This investigation employed information from the National Health and Nutrition Examination Survey (NHANES) database (https://www.cdc.gov/nchs/nhanes/index.htm). NHANES operates as a nationally representative, multi-stage investigation executed by the National Center for Health Statistics (NCHS) and the Centers for Disease Control and Prevention (CDC), established to evaluate nutritional and health conditions among different age populations in the United States. The NCHS Ethics Review Board sanctioned all NHANES research protocols, with all participants providing informed consent. Since NHANES data are publicly available and fully de-identified, studies involving secondary analyses of the NHANES database are considered exempt from additional ethical approval according to the policy of our Institutional Review Board. The research selected adults aged 20 years and above as eligible subjects. Participants were excluded based on: (1) missing data on monocyte count or serum albumin; (2) pregnancy; (3) missing data on CVD; (4) absence of mortality data or loss to follow-up; and (5) missing key covariate information. Of the 91,351 individuals registered in NHANES between 2001 and 2018, 23,740 met the criteria for final analysis. The participant selection process is depicted in Figure 1.

The exposure variable, MAR, was derived from the complete blood cell count and standard biochemical parameters available in the NHANES database. Given the relatively small values and wide distribution range of MAR, and to mitigate the impact of extreme values on effect size, MAR was standardized in this study, each unit increase in standardized MAR corresponds to a change of one standard deviation in MAR. Furthermore, to better illustrate the impact of MAR on the occurrence and progression of CVD at different classification levels, this study categorized all participants’ MAR values into tertiles based on the distribution of the overall data: Tertile 1 (MAR < 0.10), Tertile 2 (0.10 ≤ MAR < 0.14), and Tertile 3 (MAR ≥ 0.14). Additionally, five composite inflammation indices, known to significantly influence disease outcomes, were included as covariates for comparative analysis: the Advanced Lung Cancer Inflammation Index (ALI), Neutrophil-to-Albumin Ratio (NAR), Hemoglobin, Albumin, Lymphocyte, and Platelet score (HALP), SII, and NLR (19–23). The calculation formulas for all indices are provided in the supplementary materials, Supplementary Table S1.

CVD diagnosis was established through participants’ self-disclosed medical records indicating congestive heart failure (CHF), coronary heart disease (CHD), angina, myocardial infarction, or stroke (24). The participants indicated whether healthcare providers or physicians had previously notified them about any of these medical conditions, with positive responses confirming CVD presence.

MAR integrates the biological mechanisms of monocytes and albumin, reflecting the pathophysiological balance between pro-inflammatory activity, nutritional status, and anti-inflammatory defense. Monocytes play a pivotal role in the initiation and progression of atherosclerosis through endothelial infiltration and cytokine release, representing a central mechanism in the pathogenesis of CVD (12). In contrast, serum albumin exerts protective effects by mitigating oxidative stress, exerting anti-inflammatory and antithrombotic functions, and maintaining cardiovascular homeostasis (13–16). Therefore, an elevated MAR indicates enhanced pro-inflammatory activity accompanied by diminished protective capacity, establishing a strong biological link with the development of CVD.

Mortality data, including all-cause and cardiovascular mortality, were procured from the NDI public mortality dataset, linked with participants’ health data from NHANES 2001–2018. All-cause mortality encompassed deaths resulting from any cause, whereas cardiovascular mortality was identified per the International Classification of Diseases codes, including fatalities associated with heart disease (I00–I09, I11, I13, I20–I51) and cerebrovascular disease (I60–I69) (25). The follow-up period in person-years was calculated as the interval from the date of NHANES Mobile Examination Center (MEC) examination until either the date of mortality or the end of surveillance (December 31, 2019), whichever happened first.

