The Actinobacteria-to-Proteobacteria (A/P) Ratio: A Novel Oral Microbial Marker for Metabolic Syndrome

Abstract

Introduction and aims

Oral microbial biomarkers for metabolic syndrome (MetS), in contrast to its well-characterized relationship with the gut microbiome, have not been well defined. This study aims to systematically identify composite oral microbial indices by analyzing a large-scale cohort.

Methods

We analyzed oral rinse samples from 3911 U.S. individuals (aged ≥ 14 years), integrating 16S rRNA sequencing with metabolic phenotyping. We used principal coordinates analysis and permutational multivariate analysis of variance to assess β-diversity separation, linear discriminant analysis effect size analysis to identify microbial signatures, multivariable-adjusted logistic regression with restricted cubic splines (RCS) to evaluate MetS associations, Cox proportional hazards models to analyze all-cause mortality in the MetS population, and stratified analyses to test for effect modification by age, BMI and medication use.

Results

MetS participants exhibited higher BMI and were older than non-MetS controls (P < .001). Oral microbiome β-diversity diverged significantly in MetS despite stable α-diversity, marked by enrichment of Actinobacteria and depletion of Proteobacteria (P < .001). In adjusted models, the highest tertile of Actinobacteria was associated with 42% higher odds of MetS (aOR = 1.42, 95% CI: 1.06-1.91), whereas the highest tertile of Proteobacteria was associated with 42% lower odds (aOR = 0.58, 95% CI: 0.43-0.79). Capturing this dysbiotic shift, the composite Actinobacteria-to-Proteobacteria (A/P) ratio demonstrated a superior predictive value than either phylum alone (aOR = 1.58, 95% CI: 1.08-2.30). RCS analysis demonstrated a linear association of the A/P ratio with MetS and all-cause mortality among individuals with MetS. Kaplan–Meier analysis confirmed significantly reduced survival in the high-A/P ratio group (P < .001).

Conclusions

This study provides the first population-level evidence that an elevated the A/P ratio is independently associated with MetS, suggesting its potential as a novel oral microbiomarker.

Clinical Relevance

These findings provide a non-invasive tool to assess metabolic status and lay a foundation for future research into targeted oral microbial interventions for MetS.

Introduction

Metabolic syndrome (MetS), a clinical constellation characterized by central obesity, hypertension, insulin resistance, and dyslipidemia, has shown a rapidly rising global prevalence over the past two decades. Notably, it substantially elevates the risks of type 2 diabetes, atherosclerosis, and non-alcoholic fatty liver disease, emerging as a major public health burden worldwide.^1,2 While traditional perspectives emphasize genetic susceptibility, energy metabolism dysregulation and chronic inflammation as core drivers of MetS,^3,4 emerging research on the human microbiome suggests that the gut-oral-systemic axis may contribute to MetS pathogenesis through immune modulation and metabolic dysfunction.5, 6, 7

The composition and function of gut microbial ecosystems are not only dynamic but also deeply influenced by host lifestyle and environmental exposures. This dysbiosis, often observable as reduced microbial diversity and skewed proportions of key taxa, is increasingly recognized as a major contributor to systemic metabolic dysfunction.8, 9, 10 A systematic review in patients with metabolic syndrome demonstrates that interventions modulating the gut microbiota – including probiotics, prebiotics, synbiotics, and fecal microbiota transplantation – can improve glycemic control and insulin sensitivity. These findings underscore the pathogenic role of gut dysbiosis and reveal the therapeutic potential of microbiome-targeted strategies for improving metabolic health.¹¹

