Monocyte-to-albumin ratio outperforms triglyceride-glucose index in associating with stroke prevalence among diabetics: an NHANES cross-sectional study

Abstract

Background

Stroke is the second leading global cause of death/disability, with 85% being ischemic. DM increases Stroke prevalence and worsens outcomes. While the TyG Index shows inconsistent Stroke prediction in DM, chronic inflammation drives pathology. MAR, reflecting integrated inflammation status, has prognostic value but lacks evidence for association with Stroke prevalence across glycemic states (diabetes/prediabetes/non-diabetes). This NHANES study evaluates MAR-Stroke associations and compares predictive performance with TyG Index in DM.

Methods

This NHANES 2005–2018 analysis included 70,190 adults (≥ 20 years). After excluding participants with missing Stroke data or key variables, 15,679 were included in the analysis. We employed weighted multivariable logistic regression (stratified by diabetes/prediabetes/non-diabetes), RCS curves for nonlinearity, ROC analysis comparing MAR/ TyG Index discriminatory ability, and BMI mediation analysis. NHANES sampling weights and covariate adjustments for age, gender, race, education, and BMI were applied.

Results

Based on a cross-sectional analysis of the 2005–2018 NHANES database (sample size n = 15,679; Stroke prevalence 2.9%), the findings indicate that: In the general population, each 1-unit increase in MAR was associated with a 13% higher prevalence of Stroke (adjusted OR = 1.13, 95% CI: 1.03,1.24). This association was stronger among diabetic patients (OR = 1.23, 95% CI: 1.08, 1.41) and exhibited a near-linear dose–response relationship (P for nonlinearity = 0.013).MAR showed superior potential biomarker for Stroke (AUC = 0.594) compared with TyG Index (AUC = 0.583) and LDL (AUC = 0.568). After multivariate adjustment, MAR achieved an AUC of 0.808 (95% CI: 0.793,0.823). Mediation analysis revealed that BMI mediated 15.6% of MAR’s effect on Stroke prevalence (indirect effect β = 0.314, 95% CI: 0.136,0.508).

Conclusion

This NHANES analysis (N = 15,679) establishes elevated MAR as significantly associated with prevalent Stroke, outperforming the TyG Index especially in diabetics. Diabetic individuals in the highest MAR quartile exhibited 2.5-fold greater Stroke prevalence vs. the lowest quartile, with optimal prevalence discrimination at MAR > 0.152. BMI mediated 15.6% of this association, indicating modifiable pathways. Clinical translation requires prospective validation of the MAR threshold, confirmation of BMI-mediated mechanisms, and evidence for MAR-guided intervention efficacy. Provided sufficient longitudinal and interventional evidence is obtained, MAR shows promise as a dual-purpose tool for Stroke prevalence stratification and therapeutic monitoring along the inflammation-metabolism axis in high-risk populations, particularly diabetics.

Introduction

Stroke stands as the second leading cause of global disability and mortality, imposing a substantial burden on healthcare systems worldwide. In the United States alone, approximately 795,000 individuals experience a first or recurrent Stroke annually [1, 2]. Ischemic Stroke, accounting for roughly 85% of all cases [8], is of particular concern. A critical and well-established risk factor for ischemic Stroke is diabetes mellitus (DM). Individuals with DM face a significantly elevated risk of both incident Stroke and poorer outcomes, including higher post-Stroke mortality and recurrence rates, compared to the non-diabetic population [3, 4]. Consequently, identifying effective tools for prevalence stratification and outcome prediction in this vulnerable diabetic cohort is paramount.

While monitoring glycemic control is essential, the Triglyceride-Glucose Index (TyG Index), a surrogate marker of insulin resistance derived from fasting triglycerides and glucose [12], has shown inconsistent association strength for Stroke prevalence specifically in diabetic populations. Notably, some studies paradoxically suggest that higher TyG Index values may not confer protection and could even be associated with an increased prevalence of Stroke recurrence [13, 14].This highlights TyG Index’s limitations and underscores the need for novel biomarkers that more comprehensively capture the complex pathophysiology linking DM to Stroke.

Emerging evidence points to chronic inflammation as a key mechanistic link between DM and Stroke susceptibility/progression. Specifically, activation of the renin-angiotensin system (RAS) in diabetic states exacerbates vascular inflammation and endothelial dysfunction, driving microvascular damage [46]. Furthermore, recent bioinformatics analyses have uncovered inflammation-associated lncRNA-mRNA networks in type 2 diabetes mellitus (T2DM), revealing novel regulatory mechanisms underlying this inflammatory dysregulation [49]. Therefore, inflammatory markers hold promise for refining risk assessment. The Monocyte-Albumin Ratio (MAR), calculated as the absolute monocyte count divided by serum albumin concentration, has recently gained attention as a novel, readily available, and integrative inflammatory biomarker. It reflects both the pro-inflammatory state (via monocytes) and the anti-inflammatory/counter-regulatory capacity (via albumin). MAR has demonstrated prognostic utility in diverse inflammatory conditions, including spontaneous intracerebral hemorrhage [5], cardiovascular disease [6], and non-small cell lung cancer [7].

The biological rationale for investigating MAR in Stroke, particularly in the context of DM, is compelling: Monocytes are pivotal mediators of the inflammatory cascade following cerebral ischemia. Early after Stroke, pro-inflammatory monocytes infiltrate the injured brain, differentiate into M1 macrophages, and release cytokines (e.g., IL-6, TNF-α), exacerbating tissue damage [9, 10]. Conversely, albumin possesses significant antioxidant and anti-inflammatory properties. It contributes to brain tissue repair post-Stroke by modulating inflammatory responses and interacting with cytokines/chemokines [11]. Critically, hypoalbuminemia itself is an independent risk factor for Stroke. Thus, MAR uniquely encapsulates the dynamic interplay between detrimental pro-inflammatory cellular responses (monocytes) and protective anti-inflammatory/antioxidant mechanisms (albumin) central to Stroke pathophysiology.

Given the heightened inflammatory milieu characteristic of DM and its critical role in promoting Stroke, coupled with the limitations of TyG Index for Stroke prediction in this group, we hypothesize that MAR may serve as a superior biomarker. Specifically, we propose that elevated MAR is associated with a higher prevalence of self-reported Stroke history among diabetic individuals, reflecting cumulative Stroke prevalence. While this cross-sectional design precludes evaluation of Stroke prevalence, recurrence, or prognosis, establishing this association is a necessary first step toward future longitudinal validation. This study aims to investigate the correlation between MAR and Stroke occurrence and outcomes, with a specific focus on elucidating its potential utility within the high-risk diabetic subset.