Based on relevant studies and considering potential confounders, this study incorporated the following covariates from NHANES interview and laboratory data: age (years); sex (male or female); race/ethnicity (non-Hispanic White, non-Hispanic Black, Mexican American, and other); educational background (less than high school, high school graduate, and post-high school); relationship status (married/Living with partner, widowed/divorced/separated/never married); family poverty income ratio (PIR) (≤ 1.30, 1.31–3.50, >3.50); total cholesterol level (TC, mg/dl); smoking habits (never smokers: persons who smoked fewer than 100 cigarettes throughout their lifetime; current smokers: individuals who smoked more than 100 cigarettes and continued smoking during the survey period; former smokers: subjects who smoked more than 100 cigarettes but stopped before the survey); alcohol intake (characterized as consuming more than one 12-ounce beer, one 5-ounce glass of wine, or 1.5 ounces of liquor in the previous year); body mass index (BMI): normal weight (<25 kg/m²), overweight (25–30 kg/m²), and obese (>30 kg/m²); diabetes (identified through self-reported history, antidiabetic medication or insulin usage, fasting plasma glucose ≥7.0 mmol/L, or hemoglobin A1c ≥ 6.5%) (24); hypertension (determined by the average of four measurements with systolic blood pressure ≥140 mmHg and/or diastolic blood pressure ≥90 mmHg) (26). Cancer, COPD, and CVDs were established through self-reported physician diagnoses. Additional information appears on the NHANES website.

Statistical analysis was performed following the guidelines set by the CDC, taking into consideration the intricate, multi-stage cluster survey design. Specific NHANES sample weightings were incorporated to ensure sample representativeness. For data with normal distribution, continuous measurements are displayed as mean ± standard deviation (SD), whereas non-normal continuous measurements are shown as median (interquartile range, IQR). Categorical data elements are expressed as counts (n) and proportions (%). The analysis of variance (ANOVA) method evaluated differences among continuous measurements, while chi-square testing assessed between-group variations for categorical elements.

After standardizing MAR, a multi-stage modeling approach was employed to explore the relationship between MAR and the risk of CVD and mortality. For the CVD investigation, a three-tiered multivariable logistic regression model was established: Model 1 remained unadjusted; Model 2 adjusted for demographic parameters (age, sex, and race); while Model 3 incorporated additional adjustments for potential confounding factors. The nonlinear relationships between MAR and CVD were investigated utilizing restricted cubic spline (RCS), while threshold effects were identified through likelihood ratio testing. Subgroup analyses per the demographic and clinical characteristics were conducted to evaluate effect modification. Furthermore, receiver operating characteristic (ROC) curve evaluation measured the diagnostic performance of MAR regarding various cardiovascular endpoints (CVD, CHF, CHD, angina, myocardial infarction, and stroke), comparing variations in areas under the curve (AUCs).

The mortality evaluation employed Cox proportional hazards modeling, incorporating an identical three-tiered adjustment approach to investigate associations between MAR and mortality endpoints, encompassing both all-cause and cardiovascular outcomes. The survival differences across groups utilized Kaplan–Meier curve analysis (categorized by MAR tertiles) alongside log-rank testing methods. To illustrate the dose-response correlation between MAR and death risk, RCS regression techniques were employed.

Multiple verification steps assessed result stability: (1) stratified examinations and effect modification testing utilizing principal covariates; (2) the main analyses were repeated according to quartiles of MAR: Quartile 1 (MAR < 0.10), Quartile 2 (0.10 ≤ MAR < 0.12), Quartile 3 (0.12 ≤ MAR < 0.16) and Quartile 4 (MAR ≥ 0.16); (3) additional sensitivity evaluations incorporating CVD condition and integrated inflammatory nutrition indicators; (4) To minimize the potential reverse causation bias, participants who died within 2 years of follow-up were excluded.

To quantify the significant predictive value of MAR for CVD and mortality outcomes, we constructed a baseline model that included age, gender, hypertension, diabetes, and composite inflammatory markers (NLR, NAR, ALI, SII, HALP). An extended model was then developed by incorporating the MAR indicator into the baseline model. A net reclassification improvement (NRI) analysis was performed to evaluate the improvement in predictive performance. Risk was classified as low or high based on the predicted probabilities from both models, with a threshold of 0.5. The statistical computations were conducted using the R 4.3.2 platform, with significance thresholds established at a two-tailed P value < 0.05.

Among the 23,740 participants in this investigation, 54.28% were male, and 45.72% were female, with a mean age of 46.94 ± 17.10 years. Of these, 49.09% were non-Hispanic White, and over half had completed at least high school education. A total of 1,920 participants had a diagnosis of CVD. Following stratification by MAR tertiles, a higher proportion of individuals with CVD, CHF, CHD, angina, myocardial infarction, and stroke were detected in the higher MAR tertiles. Characteristics more prevalent in the higher MAR tertiles included male sex, advanced age, smoking, obesity, lower income and education levels, and comorbidities like hypertension, diabetes, cancer, and COPD. Statistically significant baseline differences were detected between groups (P < 0.05), as demonstrated in Table 1.