As the entry point of digestion and the body's second-largest microbial reservoir, the oral microbiome displays specific ecological niche discrepancies in its association with MetS: subgingival plaque analyses indicate increased α-diversity in MetS patients,^5,12 whereas salivary microbiome studies report decreased diversity or non-significant differences.^13,14 This divergence suggests distinct regulatory mechanisms: subgingival microbiota may promote systemic pathology via local inflammation, whereas microbes distributed in the oral mucosa and saliva participate in cross-organ metabolic communication through the oral-gut axis.¹⁵ Although saliva has been a focal point for biomarker discovery, recent evidence highlights the superior utility of oral rinse-derived microbiota for population-level diagnostics.¹⁶ Distinct from the predominantly planktonic profile of saliva, oral rinses capture sessile communities by aggregating microorganisms shed from the tongue dorsum, buccal mucosa, and supragingival biofilms, which mitigates site-specific bias to yield an integrative microbial signal.¹⁷ This signal positions oral rinse profiling as a non-invasive and highly standardizable platform for identifying the microbial fingerprints of MetS at a population level.

Despite extensive efforts to identify microbial biomarkers, prior studies have yielded inconclusive evidence regarding the specific dysbiotic features associated with MetS. A Brazilian adult cohort (n = 66) reported elevated abundance of Proteobacteria (eg, Neisseria), whereas a children and adolescent cohort from Iowa (n = 72) observed enrichment of Bacteroidota (eg, Porphyromonas), Proteobacteria (eg, Moraxella), and Fusobacteriota (eg, Leptotrichia).^14,18 The observed heterogeneity may be attributed to key methodological factors. Initially, the Brazil study’s failure to fully account for lifestyle-driven confounders – such as dietary habits, smoking, and medication – clouds the ability to draw firm causal inferences. This oversight risks misinterpreting a simple co-occurrence as a genuine pathogenic driver.19, 20, 21 Beyond these analytical gaps, a sharp demographic divide further complicates cross-study comparisons. While the Brazilian data centered on an adult population, the Iowa research focused specifically on adolescents with optimal oral health. Such stark variations in life stage and ethnicity inevitably compromise the external validity of these collective findings. Consequently, to address the inherent variability of single-taxon analyses, recent microbiome research has shifted toward composite microbial indices. By integrating functionally or phylogenetically related taxa, these indices capture ecological shifts more robustly than individual microbial signatures. A well-established example is the Firmicutes/Bacteroidetes (F/B) ratio in gut microbiome studies, which reflects host energy harvest and metabolic status.^22,23 However, a comparable, ecologically grounded composite index for the oral microbiome in the context of MetS remains unexplored.

To address these gaps, we analyzed multicycle cross-sectional data from the National Health and Nutrition Examination Survey (NHANES), using a nationally representative sample of 3911 participants aged 14 years or older. Using standardized 16S rRNA gene sequencing and multivariable-adjusted models, we systematically examined the association between oral-rinse-sourced microbiome features and MetS. This study aims to identify key composite microbial indices in MetS and assess their independence from major confounders, providing the population-level evidence to inform mechanistic and translational research.

Methods

Study population and data sources

NHANES is a comprehensive national survey employing stratified, multi-stage probability sampling to obtain representative U.S. samples. The National Center for Health Statistics (NCHS) Research Ethics Review Board approved NHANES data collection (‘NHANES-NCHS Research Ethics Review Board Approval’, 2022). This study analysed NHANES data from 2009 to 2012 (n = 19,591). Among 9848 participants with oral microbiome data, exclusions comprised individuals lacking MetS diagnosis information (n = 5875), recent antibiotic users (n = 62). This yielded a baseline cohort of 3911 patients (Figure 1). The final survival analysis cohort comprised 1082 participants with MetS. An additional 32 individuals were excluded due to loss to follow-up. The NHANES Ethics Review Committee approved the study, and all participants provided written informed consent. The study conformed to STROBE guidelines (Appendix file B_STROBE_checklist).

Fig. 1.

Flowchart of the selection of study participants.