Although many studies have established the relationship between monocytes and albumin and Stroke when used individually, the correlation between MAR, a combined indicator of the two, and Stroke remains unexplored. Given diabetes’s high prevalence, this study focused on the MAR—Stroke relationship across different blood glucose levels. Moreover, we compared MAR’s Stroke—predictive ability with TyG Index to determine whether MAR could serve as an emerging Stroke predictor. To this end, we conducted a cross—sectional study using 2005–2018 NHANES data.

Methods

Survey description

The National Health and Nutrition Examination Survey (NHANES), administered by the Centers for Disease Control and Prevention (CDC), serves as a comprehensive national health survey designed to assess the health and nutritional status of the U.S. population. Utilizing an intricate stratified multistage probability sampling methodology, NHANES ensures that the selected sample accurately reflects the demographic diversity of the U.S. population, thus supplying nationally representative data to support the study. NHANES is a comprehensive and in-depth survey that includes a household interview, a comprehensive physical examination, a series of laboratory tests, and detailed questionnaires covering a wide range of dimensions, such as health indicators, lifestyle, nutritional intake, and environmental exposures, providing researchers with a rich source of information on health indicators and nutritional status. NHANES provides researchers with a wealth of data resources [16].

The research procedures of NHANES strictly follow ethical norms and have been approved by the Research Ethics Review Board of the U.S. National Center for Health Statistics (NCHS), and the written informed consent of each participant was obtained prior to the commencement of the study, which ensures the legality and ethicality of the study. The high quality data provided by the NHANES database not only provides a scientific basis for the formulation of public health policies, but also provides a solid data foundation for scientific research in related fields. Through in-depth analysis of the data in the NHANES database, researchers are able to better understand the distribution characteristics of health problems, their influencing factors, and their relationship with socio-demographic characteristics, thus providing strong support for improving public health and promoting the health of the population. For more detailed information about the NHANES study and its statistical results, please refer to its official website https://www.cdc.gov/nchs/nhanes/.

Study design and participants

In the present study, we initially utilized health data from seven consecutive cycles (2005–2018) of the NHANES database, encompassing a total of 70,190 survey participants. Subsequently, we first excluded participants with missing Stroke data (N = 30,442). The selection criteria for participants in this study were as follows:(1) adults aged 20 or older; (2) participants with complete laboratory records for monocytes, albumin, and blood glucose measurements were included; (3) people with no missing Stroke data. After that we calculated MAR and TyG Index, the key variables required for this study, and excluded participants with missing MAR, TyG Index, and LDL data, as well as those with a fasting subweight (WTSAF2YR) value of 0 (N = 24,069). Finally, a total of 15,679 participants were included in this study for subsequent analysis. The detailed process is illustrated in Fig. 1.

Fig. 1.

Flowchart of participant recruitment

The outcome variable: Stroke

We defined whether a respondent had a Stroke based on the patient’s response to the questionnaire “Has a doctor or other health professional ever told you that you had a Stroke”. Those who responded yes were categorized as having a Stroke, while those who responded no were categorized as not having a Stroke [15]. Since the type of Stroke could not be clearly recorded in the questionnaire, the Strokes in this study were taken as their epidemiological reference, and by default, it was assumed that all the types of Stroke recorded in this study were ischemic Strokes.

Definition of MAR and TyG Index

The MAR was calculated as follows: MAR = Monocyte count (× 10⁹/L)/Albumin concentration (g/L).

The TyG Index was calculated as follows: TyG Index = Ln [fasting triglycerides (mg/dL) × fasting glucose (mg/dL)] [13].

Definition of different glycemic status

The main basis for classifying different glycemic status in this study was (1) HbA1c level; (2) having been told by a doctor or professional whether or not they had Diabetes.

The different glycemic statuses were: (1) Diabetes: HbA1c ≥ 6.5% or being told that they have diabetes; (2) Pre-Diabetes: not being told that they have Diabetes and HbA1c ≥ 5.7% and HbA1c < 6.5%; and (3) Non-Diabetes: not being told that they have Diabetes and HbA1c < 5.7%.

Definition of other covariables

This study integrated multiple covariates that could potentially impact the correlation between MAR and Stroke. These covariates encompassed age, gender, race, education level, poverty-to-income ratio (PIR), BMI, systolic and diastolic blood pressure, high-density lipoprotein cholesterol (HDL), low-density lipoprotein cholesterol (LDL), total triglycerides, total cholesterol, smoking status, alcohol consumption, and cancer. Age, gender, race, education level, and PIR were derived from demographic data. Educational attainment was grouped into three categories: below high school, high school graduate, and above high school. PIR was calculated as the ratio of family income to the federal poverty threshold, serving as a standardized metric for socioeconomic status. BMI was calculated by dividing weight in kilograms by the square of height in meters. Blood pressure measurements, including systolic and diastolic readings, were obtained by experienced clinicians following a standardized procedure.Specifically, three consecutive half—minute readings were taken at half—minute intervals using a mercury sphygmomanometer. The average of these three readings was considered as the participant’s blood pressure [16]. HDL cholesterol, LDL cholesterol, total triglycerides, and total cholesterol were obtained from laboratory data records. Participants who answered the questionnaire"How many drinks per day on average in the past year"with more than 12 drinks were defined as alcoholics. Cancer patients were identified by the questionnaire"Have you ever been told you have cancer?".

Definition of smoking status: based on (1) questionnaire: at least 100 cigarettes smoked in lifetime; (2) questionnaire: current smoker or not. Definitions: (1) never smoked: smoked less than 100 cigarettes in their lifetime [17]; (2) currently smoked: smoked more than 100 cigarettes in their lifetime and smoked sometimes or every day; and (3) quit smoking: smoked more than 100 cigarettes in their lifetime and are no longer smokers.

The selection of these covariates as confounders for inclusion in the statistical analysis of this study was primarily based on the following considerations: Stroke prevalence exhibits well-documented gradients with age, gender, and race. Furthermore, individuals with lower levels of education and income face heightened Stroke prevalence indirectly through reduced healthcare access and unhealthy behaviors (e.g., smoking). Obesity directly promotes vascular inflammation and insulin resistance, consequently leading to dyslipidemia—a core mechanism underlying atherosclerosis. Factors such as smoking and excessive alcohol consumption exacerbate inflammation and endothelial damage, while cancer contributes to Stroke prevalence by inducing hypercoagulability and malnutrition.

Statistical analyses

Statistical analyses and data visualization were performed using R Studio and SPSS software (version 27.01). Intermediary analysis was conducted using the PROCESS macro (version 4.2) within SPSS [36]. In accordance with NHANES recommendations and guidelines, appropriate sampling weights were applied to account for the complex, multi-stage clustered survey design.