Table 2 demonstrates the findings from a multivariable logistic regression analysis investigating the link between MAR and CVD across three distinct models. The investigation identified a substantial positive link between MAR and CVD. After standardizing MAR, the unadjusted Model 1 indicated that for each single-unit elevation in MAR, there was a 1.36-unit enhancement in CVD risk (95% CI: 1.26–1.47, P < 0.001). After adjusting for confounders in Models 2 and 3, the association weakened, although it remained significant in both adjusted models. Within the completely adjusted Model 3, every single-unit increment in MAR correlated with a 1.09-fold rise in CVD occurrence (95% CI: 1.03–1.16, P = 0.004), indicating that MAR remains significantly associated with CVD after accounting for confounding factors. Upon stratification of MAR into tertiles, subjects in the highest tertile exhibited a 35% elevated probability of CVD development vs. individuals in the lowest tertile (95% CI: 1.13–1.61, P = 0.001). Trend tests using MAR as a continuous variable revealed a notable increasing trend in CVD risk across all three models (P < 0.05), suggesting a statistically significant rise in CVD risk as MAR levels increase.

For other composite inflammatory markers, the unadjusted model revealed significant associations with CVD risk for all markers except ALI. In the crude model, the second and third tertiles of ALI were significantly correlated with CVD. After adjusting for confounders, only SII in the second and third tertiles remained associated with CVD, indicating that the relationship between SII and CVD may be primarily influenced by confounding factors. In contrast, the link between MAR and CVD exhibited stability, demonstrating a notable connection following comprehensive adjustment. Compared to other composite inflammatory markers, MAR showed a superior capacity to predict CVD risk. To explore the nonlinear correlation between MAR and CVD, an RCS curve was constructed, revealing a potential nonlinear association (Figure 2). However, threshold effect analysis did not identify a significant inflection point, suggesting the absence of a clear threshold effect between MAR and CVD. This may be due to the limited moderating effects of potential covariates or the restricted range of moderator variable values, which could obscure the manifestation of threshold effects (Table 3).

Subgroup analyses were performed based on nine variables—age, sex, BMI, cancer, COPD, alcohol consumption, hypertension, diabetes, and smoking status—to examine the robustness and possible variability of the link between MAR and CVD across different demographic and clinical subpopulations (Figure 3). A notable correlation between MAR and CVD remained evident across all categories, excluding patients with COPD and participants aged 20–40 years (P < 0.05). The findings indicate that the link between MAR and CVD exhibits stronger significance among participants aged above 40 years. Conversely, the relationship appears weaker or insignificant in those with COPD, likely due to confounding factors associated with the disease. Interaction analyses suggested no significant interaction effects between the eight subgroup factors and the MAR-CVD association.

The results (Supplementary Table S2) of the sensitivity analysis based on MAR quartile grouping showed that, in all three models, as MAR increased, the risk of CVD progressively elevated after reaching a certain threshold. In Model 3, the odds ratios (OR) for the third and fourth quartiles were 1.25 (95% CI: 1.01–1.54) and 1.39 (95% CI: 1.12–1.73), respectively (P for both < 0.05). The trend test yielded a P-value of < 0.05. This indicates that after stratification by quartiles, the relationship between MAR and CVD remains consistent, further confirming the stability of the results in this study.

To evaluate the prognostic performance of MAR vs. other composite inflammatory markers (encompassing NLR, NAR, ALI, SII, and HALP) for CVD, the AUC was calculated, as shown in Figure 4. The results confirm that MAR exhibits considerable predictive value relative to other inflammatory biomarkers. Additionally, ROC curve analyses of MAR for angina, myocardial infarction, CHD, CHF, and stroke are presented in Figure 5. Among these outcomes, CHF demonstrated the highest AUC, followed by CHD, suggesting that MAR may offer superior discriminatory power for predicting CHF and CHD compared to other CVD outcomes. This may indicate a stronger association between MAR and CHF or CHD or a more prominent role of MAR in their pathophysiological mechanisms. The NRI analysis showed that the addition of MAR significantly improved risk reclassification. The NRI value was 0.9, suggesting that the inclusion of MAR enhanced the ability to correctly reclassify individuals into higher or lower risk groups.