Outcome: MetS and all-cause mortality within the MetS population

The primary outcome was MetS, defined by the National Cholesterol Education Program Adult Treatment Panel III (NCEP ATP III) and the International Diabetes Federation (IDF) criteria as the presence of at least three of the following components: abdominal obesity (waist circumference thresholds: ≥ 90 cm for men, ≥ 80 cm for women), hypertension (systolic BP ≥ 130 mmHg, diastolic BP ≥ 85 mmHg, or use of antihypertensive medication), elevated fasting glucose (FPG ≥ 100 mg/dL or a diagnosis of diabetes/use of antidiabetic drugs), hypertriglyceridemia (TG ≥ 150 mg/dL or use of lipid-lowering medication), and low HDL-C (< 40 mg/dL for men and < 50 mg/dL for women).²⁴ Mortality outcomes, including vital status, specific cause of death, and duration of follow-up, were ascertained through linkage with the National Death Index (NDI). The primary endpoint of this study was all-cause mortality among participants with MetS. Follow-up spanned from the NHANES 2009-2012 baseline examination until death or the end of the study period (December 31, 2019).

Exposure variable: oral microbacteria diversity and abundance

The primary exposure variable was defined as the relative abundance of oral bacterial phyla, derived from the pre-processed NHANES dataset ‘Relative Abundance Phylum-DADA2 (Unassigned Taxa Removed)’. Oral rinse collection and laboratory processing were conducted by NHANES as follows: Oral rinse samples from the NHANES 2009-2012 cohorts were processed using the CDC standardised DNA extraction protocol. Participants were instructed to rinse their mouths with 10 mL of sterile saline and then expectorate into sterile collection tubes. Extracted DNA was distributed into 96-well plates (72 plates for 2009-2010 and 60 plates for 2011-2012), with each plate containing experimental samples, blank controls, and synthetic microbial community standards (oral and gut) to assess contamination risks and analytical reproducibility. Following sequencing, microbial profiles were generated by clustering reads into amplicon sequence variants (ASVs). Taxonomic classification was performed from phylum to genus levels by running the QIIME28 command, and relative abundance/read count tables were compiled for each hierarchical rank.²⁵ These tables were subsequently standardised across phylogenetic levels and analytical methods using a custom R script. Microbial diversity was evaluated using both α- and β-diversity measures. We quantified α-diversity, a measure of within-sample species heterogeneity that incorporates richness and evenness, using four indices: observed Amplicon Sequence Variants (ASVs), the Shannon-Wiener Index (SWI), the inverse Simpson index, and Faith’s Phylogenetic Diversity (FPD). All α-diversity analyses were performed using rarefied data normalized to 10,000 sequences per sample and averaged over 10 iterations to ensure stability. To assess between-sample compositional differences (β-diversity), we employed three distance metrics: Bray–Curtis dissimilarity (based on abundance profiles), unweighted UniFrac (considering only presence or absence of taxa), and weighted UniFrac (which additionally incorporates relative abundance into phylogenetic branch lengths). To quantify the core structural dysbiosis in MetS, characterized by the concurrent enrichment of Actinobacteriota and depletion of Proteobacteria, we established a composite microbial index – the A/P ratio, defined as the relative abundance of Actinobacteriota divided by that of Proteobacteria.

Covariates

The covariates encompassed demographic, lifestyle, and clinical domains. Demographic variables included age (RIDAGEYR, continuous), sex (RIAGENDR: male/female), race/ethnicity (RIDRETH1) and Body Mass Index (BMI). Lifestyle factors were operationalized as: (1) smoking risk (high: ≥ 100 cigarettes; low: < 100 cigarettes in the past year); (2) current alcohol consumption (ALQ101: yes/no); (3) sedentary behavior categorized per WHO guidelines as low-risk ( < 4 hr/day), moderate-risk (4-8 hr/day), or high-risk ( > 8 hr/day); and (4) dietary fiber patterns derived via energy-adjusted residuals. Specifically, total caloric intake (DR1TKCAL) served as the independent variable in a linear regression model predicting dietary fiber consumption (DR1IFIBE), with positive residuals indicating high-fiber-density diets and negative residuals denoting low-fiber-density diets. Clinical covariates included documented prescriptions for antidiabetic, antihypertensive, or lipid-lowering medications within the previous 30 days. Following established practices in chronic disease epidemiology, we categorized participants into three age strata: young adulthood (14-44 years), middle adulthood (45-65 years), and the elderly cohort ( > 65 years). Participants were classified by BMI category: non-overweight ( < 24.9 kg/m²), overweight (24.9-29.9 kg/m²), and obese ( ≥ 30.0 kg/m²). Periodontitis severity was classified according to the CDC/AAP 2012 criteria using full-mouth clinical attachment loss (CAL) and probing depth (PD). Mild periodontitis was defined as CAL ≥ 3 mm at ≥ 2 interproximal sites on adjacent teeth (mesial and distal surfaces of the same interproximal space) or PD ≥ 4 mm at ≥ 2 non-adjacent sites (excluding pseudopockets from gingival edema). Moderate cases required CAL ≥ 4 mm at ≥ 2 interproximal sites or PD ≥ 5 mm at ≥ 2 non-adjacent sites. Severe periodontitis was characterized by CAL ≥ 6 mm at ≥ 2 interproximal sites with simultaneous PD ≥ 5 mm at ≥ 1 corresponding site. Data entries with CAL/PD coded as 99 or NA were excluded to address measurement errors and implausible values.