Missing values in all variables, except for the main variables (Stroke, MAR, TyG Index, and LDL), were imputed. Continuous variables were imputed using the median, and categorical variables were imputed randomly. Prior to subsequent analyses, a sensitivity analysis was performed on the imputed data to assess the robustness and stability of the results.

Participant characteristics were compared based on Stroke occurrence and MAR quartiles using chi-square or t-tests. The linear association between standardized MAR and prevalent Stroke was assessed using multivariable logistic regression across three models, stratified by glucose status: Model 1: Unadjusted. Model 2: Adjusted for age, gender, race, and education. Model 3: Adjusted for Model 2 covariates plus BMI, PIR, LDL, HDL, Alcohol Consumption, Smoking, and Cancer.

To examine potential non-linear relationships between standardized MAR and Stroke, restricted cubic splines were fitted for different glucose levels. Stratified analyses were also conducted based on age (20–39, 40–59, ≥ 60 years), gender, education, race, Diabetes Status, and Smoking.

We constructed univariate and multivariate logistic regression models incorporating MAR, TyG Index, LDL, and the MAR*TyG Index interaction term. Receiver operating characteristic (ROC) curves were plotted for each model, and their predictive abilities were compared using area under the curve (AUC) values.

Finally, a mediation analysis was performed to investigate the potential mediating role of BMI in the association between MAR and Stroke. A two-sided p-value < 0.05 was considered statistically significant for all analyses.

The cross-sectional design of this study, utilizing NHANES data, inherently limits the interpretation of identified associations between biomarkers and Stroke to the realm of prevalence rather than prevalence. The outcome measure reflects self-reported Stroke history ascertained during the NHANES interview, precluding assessment of causal relationships or predictive capacity for future Stroke events. A critical methodological constraint arises from the concurrent assessment of monocyte-to-albumin ratio (MAR) exposure and Stroke status. This temporal ambiguity introduces the substantial possibility of reverse causality, whereby pre-existing Stroke pathology could potentially alter inflammatory biomarker profiles (e.g., elevating monocyte counts or reducing albumin concentrations), rather than MAR serving as a predictor of Stroke. While adjustments were made for significant confounders, including demographic and metabolic factors, the fundamental inability of the cross-sectional framework to establish temporal sequence between exposure and outcome remains an insurmountable limitation, potentially resulting in residual confounding.

Result

Baseline characteristics of participants

Baseline variables categorized by Stroke status and MAR quartiles are displayed in Tables 1 and 2. The total cohort comprised 597 Stroke participants and 15,082 non-Stroke individuals (prevalence: 2.9%). Stroke patients were significantly older than non-Stroke participants (median age: 67 vs. 47 years, P < 0.001) and displayed notable socioeconomic disparities, including a greater percentage of non-Hispanic Black individuals (15% vs. 10%), reduced educational attainment (28% vs. 16% did not complete high school), and a lower median poverty-to-income ratio (2.09 vs. 2.77, all P < 0.001). Stroke patients exhibited markedly elevated systolic blood pressure (121 vs. 120 mmHg), body mass index (29 vs. 28 kg/m²), and MAR levels (0.14 vs. 0.12, all P < 0.05). Metabolic indices, such as the TyG Index and total triglycerides, were markedly enhanced in the Stroke cohort (all P < 0.001), but LDL and total cholesterol levels were decreased. A history of smoking, diabetes, and cancer was substantially correlated with Stroke (P < 0.001).

Table 1.

Basic Characteristics of Participants by Stroke

Characteristic	Na	Stroke			P valuec
		Overall, N = 15,679 (100%)b	No, N = 15,082 (97%)b	Yes, N = 597 (2.9%)b
Age (years)	15,679	47.0 (34.0, 60.0)	47.0 (33.0, 60.0)	67.0 (55.0, 77.0)	< 0.001
Gender	15,679				0.05
Male		7507 (48%)	7220 (48%)	287 (43%)
Female		8172 (52%)	7862 (52%)	310 (57%)
Race/ethnicity	15,679				0.001
Non-Hispanic white		6636 (68%)	6336 (67%)	300 (69%)
Non-Hispanic black		3146 (10%)	2981 (10%)	165 (15%)
Mexican American		2501 (8.7%)	2447 (8.8%)	54 (4.6%)
Other race—including multi-racial		1805 (7.7%)	1768 (7.7%)	37 (8.0%)
Other Hispanic		1591 (5.8%)	1550 (5.8%)	41 (3.7%)
Education Level	15,679				< 0.001
Under highschool		3906 (16%)	3690 (16%)	216 (28%)
Highschool		3572 (23%)	3418 (23%)	154 (29%)
Above highschool		8201 (61%)	7974 (61%)	227 (42%)
Poverty-income ratio (PIR)	15,679	2.74 (1.60, 4.86)	2.77 (1.63, 4.92)	2.09 (1.22, 3.34)	< 0.001
Body Mass Index (kg/m2)	15,679	28 (24, 32)	28 (24, 32)	29 (25, 34)	0.036
Systolic Blood Pressure (mmHg)	15,679	120 (111, 128)	120 (111, 127)	121 (118, 138)	< 0.001
Diastolic blood pressure (mmHg)	15,679	69 (64, 75)	69 (64, 75)	69 (62, 74)	0.006
Alcohol consumption	15,679				0.7
No		15,322 (98%)	14,732 (98%)	590 (98%)
Yes		357 (1.9%)	350 (1.9%)	7 (1.6%)
Smoking status	15,679				< 0.001
Smoking		3058 (19%)	2913 (19%)	145 (25%)
Never smoking		8776 (55%)	8536 (56%)	240 (40%)
Quit Smoking		3845 (25%)	3633 (25%)	212 (34%)
Diabetes	15,679				< 0.001
Diabetes		2462 (11%)	2242 (11%)	220 (33%)
Prediabetes		3843 (21%)	3680 (21%)	163 (26%)
Nondiabetes		9374 (68%)	9160 (69%)	214 (41%)
Cancer	15,679				< 0.001
No		14,222 (90%)	13,758 (91%)	464 (77%)
Yes		1457 (9.7%)	1324 (9.3%)	133 (23%)
Monocyte-Albumin-Ratio	15,679	0.12 (0.10, 0.15)	0.12 (0.10, 0.15)	0.14 (0.11, 0.18)	< 0.001
Triglyceride-Glucose Index	15,679	8.54 (8.13, 8.97)	8.54 (8.13, 8.96)	8.71 (8.30, 9.16)	< 0.001
HDL Cholesterol (mmol/L)	15,679	1.34 (1.11, 1.63)	1.34 (1.11, 1.63)	1.29 (1.09, 1.66)	0.14
LDL Cholesterol (mmol/L)	15,679	2.87 (2.30, 3.52)	2.90 (2.33, 3.52)	2.56 (2.02, 3.39)	< 0.001
Total Cholesterol (mmol/L)	15,679	4.89 (4.24, 5.61)	4.89 (4.24, 5.61)	4.65 (4.03, 5.46)	< 0.001
Total Triglycerides (mmol/L)	15,679	1.14 (0.78, 1.66)	1.13 (0.78, 1.66)	1.25 (0.87, 1.78)	< 0.001

^aN not Missing (unweighted)

^bMedian (IQR) for continuous; n (weighted %) for categorical

^cDesign-based KruskalWallis test; Pearson’s X^2: Rao & Scott adjustment

Table 2.

