/lymphocyte count

CVD is a composite of a group of specific cardiovascular diseases, including heart attack, angina, congestive heart failure, coronary heart disease, and stroke (19). The diagnosis of specific CVD was based on self-reported medical history in the medical conditions section of NHANES. The primary outcome of this study is the prevalence of CVD. The secondary outcome is the prevalence of five specific CVDs, including heart attack, angina, congestive heart failure, coronary heart disease, and stroke.

Standardized questionnaires were used to collect information on age, gender, ethnicity, education level, family income, smoking and drinking behavior, medical history, and medication use. Medical history was based on self-reported previous medical records from a healthcare professional or a physician. Biochemical parameters were measured using a rigorous procedure, the details of which are provided in the NHANES Procedures Manual for Laboratory/Medical Technologists (16). To facilitate data integration, the following variables were further classified: (i)

Ethnicity: Non-Hispanic white, non-Hispanic black, Mexican American, or other ethnicities

To provide nationally representative estimates, MEC weights were utilized in data analysis to account for oversampling, non-response, and non-coverage. Details on weighting methods are available on the NHANES website (https://wwwn.cdc.gov/nchs/nhanes).

Baseline demographic characteristics were presented as means (standard error, SE) and continuous variables and weighted percentages (95% confidence interval, CI) for categorical variables. Spearman's rank correlation coefficient was used to determine the association between SII and SIRI and baseline variables, including age, BMI, alanine aminotransferase (ALT), aspartate aminotransferase (AST), total cholesterol (TC), high-density lipoprotein cholesterol (HDL), estimated glomerular filtration rate (eGFR), and medical history. Quartiles were used to categorize the level of SII and SIRI as categorical variables. A standard normal distribution of SII and SIRI [mean = 0, standard deviation (SD) = 1] was created by standardizing the Z-scores.

Multivariate logistic regression models were used to calculate odds ratios (ORs) and 95% CIs between SII and SIRI levels (quartiles and per-SD) and CVD prevalence. Baseline variables were considered as candidate predictors for the multivariate regression model. Confounding covariates were progressively added to different models. Subgroup analyses were performed stratified by clinical characteristics, including sex (male or female), age (<65 or ≥65 years), body mass index (30–34.9, 35–39.9, or ≥40 kg/m²), ethnicity (non-Hispanic white, non-Hispanic black, Mexican American, and others), DM (no or yes), hypertension (no or yes), hyperlipidemia (no or yes), and asthma (no or yes). In subgroup analyses, the effect of SIRI level on the odds ratio of CVD prevalence with each 1 SD increase was assessed. Meanwhile, interaction P values between each stratification variable (e.g., male/female) were calculated to examine the consistency of the trend toward CVD prevalence with each 1 SD increase in SIRI level. If an interaction p-value of less than 0.05 is observed, this suggests that the association between SIRI and CVD prevalence in this stratification is significantly different across subgroups, with the association likely to be more pronounced in one subgroup than the others.

A generalized additive model (GAM) with a spline smoothing function was used to assess the dose-response relationship between SIRI levels and CVD prevalence. Nonlinear P-values were obtained using the log-likelihood ratio test (22). If nonlinear P-value is less than 0.05, the fitted curve is nonlinear and vice versa. For nonlinear association, threshold effect analysis based on a two-piecewise linear regression model was performed to calculate the inflection point in the smoothing curve at which the relationship between SIRI and CVD prevalence changed significantly (23).

Nomograms were constructed for clinical applications using logistic regression models. Variables were screened using univariate analysis and multivariate logistic regression. The discriminative power of the model was assessed using the area under the curve (AUC) and calibration of risk predictions via the bias-corrected calibration intercept (i.e., predicted vs. observed outcome) (24). Furthermore, decision curve analysis (DCA) was performed by estimating the net benefit at different threshold probabilities to determine the suitability of the established column line graphs for clinical application (25).

All analyses were conducted using the statistical packages R (version 4.2.1, R Foundation) and EmpowerStats (version 4.2.0, www.R-project.org, X&Y Solutions, Inc., Boston, MA). Bilateral P-values less than 0.05 were considered statistically significant.

A total of 17,261 participants with obesity were included, representing 68,648,287 individuals. The average age of the participants was 48.3 years, and 44.9% were male. Participants with obesity were divided into Q1, Q2, Q3, and Q4 based on SII and SIRI values. Significant differences were found in age, poverty income ratio (PIR), BMI, HDL, TC, ALT, AST, gender, ethnicity, education levels, DM, hyperlipidemia, smoking, and antihypertensive use (all P < 0.05) for SII. Similarly, for SIRI, significant differences were observed between participants with higher and lower SIRI values in age, PIR, BMI, HDL, TC, ALT, gender, ethnicity, education levels, DM, hypertension, smoking, drinking, antihypertensive use and glucose-lowering drug use (all P < 0.05). The detailed results of baseline characteristics of participants with obesity are presented in Supplementary Table S1 (for SII) and Table 1 (for SIRI).