Statistical analysis

All statistical analyses were performed using R (version 4.3.1) and SPSS (version 26.0). To handle missing data in the covariates, we employed a machine learning approach using the random forest algorithm for multiple imputation. Only covariates with missing values were imputed. Continuous variables are presented as weighted mean ± standard deviation (SD) or median and interquartile range (Q1, Q3), and categorical variables as weighted frequency (%). Nationally representative analytic weights, derived from WTSAF2YR/2 in accordance with NHANES protocol guidelines, were applied to account for the complex multistage probability cluster sampling design. Intergroup comparisons of baseline characteristics were assessed using weighted ANOVA and the Mann-Whitney U test for continuous variables and Rao-Scott corrected χ² tests for categorical variables.

Principal coordinates analysis (PCoA) was employed to visualize microbiome beta-diversity based on distance matrices, including Bray-Curtis dissimilarity, unweighted UniFrac, and weighted UniFrac. Differences in microbial community composition (beta-diversity) between individuals with and without MetS were formally tested using permutational multivariate analysis of variance (PERMANOVA), implemented via the adonis2 function (10,000 permutations). To identify microbial taxa exhibiting differential abundance between defined groups, we applied the Linear Discriminant Analysis Effect Size (LEfSe) biomarker discovery algorithm. This method utilizes the non-parametric Kruskal-Wallis rank-sum test (α = 0.05) to screen for features with significant abundance differences, followed by Linear Discriminant Analysis (LDA) to estimate the effect size of these discriminative features (reported as LDA score > 2.0). We summarized the results using phylogenetic cladograms to depict significant enrichments in taxa across different taxonomic levels.

The associations between individual alpha-diversity metrics, oral microbiome abundance, the Actinobacteria-to-Proteobacteria (A/P) ratio and MetS were evaluated through weighted univariate and multivariable logistic regression analyses. Multivariable weighted logistic regression models employed hierarchical covariate adjustment: Model 0 examined microbial exposures alone; Model 1 added sociodemographic factors (age, sex, race/ethnicity) and BMI; Model 2 further incorporated lifestyle factors (dietary patterns, smoking status, alcohol intake, sedentary behavior) and systemic medications (antihypertensives, antidiabetics, lipid-lowering agents). Weighted restricted cubic splines (RCS) were used to model potential nonlinear dose-response relationships. Cox proportional hazards regression models were fit to compute hazard ratios (HRs). All-cause mortality risk among patients with MetS was analyzed using Kaplan-Meier survival curves with log-rank testing for significance. Effect estimates are reported as odds ratios (OR), hazard ratios (HR), or beta coefficients with corresponding 95% confidence intervals (CI). To satisfy the distributional assumptions of parametric regression models, relative abundances were included as continuous variates only after a log-transformation.