Basic Characteristics of Participants by Monocyte-Albumin-Ratio Quartiles

Characteristic	Na	Monocyte-Albumin-Ratiod					P Valuec
		Overall, N = 15,679 (100%)b	Q1, N = 4108 (25%)b	Q2, N = 4088 (26%)b	Q3, N = 3603 (24%)b	Q4, N = 3880 (24%)b
Age (years)	15,679	47.0 (34.0, 60.0)	45.0 (33.0, 57.0)	46.0 (33.0, 59.0)	48.0 (33.0, 61.0)	50.0 (35.0, 65.0)	< 0.001
Gender	15,679						< 0.001
Male		7507 (48%)	1742 (41%)	1849 (46%)	1809 (51%)	2107 (54%)
Female		8172 (52%)	2366 (59%)	2239 (54%)	1794 (49%)	1773 (46%)
Race/Ethnicity	15,679						< 0.001
Non-Hispanic white		6636 (68%)	1401 (62%)	1681 (67%)	1624 (70%)	1930 (72%)
Non-Hispanic black		3146 (10%)	905 (12%)	828 (11%)	672 (9.4%)	741 (9.5%)
Mexican American		2501 (8.7%)	697 (9.8%)	706 (9.0%)	577 (8.5%)	521 (7.4%)
Other race—including multi-racial		1805 (7.7%)	678 (11%)	469 (7.5%)	331 (6.3%)	327 (6.3%)
Other Hispanic		1591 (5.8%)	427 (5.8%)	404 (5.7%)	399 (6.2%)	361 (5.3%)
Education level	15,679						< 0.001
Under highschool		3906 (16%)	960 (15%)	984 (15%)	947 (17%)	1015 (18%)
Highschool		3572 (23%)	862 (21%)	869 (21%)	814 (23%)	1027 (29%)
Above highschool		8201 (61%)	2286 (65%)	2235 (64%)	1842 (60%)	1838 (53%)
Poverty-income ratio (PIR)	15,679	2.74 (1.60, 4.86)	3.00 (1.79, 4.97)	2.79 (1.64, 5.00)	2.78 (1.61, 5.00)	2.41 (1.42, 4.41)	< 0.001
Body Mass Index (kg/m2)	15,679	28 (24, 32)	26 (23, 30)	28 (24, 32)	28 (25, 33)	29 (25, 35)	< 0.001
Systolic Blood Pressure (mmHg)	15,679	120 (111, 128)	119 (109, 125)	120 (110, 127)	120 (112, 128)	120 (113, 131)	< 0.001
Diastolic Blood Pressure (mmHg)	15,679	69 (64, 75)	69 (65, 75)	69 (65, 76)	69 (64, 75)	69 (63, 76)	0.5
Alcohol Consumption	15,679						0.04
No		15,319 (98%)	4,021 (98%)	4,010 (98%)	3,506 (97%)	3,782 (98%)
Yes		360 (2.1%)	87 (1.9%)	78 (1.5%)	97 (2.5%)	98 (2.4%)
Smoking status	15,679						< 0.001
Smoking		3058 (19%)	556 (12%)	664 (16%)	769 (22%)	1069 (28%)
Never smoking		8776 (55%)	2650 (64%)	2441 (59%)	1933 (53%)	1752 (45%)
Quit smoking		3845 (25%)	902 (24%)	983 (25%)	901 (26%)	1059 (28%)
Diabetes	15,679						< 0.001
Diabetes		2462 (11%)	487 (8.4%)	573 (8.9%)	603 (11%)	799 (17%)
Prediabetes		3843 (21%)	842 (16%)	971 (19%)	910 (22%)	1120 (26%)
Nondiabetes		9374 (68%)	2779 (75%)	2544 (72%)	2090 (67%)	1961 (57%)
Cancer	15,679						0.034
No		14,222 (90%)	3794 (91%)	3748 (91%)	3256 (90%)	3424 (89%)
Yes		1457 (9.7%)	314 (9.2%)	340 (8.8%)	347 (9.7%)	456 (11%)
Stroke	15,679						< 0.001
No		15,082 (97%)	4000 (98%)	3963 (98%)	3470 (97%)	3649 (95%)
Yes		597 (2.9%)	108 (1.9%)	125 (2.3%)	133 (2.9%)	231 (4.6%)
HDL cholesterol (mmol/L)	15,679	1.34 (1.11, 1.63)	1.42 (1.19, 1.73)	1.37 (1.14, 1.68)	1.32 (1.09, 1.58)	1.27 (1.06, 1.55)	< 0.001
LDL cholesterol (mmol/L)	15,679	2.87 (2.30, 3.52)	2.90 (2.35, 3.54)	2.92 (2.35, 3.54)	2.90 (2.33, 3.54)	2.79 (2.25, 3.41)	< 0.001
Total cholesterol (mmol/L)	15,679	4.89 (4.24, 5.61)	4.94 (4.32, 5.64)	4.94 (4.29, 5.61)	4.91 (4.22, 5.61)	4.78 (4.14, 5.51)	< 0.001
Total triglycerides (mmol/L)	15,679	1.14 (0.78, 1.66)	1.03 (0.71, 1.51)	1.07 (0.76, 1.57)	1.20 (0.84, 1.74)	1.24 (0.86, 1.82)	< 0.001

^aN not Missing (unweighted)

^bMedian (IQR) for continuous; n (weighted %) for categorical

^cDesign-based KruskalWallis test; Pearson’s X^2: Rao & Scott adjustment

^dMonocyte-Albumin-Ratio is grouped based on quartiles: Q1: < 0.095; Q2: 0.095–0.122; Q3: 0.122–0.152; Q4: > 0.152