Spearman correlation analysis revealed that, for SII, the prevalence of CVD was negatively correlated with age [Spearman correlation coefficient (rho) = −0.05, P < 0.001], gender (rho = −0.10, P < 0.001), ethnicity (rho = −0.06, P < 0.001), ALT (rho = −0.09, P < 0.001), and AST (rho = −0.13, P < 0.001). However, the prevalence of CVD was positively correlated with BMI (rho = 0.1, P < 0.001) and hyperlipidemia (rho = 0.03, P < 0.001). Moreover, the prevalence of CVD was not statistically correlated with the variables, including HDL (P = 0.31), TC (P = 0.21), eGFR (P = 0.09), DM (P = 0.14), hypertension (P = 0.54), and asthma (P = 0.70). Additionally, for SIRI, CVD prevalence was negatively correlated with ethnicity (rho = −0.12, P < 0.001), HDL (rho = −0.09, P < 0.001), TC (rho = −0.08, P < 0.001), and eGFR (rho = −0.11, P < 0.001) and positively correlated with age (rho = 0.07, P < 0.001), gender (rho = 0.13, P < 0.001), BMI (rho = 0.06, P < 0.001), DM (rho = 0.07, P < 0.001), hypertension (rho = 0.07, P < 0.001), and hyperlipidemia (rho = 0.02, P < 0.001) but not significantly correlated with ALT, AST, and asthma (all P > 0.05). The results are presented in Supplementary Table S2.

Survey-weighted logistic regression was performed to examine the relationship between SII and SIRI levels and the prevalence of CVD among the obese population. However, we found no evidence to suggest that SII was an independent risk factor for the prevalence of CVD (Supplementary Table S3). Additionally, in the crude model, the odds ratios of CVD prevalence increased significantly with the increase in SIRI level, including heart attack, angina, coronary heart disease, congestive heart failure, and stroke in the crude model (P for trend <0.001). Moreover, per-SD increase in SIRI levels increased the prevalence of CVD by 35% (1.35 [1.27–1.44]), heart attack by 32% [1.32 (1.24–1.41)], angina by 23% [1.23 (1.14–1.32)], coronary heart disease by 37% [1.37 (1.27–1.47)], congestive heart failure by 41% [1.41 (1.31–1.53)], and stroke by 24% [1.24 (1.16–1.32)], respectively. Besides, this relationship persisted in model 1 after adjusting for demographic variables, including age, gender, ethnicity, education levels, and PIR, except for angina (P for trend = 0.067). Furthermore, in models 2 and 3, multivariable regression analysis revealed that SIRI levels were independent risk factors for CVD, including coronary heart disease and congestive heart failure. Moreover, per-SD increase in SIRI was associated with a 13%, 15%, and 28% increase in the risk of CVD (OR = 1.13, 95% CI: 1.04–1.22, P = 0.01), coronary heart disease (OR = 1.15, 95% CI: 1.05–1.26, P = 0.002), and congestive heart failure (OR = 1.28, 95% CI: 1.16–1.41, P < 0.001), respectively, in the corresponding model 3, with maximum adjusted covariates (Table 2).

Subgroup analyses assessed the performance of SIRI with respect to the prevalence of CVD in some subpopulations. Our findings were robust in most subgroups, including gender, age, BMI, ethnicity, DM, asthma, hyperlipidemia, and hypertension (all P for interaction > 0.05). However, the association between SIRI and CVD differed by age in the subgroups (Figure 1). The findings revealed that SIRI was more likely to promote the development of CVD in participants aged ≥65 years (OR = 1.23, 95% CI: 1.13–1.34, P = 0.01).

We constructed a nomogram to assess the probability of CVD prevalence using baseline variables that were screened by univariate analysis and multivariate logistic regression models (Supplementary Tables S4 and S5). Finally, ten variables were selected for nomogram construction, including SIRI, age, gender, PIR, BMI, eGFR, DM, hyperlipidemia, hypertension, and smoke. Each variable had a corresponding hazard point, and total points were obtained by summing the points of each variable (Figure 2A).

Moreover, ROC analyses were used to evaluate the predictive value of SIRI in CVD prevalence. SIRI showed a certain accuracy in predicting CVD prevalence (AUC = 0.604, 95% CI: 0.59–0.62) (Figure 2B). In addition, a robust result was revealed when we performed ROC analysis in combination with other variables in the nomogram (AUC = 0.832, 95% CI: 0.82–0.84) (Figure 2C). Finally, calibration curves (Supplementary Figure S1A) and decision curve analysis (DCA) (Supplementary Figure S1B) demonstrated a stable consistency between predicted and actual probabilities.

Smooth curve fitting regression analysis was conducted to estimate the relationship between the odds ratio of SIRI and CVD prevalence based on model 3 (Figure 3). A significant linear association was observed between the odds ratio of SIRI and CVD prevalence (P for nonlinear = 0.275). Specifically, for each 1-unit increase in SIRI, the prevalence of CVD increased by 20% (OR = 1.20, 95% CI: 1.12–1.30, P < 0.001).