National Health and Nutrition Examination Survey (NHANES) is a population-based representative survey performed by the National Center for Health Statistics (NCHS) of the Centers for Disease Control and Prevention (CDC) (12). Its primary goal is to assess the health status and disease risk factors in the American population (13). NHANES collects comprehensive information, including demographic data, disease-related questionnaire responses, examination results, and laboratory data, through structured home interviews, physical examinations performed at mobile centers, and laboratory tests conducted at analytical facilities. The datasets analyzed in this study can be accessed through the NHANES website (https://www.cdc.gov/nchs/nhanes/index.htm). The NHANES protocol was approved by the research ethics review board of NCHS, and informed consent was obtained from all participants (14).

In brief, we analyzed 6 cycles of NHANES data spanning 11 years, from 2007 to 2018 (2007–2008, 2009–2010, 2011–2012, 2013–2014, 2015–2016, 2017–2018). A total of 59,842 participants were identified from the NHANES 2007–2018 data. We excluded individuals under the age of 60 (n = 47,932) and those without hypertension (n = 4,585). After further excluding participants with missing data on all-cause mortality (n = 14), BMI (n = 157), waist circumference (n = 420), triglycerides (n = 152), complete blood count (n = 672), medical history (n = 458), the final analysis included 5,433 individuals (see Supplementary Documents for screening process plots).

The NHANES public-use linked mortality file as of December 31, 2018, correlated with NCHS with the National Death Index, was used to determine mortality status in the follow-up individuals (15). The main outcomes in this study were all-cause mortality and cardiovascular mortality.

The parameters involving AVI, SIRI, NLR, and MHR were calculated using the below formulas:

The SIRI (systemic inflammatory response index) was defined as a neutrophil count × monocyte count/lymphocyte count and expressed as × 10^9 cells/µl (16).

The NLR (neutrophil-to-lymphocyte ratio) was defined as neutrophil count/lymphocyte count (17).

The MHR (monocyte to high-density lipoprotein) was defined as monocyte count/high-density lipoprotein (18, 19).AVI=2×(WC(cm))2+0.7×(WC(cm)−HC(cm))21000WC waist circumference, HC height circumference. CM centimeter (the unit).

We downloaded various data from NHANES, including demographic information [age, gender, race, education levels, family income-to-poverty ratio (PIR)], smoking status, alcohol consumption, and medical history [heart failure (HF), coronary artery disease (CAD), angina, stroke, diabetes, and hypercholesterolemia]. At a mobile examination center, participants' waist circumference (WC), hip circumference (HC), and body mass index (BMI) were measured. BMI was calculated as the ratio of weight (kg) to height (cm) squared. Additionally, Serum specimens were collected as part of the NHANES laboratory examination component for 2007–2018, processed, stored, and sent to Collaborative Laboratory Services for analysis. At baseline, we measured complete blood count, plasma albumin (g/L), blood urea nitrogen (mg/dl), creatinine (µmol/L), uric acid (UA) (µmol/L), alanine aminotransferase (AST) (U/L), aspartate aminotransferase (U/L), glycohemoglobin (%), glucose (mmol/L), cholesterol (mmol/L), HDL-cholesterol (mmol/L), and triglyceride (mmol/L) levels. Detailed information regarding these methods is publicly available on the NHANES website (20). Race was classified into categories including Mexican American, other Hispanic, non-Hispanic White, non-Hispanic Black, or another race (21). Education level was categorized into three groups: less than high school, high school, and higher than high school. The family income-to-poverty ratio (PIR) was divided into three categories: less than 1.3 for low-income, between 1.3 and 3.5 for middle-income, and above 3.5 for high-income households (22). Smoker was defined as participants who had smoked more than 100 cigarettes in their lifetime (23), regardless of whether they had quit smoking at the time of the interview (24). Alcohol drinkers were defined as individuals who consumed at least 12 drinks in the year preceding the survey (25). Questionnaires were used to diagnose hypertension (BPD035), heart failure (MCQ160b), CAD (MCQ160c), angina (MCQ160d), heart attack (MCQ160e), and stroke (MCQ160f) (26).