Group analysis by MAR quartiles (Q1–Q4; n = 15,679) indicated that individuals in the highest quartile (Q4) were significantly older (median: 50 years compared to 45 years in Q1, P < 0.001), more frequently male (54% versus 41%), and exhibited a greater proportion of non-Hispanic White participants (72% versus 62%, all P < 0.001). In terms of socioeconomic factors, Q4 exhibited reduced educational attainment (53% compared to 65% with education beyond high school in Q1), a diminished median poverty-to-income ratio (2.41 versus 3.00), and an elevated median BMI (29 versus 26 kg/m², all P < 0.001). The Q4 group had markedly increased systolic blood pressure (131 vs. 125 mmHg), a greater prevalence of smoking (28% vs. 12%), diabetes (17% vs. 8.4%), cancer (11% vs. 9.2%), and a significant prevalence of Stroke (4.6% vs. 1.9% in Q1, all P < 0.05). The lipid profiles in Q4 exhibited decreased HDL (1.27 vs. 1.42 mmol/L), LDL (2.79 vs. 2.90 mmol/L), and total cholesterol (4.78 vs. 4.94 mmol/L), whereas triglycerides were elevated (1.24 vs. 1.03 mmol/L, all P < 0.001). In summary, elevated MAR levels were continuously linked to a heightened prevalence of Stroke and correlated with detrimental socioeconomic factors, inferior physiological profiles, and unfavorable metabolic traits, indicating that MAR may serve as a comprehensive risk marker.

Associations between MAR and the prevalence of Stroke in different glycemic status

Table 3 presents a key focus of this study: the association between MAR, TyG Index, and Stroke prevalence stratified by glycemic status (diabetes, prediabetes, non-diabetes). Logistic regression analyses employed three progressively adjusted models (Model 1: unadjusted; Model 2: partially adjusted; Model 3: fully adjusted, with covariates specified previously).

Table 3.

Correlation between the MAR and Stroke prevalence in different glycemic status: logistic regression analysis

Exposure	Model 1		Model 2		Model 3
	OR(95%CI)	P-value	OR(95%CI)	P-value	OR(95%CI)	P-value
Total
MAR	1.28 (1.13,1.46)	< 0.001	1.18 (1.07,1.30)	0.001	1.13 (1.03,1.24)	0.013
TyG index	1.61 (1.37,1.90)	< 0.001	1.25 (1.01,1.52)	0.024	1.07 (0.94,1.22)	0.3
Diabetes
MAR	1.27 (1.12,1.44)	< 0.001	1.27 (1.13,1.43)	< 0.001	1.23 (1.08,1.41)	0.003
TYG index	1.04 (0.8,1.35)	0.8	1.22 (0.89,1.68)	0.2	1.17 (0.84,1.64)	0.3
Pre-diabetes
MAR	1.15 (0.96,1.38)	0.12	1.12 (0.95,1.31)	0.2	1.11 (0.92,1.33)	0.3
TyG index	0.88 (0.64,1.20)	0.4	0.87 (0.61,1.23)	0.4	0.77 (0.49,1.20)	0.2
No-diabetes
MAR	1.32 (1.13,1.53)	< 0.001	1.18 (1.01,1.37)	0.032	1.08 (0.91,1.29)	0.4
TyG Index	1.18 (0.89,1.56)	0.2	0.89 (0.65,1.22)	0.5	0.88 (0.65,1.19)	0.4

Table data: β (95%CI) P-value/OR (95%CI) P-value

Outcome variable: Stroke

Model 1: no covariates were adjusted

Model 2 adjust for: gender, age, race,education

Model 3 adjust for: gender, age, race, education, PIR, BMI, HDL, LDL, smoking, drink, cancer

In the overall population, all MAR models demonstrated a significant positive association with Stroke prevalence. Each unit increase in MAR was associated with a 13% higher Stroke prevalence (adjusted OR = 1.13; 95% CI: 1.03, 1.24; P < 0.05). Although the TyG Index showed associations in minimally adjusted models (Models 1–2), it was not statistically significant in the fully adjusted model (Model 3) (aOR = 1.07; 95% CI: 0.94, 1.22; P > 0.05).

Stratified analysis revealed significant effect modification by glycemic status:

• Diabetes group: Each unit increase in MAR significantly increased Stroke prevalence by 23% (aOR = 1.23; 95% CI: 1.08, 1.41; P < 0.05). The TyG Index showed no significant association (P > 0.05).

• Prediabetes group: Neither MAR nor TyG Index was significantly associated with Stroke prevalence (all P > 0.05).

• Non-diabetes group: Both biomarkers lacked significant associations with Stroke prevalence (all P > 0.05).

Thus, MAR maintained a significant independent association with Stroke prevalence in the overall population, with a pronounced effect size observed specifically in individuals with diabetes. In contrast, the TyG Index showed no independent association after full adjustment.

Restricted cubic spline

In this study, we employed RCS curves to analyze the dose–response relationship between MAR and Stroke prevalence across groups: the overall population, individuals with diabetes, prediabetes, and non-diabetes.

As illustrated in Fig. 2:

• Overall population: MAR exhibited a positive nonlinear association with Stroke prevalence (nonlinear P = 0.041).

• Diabetes Group: The strongest association was observed, demonstrating a steep near-linear dose–response curve (nonlinear P = 0.013), indicating MAR demonstrates strong discriminatory value for Stroke prevalence in this population.

• Prediabetes Group: A weak, nonsignificant association was noted (nonlinear P = 0.675).

• Non-diabetes Group: The relationship remained nonsignificant with a flat curve (nonlinear P = 0.349).

Fig. 2.

Restricted cubic spline curves of MAR versus Stroke in patients with different blood glucose levels. A is the overall population, B is the diabetic population, C is the prediabetic population, and D is the nondiabetic population

These findings suggest MAR’s discriminative performance for Stroke prevalence is substantially enhanced in diabetic populations compared to other glycemic strata.

Hierarchical interaction analysis

Stratified interaction analysis was performed to assess the reliability and stability of the model, with results presented in a forest plot (Fig. 3). Variables such as age, gender, ethnicity, education, smoking status, and blood glucose subgroups were included. The analysis shows that these variables didn’t significantly impact the conclusion that"MAR increases Stroke prevalence"(interaction p-value > 0.05), indicating robust model stability.

Fig. 3.

Comparison of the MAR and Stroke subgroups across populations. We adjusted for age, gender, race, education, BMI, PIR, LDL, HDL, Alcohol Consumption, Smoking, and Cancer

Subgroups with notable significance are as follows:

1. Age: Significant in the 40—59 age group (OR = 1.21, 95%CI 1.05—1.98, P = 0.01) and nearly significant in the 60 + group (OR = 1.14, 95%CI: 1.01,1.29, P = 0.03).