Throughout a median follow-up of 8.3 years (interquartile range: 4.9–13.0 years), the study documented 2,338 all-cause deaths and 705 CVD-related deaths. The link between MAR and mortality underwent examination via Cox regression analysis utilizing three sequential adjustment models. The MAR values were divided into tertile categories. Model 1 (unadjusted) demonstrated that subjects in the highest MAR tertile exhibited markedly increased risks for all-cause mortality [hazard ratio [HR]: 1.94, 95% confidence interval [CI]: 1.70–2.23] and CVD mortality (HR: 2.40, 95% CI: 1.84–3.12), with P < 0.001. After comprehensive covariate adjustment, Model 3 indicated that within the highest tertile, each incremental unit of MAR corresponded to a 25% heightened risk of all-cause mortality and a 46% elevated risk of CVD mortality (Table 4). The Kaplan–Meier analysis illustrated notably diminished survival probabilities for both all-cause and CVD mortality in participants within the highest MAR tertile vs. those in the lowest tertile (P < 0.001; Figures 6A,6B).

The RCS curve depicted in Figure 7A shows a nonlinear connection between continuous MAR levels and all-cause mortality (P for nonlinearity = 0.030). Specifically, lower MAR values were linked to a reduced mortality risk, but once MAR exceeded 0.125, the all-cause mortality risk showed progressive elevation. The RCS analysis in Figure 7B indicates an ascending pattern in cardiovascular mortality risk with elevating MAR levels, with the link between MAR and cardiovascular mortality appearing either linear or without significant nonlinearity (P for nonlinearity = 0.370).

Subgroup analysis stratified by nine factors—age, sex, BMI, cancer, COPD, alcohol consumption, hypertension, diabetes, and smoking status—indicated a positive link between higher MAR and elevated all-cause mortality risk in the subgroups defined by sex, BMI, cancer, alcohol consumption, diabetes, and hypertension. This association was particularly pronounced in individuals aged ≥60 years. No notable association was detected among smokers or those with COPD (P > 0.05). Interaction analysis identified a potential interaction between alcohol consumption and MAR regarding all-cause mortality risk, suggesting that alcohol consumption may modulate the predictive value of MAR for mortality (P for interaction < 0.05). No notable interactions were detected between MAR and the other factors (P for interaction >0.05) (Figure 8). In the subgroup analysis of MAR and CVD mortality, no significant associations were observed among individuals with cancer, COPD, diabetes, current smoking status, or those without hypertension. Conversely, MAR demonstrated greater predictive accuracy for CVD mortality risk in individuals aged ≥60 years. No notable interactions between MAR and the aforementioned stratification factors were observed regarding CVD mortality (P for interaction > 0.05) (Figure 9).

The sensitivity analysis based on MAR quartile grouping (Supplementary Table S3) demonstrated a positive association between higher MAR levels and the risk of both all-cause mortality and CVD mortality in the general population (P for the fourth quartile < 0.05). These findings were consistent with those obtained from the analysis using MAR tertile grouping, indicating that elevated MAR levels are associated with an increased risk of mortality outcomes. Furthermore, the sensitivity analysis results, adjusted for cardiovascular disease history (Supplementary Table S4) and other inflammatory composite markers (Supplementary Table S5), remained consistent, further reinforcing the positive association between elevated MAR levels and increased all-cause and cardiovascular mortality in the general population (P-values for the third quartile were all < 0.05). Additionally, after excluding deaths occurring within the first 2 years of follow-up, Cox regression analysis showed that the risks of both all-cause and cardiovascular mortality remained significantly higher in the third quartile of MAR (OR: 2.24, 95% CI: 1.69–2.97, P < 0.05), with results remaining consistent after adjustment for confounding factors (Supplementary Table S6). These sensitivity analyses confirm that the associations observed in this study are robust and consistent across different models.

Finally, the NRI analysis confirmed the incremental value of MAR in predicting mortality outcomes. Incorporating MAR into the baseline model significantly improved risk reclassification (NRI >0.5). These findings indicate that MAR provides additional predictive power beyond established risk factors and composite inflammatory markers.