All data were analyzed using R statistics (version 4.2.2, https://cran.r-project.org/). The Kolmogorov–Smirnov test indicated that all continuous variables did not follow a normal distribution; therefore, these variables are presented as medians with interquartile ranges. Categorical variables are expressed as frequencies and percentages. To examine differences in variables across different levels of the Atrial Volume Index (AVI), we categorized AVI into tertiles, from the lowest to the highest. The Chi-squared test or the Kruskal-Wallis H-test was utilized to analyze differences in AVI across these four categories. The one-way ANOVA test, Kruskal-Wallis H-test, or Chi-squared test were employed to compare continuous or categorical variables within the tertile groups of AVI (27). A p-value of <0.05 was considered statistically significant.

Multivariate Cox regression models were used to estimate the associations between AVI and all-cause as well as cardiovascular mortality, calculating hazard ratios (HRs) and 95% confidence intervals (CIs). Model 1 was unadjusted, while Model 2 was adjusted for gender, age, race, and education level. Model 3 included further adjustments for poverty income ratio (PIR), body mass index (BMI), smoking status, alcohol consumption, diabetes, hypercholesterolemia, cholesterol, and triglycerides. Subgroup analyses were conducted to explore potential differences among specific populations based on gender, race, education, PIR, diabetes, coronary artery disease (CAD), angina, stroke, smoking, and drinking status. Kaplan-Meier curves were plotted to estimate survival probabilities over time across different AVI levels, with the log-rank test employed to assess disparities among the tertile curves. A restricted cubic spline (RCS) model was applied to investigate the non-linear relationships between AVI, inflammatory parameters, and all-cause or cardiovascular mortality. Additionally, multivariate logistic regression was conducted to assess the associations between inflammatory parameters and AVI. The receiver operating characteristic (ROC) curve was utilized to determine the cutoff value of AVI for identifying mortality. The same statistical methods were applied in multivariate Cox regression to explore the associations between inflammatory parameters and mortality. Subsequently, mediation analysis was performed to assess whether the relationship between AVI and mortality was partially mediated by inflammatory parameters. The presence of a mediating effect was defined by significant indirect, direct, and total effects (26).

Eventually, we developed a predictive model for AVI to assess mortality risk. To eliminate collinearity among variables, we employed the least absolute shrinkage and selection operator (LASSO) regression model to exclude certain variables (28). The model's evaluation was conducted using 10-fold cross-validation, and we plotted a curve concerning lambda values. A risk prediction nomogram plot was created based on key mortality-related variables, including AVI and inflammatory parameters (SIRI and MHR). The model's forecasting performance was quantified by the area under the curve (AUC) with a 95% confidence interval from the ROC curve. The clinical applicability of the model was assessed through decision curve analysis.

Table 1 presents the baseline characteristics of the individuals enrolled in this study across different AVI tertiles. Among the participants, the median AVI values were 23.93, with ranges from 19.91 in T1, 23.93 in T2, to 29.49 in T3. The median age of the 5,433 senior citizens was 70 years, with approximately 47.78% being male. The incidence of all-cause and cardiovascular mortality among these participants during a median follow-up of 69 months was 23.19% and 6.20%, respectively. Compared to those in the low AVI group, participants with elevated AVI levels exhibited a higher prevalence of smoking, alcohol consumption, lower education levels, and PIR. Additionally, in terms of medical history, participants in higher quartiles reported a greater prevalence of diabetes, hypercholesterolemia, heart failure (HF), CAD, angina, and heart attacks. Regarding physical examination metrics, the T3 group demonstrated significantly higher levels of weight, height, BMI, and WC. Participants with higher AVI levels also showed significantly elevated inflammatory parameters (including SIR, MHR, and NLR), complete blood cell counts (including leukocyte, lymphocyte, and monocyte counts), as well as higher levels of ALT, creatinine, uric acid (UA), glucose, triglycerides, and HbA1c.