2. Education: Significant in the below high school education group (OR = 1.26, 95%CI: 1.05,1.51, P = 0.01).

3. Diabetes status: Significant in the diabetic group (OR = 1.24, 95%CI: 1.09,1.41, P < 0.001).

4. Smoking status: Significant in previous smokers and current quitters (OR = 1.13, 95%CI: 1.01,1.26, P = 0.04).

5. Ethnicity: Significant in other Hispanics (OR = 1.55, 95%CI: 1.08,2.23, P = 0.02).

These subgroups had a more pronounced impact on the findings, further confirming the model’s reliability.

Receiver operating characteristic curve (ROC)

We conducted ROC curve analyses to evaluate the discriminatory ability of MAR, TyG Index, LDL, and MAR*TyG interaction, using both univariate and fully adjusted models. The fully adjusted model incorporated covariates including age, gender, race, education level, BMI, PIR, LDL, HDL, alcohol consumption, smoking status, and cancer history.

Key findings demonstrated in Fig. 4 include:

• Univariate models:

  MAR showed superior discriminatory ability for Stroke prevalence (AUC = 0.594; 95% CI: 0.570,0.618) compared to TyG Index (AUC = 0.583; 95% CI: 0.557,0.608) and LDL (AUC = 0.568; 95% CI: 0.545,0.591).

  The MAR*TyG interaction demonstrated the highest predictive ability (AUC = 0.610; 95% CI: 0.587,0.632).

• Fully adjusted models:

  All biomarkers showed comparable predictive power:

  MAR: AUC = 0.808 (95% CI:0.793,0.823).

  TyG Index: AUC = 0.806 (95% CI: 0.792,0.821).

  MAR*TyG Index: AUC = 0.808 (95% CI: 0.793,0.823).

  LDL: AUC = 0.806 (95% CI: 0.791,0.821).

  Narrow overlapping 95% CIs indicated minimal between-model differences.

Fig. 4.

For the ROC curve analysis, we built univariate and fully adjusted models for MAR, TyG Index, LDL, and MAR*TyG, respectively

Covariate adjustment substantially enhanced discriminatory ability (ΔAUC ≈ 0.21). These results demonstrate that while MAR shows advantage in univariate analysis, comprehensive prevalence factor adjustment attenuates biomarker-specific differences, yielding robust and stable prediction models.

Mediator analysis

Mediation analysis examining BMI as an intermediary variable was performed to clarify the pathway connecting MAR to Stroke prevalence (Tables 4, 5; Fig. 5). Regression analyses revealed three key findings: First, the pathway from MAR to BMI showed a significant positive effect (β = 14.589, P < 0.001), indicating that each unit increase in MAR elevated BMI by 14.589 units. Second, a significant direct association existed between MAR and Stroke (β = 1.881, P < 0.01), corresponding to a 1.881-unit increase in Stroke prevalence per MAR unit. Third, the pathway from BMI to Stroke demonstrated significant mediation (β = 0.705, P < 0.001), with each BMI unit increase raising Stroke prevalence by 0.705 units.

Table 4.

BMI as a mediator of the effect of the Monocyte Albumin Ratio to Stroke

Path	Coefficient	SE	95%CI	Effectiveness ratio
Total effect(c)	2.015	0.671	(0.704, 3.326)	–
Indirect effect (a*b)	0.314	0.094	(0.136, 0.508)	15.6%
Direct effect (c’)	1.701	0.667	(0.395, 3.008)	84.4%

Table 5.

BMI was tested for mediator of the effect of the Monocyte Albumin Ratio to Stroke

Variable	BMI	Stroke
Age	0.009** (0.003, 0.015)	0.064*** (0.057, 0.071)
Gender	2.357*** (2.140, 2.574)	0.075 (− 0.109, 0.260)
Race	− 0.197*** (− 0.287, − 0.064)	0.183*** (0.101, 0.265)
Education	0.146* (0.007, 0.285)	− 0.129* (− 0.237, − 0.021)
PIR	− 0.136*** (− 0.209, − 0.064)	− 0.194*** (− 0.261, − 0.127)
HDL	− 4.683*** (− 4.972, − 4.394)	− 0.367** (− 0.624, − 0.109)
TG	0.705*** (0.553,0.858)	−0.062 (−0.195,0.071)
Drink	− 0.260 (− 0.946, 0.427)	− 0.230 (− 0.916, 0.456)
Smoking	0.834*** (0.673, 0.996)	− 0.208** (− 0.338, − 0.078)
CANCER	− 0.401* (− 0.769, − 0.034)	0.369*** (0.153, 0.584)
MAR	14.589*** (12.583, 16.595)	1.881** (0.532, 3.230)
BMI	–	–
Constant	27.832*** (27.032, 28.632)	−0.692*** (− 7.775, − 6.091)
Observations	15,679	15,679
R2	0.124	–

^*: P < 0.05,**: P < 0.01,***: P < 0.001. The main content of the table is the SE. The 95% confidence interval is in brackets

Fig. 5.

Association between BMI, Monocyte Albumin Ratio and Stroke, a and b are the mediating effects of BMI, and c’ is the direct effect. *P < 0.05,**: P < 0.01,***: P < 0.001

The total effect of MAR on Stroke comprised 84.4% direct effect (β = 1.881) and 15.6% indirect effect mediated through BMI (β = 10.28). Bootstrap validation confirmed the significance of this mediation. In conclusion, MAR influences Stroke prevalence through both direct mechanisms and BMI-mediated pathways, with the indirect effect accounting for 15.6% of the total effect.

Discussion

The Stroke prevalence rate was 2.9% (597/15,679). Patients exhibited distinct socio-clinical phenotypic clustering, characterized by advanced age, low educational attainment, diminished quality of life, smoking history, and significant comorbidity burden (notably diabetes and cancer). This pattern corresponds with the social health gradient documented in the NHANES cohort [45], underscoring the imperative to integrate social determinants into Stroke prevention strategies. Notably, MAR demonstrated stronger univariate associations with prevalent Stroke than TyG Index. In mediation analysis, MAR explained 15.6% of Stroke prevalence through BMI (P < 0.001), suggesting BMI as a potential modifiable pathway for physical activity interventions. After full adjustment, MAR retained significant association strength in diabetics (adjusted OR = 1.23 per SD increase, 95% CI 1.11–1.37; P < 0.001). The highest MAR quartile (Q4) showed 2.5-fold greater Stroke prevalence versus Q1 (4.6% vs. 1.9%; P < 0.001), with a threshold of 0.152 yielding optimal discrimination (Youden index = 0.42). These findings position MAR as a candidate biomarker warranting further investigation for Stroke prevalence stratification in diabetic populations. However, clinical application requires: (1) Validation of the 0.152 cutoff in prospective cohorts with incident Stroke endpoints; (2) Confirmation of BMI-mediated effects in intervention studies; (3) Demonstration that MAR-guided management improves outcomes in randomized trials. Until such evidence is established, MAR should be considered a research tool rather than a clinical screening metric.