In this study, after a median follow-up of 69 months, we observed 1,260 deaths, including 337 attributed to cardiovascular causes, within the older adult population. Table 2 illustrates the associations between the Atherogenic Index of Plasma (AVI) and both all-cause and cardiovascular mortality. In the crude model, AVI demonstrated a significant association with an increased risk of all-cause mortality (HR = 1.03, 95% CI = 1.02–1.04) and cardiovascular mortality (HR = 1.04, 95% CI = 1.02–1.06). These associations remained robust and statistically significant after multivariable adjustments, with Model 2 showing HRs of 1.04 (95% CI = 1.03–1.05) for all-cause mortality and 1.06 (95% CI = 1.03–1.08) for cardiovascular mortality, and Model 3 showing HRs of 1.09 (95% CI = 1.06–1.11) and 1.07 (95% CI = 1.03–1.12) respectively. Compared to the first tertile of AVI, the hazard ratios for participants in the second, third, and fourth tertiles were consistently higher, regardless of variable adjustments. The Kaplan-Meier survival plots presented in Figure 1 indicate significant differences in all-cause and cardiovascular mortality across AVI tertiles (P < 0.01), with an overall median survival time of 144 months in tertile 3 for all-cause mortality.

Using restricted cubic spline regression (RCS) models adjusted for the aforementioned confounders, we identified an L-shaped association between Dietary Inflammatory Index (DII) and both all-cause and cardiovascular mortality (Figure 2). The cut-off values were determined to be 19.26 for all-cause mortality and 19.43 for cardiovascular mortality.

We conducted subgroup analyses to further explore the relationships between AVI and mortality, with detailed results presented in Figure 3. Most analyses revealed differences within groups; however, the association between AVI and all-cause mortality was not significant among participants of other races, those with a PIR ≥ 3.5, individuals with only a high school diploma, and those with coronary artery disease (CAD) or angina. For cardiovascular mortality, significant differences were mostly absent across subgroup analyses, except for female participants, Hispanic participants, those with a PIR < 3.5, individuals with a high school education or less, and non-diabetics who did not have CAD, angina, or a smoking and drinking history.

Table 3 displays the associations between AVI and inflammatory parameters as determined by multivariate logistic regression. The results indicate that AVI is positively associated with Systemic Inflammation Response Index (SIRI) (OR = 1.12, 95% CI = 1.07–1.18), Monocyte to High-Density Lipoprotein Ratio (MHR) (OR = 5.69, 95% CI = 4.47–7.24), and Neutrophil to Lymphocyte Ratio (NLR) (OR = 1.04, 95% CI = 1.01–1.08). We employed RCS models with three knots (10th, 50th, and 90th percentiles) to assess the nonlinearity of these associations, confirming an L-shaped relationship (see Supplementary Documents).

Cox regression results examining the impact of inflammatory parameters on all-cause and cardiovascular mortality are summarized in Table 4. All indicators were positively associated with all-cause mortality, and most factors also correlated with cardiovascular mortality, with the exception of MHR in Model 3. Additionally, RCS models adjusted for confounders revealed an anhaped association between the two (see supplementary documents.

Figure 4(a) illustrates the mediation effects of SIRI, MHR, and NLR on the relationship between AVI and all-cause mortality, accounting for 27.15%, 35.15%, and 16.55% of the association, respectively. Figure 4(b) shows that these parameters mediated 14.08%, 9.71%, and 8.17% of the association between AVI and cardiovascular mortality.

We developed a risk prediction model utilizing LASSO penalized regression, incorporating 33 additional covariates (including age, gender, PIR, BMI, and laboratory examination results) to assess the impact of AVI on all-cause mortality. Through LASSO regression and a 10-fold cross-validation method for model evaluation, we identified 12 significant variables for inclusion in the nomogram. The predictive performance for AVI was validated using the ROC curve (AUC = 0.76 for the total), with performance metrics for 2, 3, and 4 years being 0.62, 0.56, and 0.52, respectively (see Supplementary Documents).