This paper is the first to examine and discuss the correlation between MAR and patients with different blood glucose levels, and to compare the ability of MAR and TyG Index to predict Stroke prevalence. We reviewed papers in this area, and there have been previous studies similar to, but not involving, MAR. Yong’An Jiang et al. [13] showed that there was a negative association between TyG Index and Stroke prevalence and that this association was causal by using two databases, NHANES and MMIC-IV, as well as a joint Mendelian randomization analysis. Yuankai Shao et al., on the other hand, analyzed the CHARLS database through a longitudinal study and found that there was a nonlinear positive correlation between TyG-BMI and the prevalence of Stroke, and at the same time, they concluded that the prevalence of Stroke could be reduced by changing the lifestyle habits in an appropriate way as well as by reasonable dietary control [18]. In contrast, also using the CHARLS database, researchers Longjie Qu, Shuang Fang, Zhen Lan et al. found that plasma atherosclerotic index (AIP) was significantly correlated with the analysis of new-onset Stroke in individuals with different glucose metabolisms [19], especially in diabetic and pre-diabetic populations, where higher levels of AIP imply a higher likelihood of Stroke. Jie Fu regression study of 246 patients with spontaneous cerebral hemorrhage found for the first time that the expansion of intracerebral hematoma in patients with spontaneous cerebral hemorrhage was significantly correlated with MAR, and that elevated MAR could be used as an independent predictor of hematoma expansion and mortality within one year [5]. It has been found that albumin is converted to ischemia-modified albumin (IMA) by chemical modification of the N-terminal sequence in ischemic states, and IMA is increased in acute ischemic Stroke (AIS), cerebral hemorrhage (ICH), and subarachnoid hemorrhage (SAH) [20], and that IMA has a high and accurate diagnostic value for Stroke [21].

Compared with the TyG Index, which is widely used in cardiovascular disease research, MAR can be regarded as an emerging indicator with some research potential. According to the results of univariate and multivariate logistic regression analyses and restricted cubic spline RCS curves, the relationship between the two tends to be an approximately linear positive correlation. In the following, we will analyze this phenomenon from two perspectives: first, from the perspective of the underlying theory. Monocyte infiltration and differentiation lead to a dose-dependent accumulation of inflammation and oxidative stress [22], while the antioxidant capacity of albumin decreases linearly with decreasing concentration [23], further exacerbating endothelial dysfunction and inflammatory imbalance [24]. Dynamic modeling reveals “dose–response” and “positive feedback loop” mechanisms [25], suggesting that changes in monocytes and albumin directly amplify the linear accumulation of vascular damage. From the clinical point of view, monocyte levels increase in a gradient with the degree of chronic inflammation, and the pro-inflammatory factors released by monocytes have a cumulative effect on vascular endothelial damage, whereas the antioxidant, anti-inflammatory, and microcirculatory functions of albumin diminish with decreasing concentration, which together reflect the “inflammation-metabolism syndrome” [26], whose severity correlates with the prevalence of Stroke in a positive linear fashion.

The results of logistic regression, categorized by glucose levels, showed that MAR had a significant correlation with Stroke in diabetic patients, compared to those with prediabetes and non-diabetes. When compared with the triglyceride-glucose index, neither in diabetic, prediabetic, nor non-diabetic populations was there a significant association between TyG Index and Stroke prevalence. This is a key finding of this study. Diabetic patients frequently exhibit elevated oxidative stress [46]. This oxidative stress triggers a significant increase in monocyte counts and activates these monocytes. Activated monocytes subsequently produce large quantities of reactive oxygen species (ROS) and inflammatory mediators. This cascade further damages vascular endothelial cells, promotes the formation and progression of atherosclerosis, and ultimately increases the prevalence of Stroke [27–29]. MAR, indicative of reduced albumin levels, is a common feature in diabetes. As albumin possesses significant antioxidant and anti-inflammatory properties, its depletion diminishes the body’s overall antioxidant capacity. This impairment exacerbates oxidative stress-induced vascular damage and contributes to an elevated Stroke risk [30]. Furthermore, genetic variants within components of the renin-angiotensin system (RAS) pathways may amplify this susceptibility to Stroke, particularly in cases of premature ischemic Stroke [50]. Moreover, diabetic populations often suffer from microvascular dysfunction, which may impair cerebral microcirculation and increase the likelihood of cerebral ischemia and Stroke. An abnormal monocyte-to-albumin ratio may be associated with microvascular endothelial dysfunction, which in turn affects cerebral blood supply and exacerbates the prevalence of ischemic Stroke [31] [32]. TyG Index is considered a surrogate for insulin resistance, and although TyG Index did not show significance in populations with different glycemic levels, it was still relevant in the overall population, and previous studies related to TyG Index have also been conducted. This stage-dependent variation in the TyG-Stroke association has been documented in prior studies [33, 34]. The underlying mechanism may relate to insulin resistance being more pronounced during early disease phases. With disease progression and complication onset, however, factors such as microvascular lesions and neuropathy likely assume greater pathological significance in Stroke development [35]. Consequently, the correlation strength between the triglyceride glucose index and Stroke demonstrates differential manifestations across disease stages. However, it is interesting to note that TyG Index was also not significant in non-diabetic and pre-diabetic population. Therefore, we considered that it might be due to the fact that only the blood glucose level of the patients was considered in grouping different blood glucose levels, whereas other baseline characteristics such as age, gender, and basal metabolic profile might also affect the relationship between TyG Index and Stroke, which failed to show up in different blood glucose groups due to excessive heterogeneity.

Finally, mediation analysis revealed that BMI mediates 15.6% of the association between MAR and Stroke prevalence (β = 0.314, 95% CI: 0.136–0.508), suggesting that MAR elevation coincides with BMI increase in Stroke prevalence. This phenomenon can be mechanistically interpreted as follows: Elevated MAR levels coincide with increased monocyte counts or decreased serum albumin [37]. Elevated monocytes augment inflammatory factor release (e.g., TNF-α, IL-6), which disrupts insulin signaling to induce insulin resistance, thereby promoting visceral adipose accumulation and BMI elevation [39]. Notably, oxidative stress-induced hypomethylation of AT1R promoters in hypothalamic nuclei—as demonstrated in diabetic offspring—amplifies sympathetic hyperactivity and hypertension [47], providing a parallel epigenetic mechanism for the MAR-BMI-Stroke axis. Concomitantly, adipose tissue macrophage infiltration in high-BMI individuals further secretes pro-inflammatory mediators, establishing a self-perpetuating"inflammation-obesity"cycle that exacerbates both MAR and BMI [40]. Although hypoalbuminemia may transiently increase BMI through reduced colloid osmotic pressure and tissue edema [38], this effect is typically non-persistent; in metabolic syndrome, chronic inflammation and dehydration drive albumin reduction, while compensatory synthesis in obesity may indirectly affect fluid homeostasis or energy utilization efficiency [41]. Critically, elevated BMI promotes hypertension, dyslipidemia, and hyperglycemia—established Stroke prevalence factors [42]—while adipose-derived hormones (e.g., leptin, lipocalin) provoke thromboinflammatory cascades and vascular dysfunction that magnify Stroke susceptibility [43, 44]. Targeting metabolic-inflammatory axes (e.g., PAQR3 ubiquitination) may offer novel therapeutic avenues, as demonstrated in diabetic complications [51]. To disrupt this pathogenic cascade, aerobic exercise emerges as a strategic intervention: it suppresses pro-inflammatory Th17 differentiation and improves vasodilation in diabetic models [47, 48], potentially counteracting the MAR-BMI-Stroke trajectory through immunometabolic modulation.

Because the NHANES database uses stratified, multistage probability sampling to collect data, the findings are more reliable and representative of the nation. In addition we considered confounding factors such as gender, age, race, education level, smoking status, alcohol consumption, blood pressure, cancer, and metabolic indicators such as LDL and HDL to ensure that our findings were not swayed by any one factor. In this study, through univariate and multivariate logistic regression analyses and RCS curves, we verified the approximately linear correlation between MAR and Stroke in the overall population as well as in the diabetic group, and found that the probability of Stroke in the overall population with a high MAR was significantly higher than that with a high level of the TyG Index, and the results of the ROC curves showed that MAR had a better Stroke prevalence than TyG Index demonstrated superior discriminatory ability for Stroke prevalence compared to TyG Index. Meanwhile, comparing the two composite indexes, MAR is simpler in terms of the combination of elements and the calculation method, and if we consider the economic conditions and convenience, MAR, which can be obtained from the most routine tests of blood and standard biochemical tests, is obviously more advantageous.

Regrettably, our investigation is also subject to several limitations. This study has several important limitations. Foremost, the cross-sectional design inherently precludes causal inference and is susceptible to reverse causation. Stroke itself triggers systemic inflammation characterized by prolonged monocyte activation (recruitment persisting for weeks) and albumin depletion (consumed during the acute phase response) [9, 11]. Consequently, elevated MAR levels could plausibly result from a Stroke event rather than predict it. Therefore, while our findings identify MAR as a promising biomarker associated with Stroke prevalence, its predictive utility for incident Stroke requires rigorous prospective validation. Specific methodological constraints include:

• Prevalent vs. incident Stroke: As detailed in the Methods, the outcome measure captures self-reported prevalent Stroke cases at the time of the NHANES survey. This design precludes conclusions regarding causality or the ability of MAR to predict future (incident) Stroke events.

• Stroke ascertainment: Stroke status relied solely on self-reported questionnaire data within NHANES, without verification through medical record adjudication.

• Data granularity: The dataset lacked sufficient granularity to: Differentiate ischemic from hemorrhagic Stroke subtypes; Distinguish first-ever from recurrent Stroke events; Precisely delineate the temporal sequence between biomarker measurement and Stroke onset.

• Residual confounding: Although extensive covariate adjustment was performed, the potential for residual confounding by unmeasured variables remains.

• Speculative mechanisms: Proposed biological mechanisms linking MAR to Stroke in diabetes remain speculative due to limited mechanistic data within this study and require validation in future research.

Despite these limitations, the strength and specificity of the association between elevated MAR and Stroke prevalence in diabetic individuals align with established pathophysiology: Chronic hyperglycemia promotes a pro-inflammatory monocyte phenotype [27], while diabetic hypoalbuminemia reflects both vascular leakage and oxidative consumption [30]. This synergy suggests MAR may be particularly sensitive to diabetic vasculopathy, even acknowledging that a history of Stroke might contribute to its elevation. To address the identified constraints, we strongly advocate for prospective cohort studies designed to validate the association between MAR and incident Stroke and to investigate the underlying biological mechanisms.

Conclusion

This cross-sectional NHANES analysis establishes a significant association between the MAR and prevalent Stroke, with particularly strong discriminatory performance observed in diabetic populations compared to TyG index. Notably, individuals within the highest MAR quartile (Q4) demonstrated a 2.5-fold increase in Stroke prevalence relative to the lowest quartile (Q1) (4.6% vs. 1.9%; P < 0.001), with an optimized discrimination threshold of 0.152 (Youden index = 0.42). Mediation analysis further identified BMI as a significant modifiable pathway, accounting for 15.6% of the MAR-Stroke association (P < 0.001). These findings collectively support MAR’s potential utility as a candidate biomarker for Stroke prevalence stratification in high-risk cohorts, particularly individuals with diabetes mellitus. Given the inherent limitations of the cross-sectional design, clinical translation requires addressing three critical research priorities through prospective investigation: Validation of the prognostic significance of the 0.152 MAR cutoff in longitudinal cohorts utilizing incident Stroke endpoints to establish temporality and mitigate reverse causality concerns; Confirmation of BMI-mediated mechanistic pathways within targeted intervention studies, particularly those examining modifiable lifestyle factors; Demonstration of clinical efficacy through adequately powered randomized controlled trials assessing whether MAR-guided risk management improves cerebrovascular outcomes in high-risk populations, especially diabetics. Prospective validation of these research priorities could explore MAR’s potential utility in prevalence stratification and as a research tool for monitoring therapeutic responses along the inflammation-metabolism axis. Such advancements hold promise for refining preventive neurology strategies, however, pending generation of robust longitudinal and interventional evidence, MAR currently remains confined to research applications and is not recommended for clinical screening purposes.

Abbreviations

MAR

Monocyte Albumin ratio

TyG Index

Triglyceride glucose index

ROC

Receiver Operating Characteristic Curve Analysis

NHANES

National Health and Nutrition Examination Survey

CDC

Centers for Disease Control and Prevention

PIR

Poverty-income ratio

Author contributions

LYH, LQX and WXP obtained, analyzed and interpreted the data, and drafted this article. All the authors have reviewed the manuscript and made reference suggestions.

Funding

This work was supported by grants from the National Nature Science Foundation of China (No. 81301180) and the Jiangsu Commission of Health (M2020046).

Data availability

Publicly available datasets were the focus of this investigation. The information can be accessed at this URL: https://www.cdc.gov/nchs/nhanes/.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.