Subgroup and sensitivity analyses

To assess the robustness and generalizability of the primary findings, we performed a series of prespecified subgroup and sensitivity analyses. To identify potential effect modifiers, stratified analyses and interaction testing were conducted across subpopulations defined by age, BMI, and the use of metabolic medications (antidiabetics, lipid-lowering agents, and antihypertensives). Potential effect modification was assessed by including multiplicative interaction terms in logistic regression for MetS and Cox models for all-cause mortality. All P for interaction were derived from the fully adjusted models (Model 2). Statistical significance was defined as a two-sided P < .05.

Sensitivity analyses were further executed to substantiate the stability of the association between the A/P ratio and MetS against potential confounding and bias. First, to address age-dependent socioeconomic factors (education, marital status, and family income-to-poverty ratio), which contained substantial missing data and were inapplicable to adolescents, we performed a restricted analysis among adults (aged ≥ 20 years) with full adjustment for these covariates. Second, to mitigate bias from multiple imputation, we conducted a complete-case analysis (n = 1829) restricted to participants with no missing data. Finally, to mitigate residual confounding from oral health status, the models were further adjusted for periodontal condition and rinsing frequency within the complete-case dataset, addressing the 30% missingness in periodontal diagnosis observed in the main model.

Results

Baseline characteristics

This cross-sectional analysis included 3911 participants, of whom 1114 (28.5%) were classified with MetS. Individuals with MetS exhibited markedly higher BMI (33.28 ± 6.78 vs 26.14 ± 5.58 kg/m²; P < .001) and were older (48.68 ± 14.20 vs 35.12 ± 15.70, P < .001) compared to those without MetS. High-risk smoking was more prevalent in the MetS group (49.53% vs 39.34%; P < .001). Pronounced disparities were observed in medication use, with significantly higher rates of antidiabetic (13.37% vs 1.16%; P < .001), antihypertensive (31.70% vs 3.71%; P < .001), and statin use (12.66% vs 1.17%; P < .001) among MetS participants. No significant differences were detected in gender distribution, ethnicity, alcohol consumption, sedentary behavior, or dietary patterns (Table 1).

Table 1.

Baseline characteristics of the study population.

	Overall	Without MetS	With MetS	P
Unweitghted number	3911	2797	1114
Weighted number	166,502,076	119,937,508	46,564,568
BMI	28.17 ± 6.76	26.14 ± 5.58	33.28 ± 6.78	<.001
Age	38.98 ± 16.46	35.12 ± 15.70	48.68 ± 14.20	<.001
Gender				.231
female	1905 (48.33%)	1365 (49.17%)	540 (46.17%)
male	2006 (51.67%)	1432 (50.83%)	574 (53.83%)
Race				.088
Mexiccan American	730 (10.16%)	506 (9.95%)	224 (10.68%)
Non-Hispanic Black	885 (12.61%)	646 (12.85%)	239 (11.99%)
Non-Hispanic White	1431 (63.90%)	1002 (63.07%)	429 (66.02%)
Other Hispanic	406 (5.75%)	273 (5.74%)	133 (5.76%)
Other Race	459 (7.59%)	370 (8.38%)	89 (5.54%)
Alcohol_use				.048
Current drinker	924 (18.30%)	632 (17.29%)	292 (20.90%)
Non-current drinker	2987 (81.70%)	2165 (82.71%)	822 (79.10%)
Smoking_risk				<.001
High-risk smoker	1578 (42.19%)	1017 (39.34%)	561 (49.53%)
Low-risk smoker	2333 (57.81%)	1780 (60.66%)	553 (50.47%)
Sedentary_risk				.35
High	1249 (35.29%)	935 (36.09%)	314 (33.24%)
Medium	1504 (38.83%)	1074 (38.10%)	430 (40.70%)
Low	1158 (25.88%)	788 (25.81%)	370 (26.07%)
diet_pattern				.387
High_Fiber_Diet	2049 (53.28%)	1476 (53.99%)	573 (51.46%)
Low_Fiber_Diet	1862 (46.72%)	1321 (46.01%)	541 (48.54%)
Antidiabetic				<.001
NO	3681 (95.43%)	2757 (98.84%)	924 (86.63%)
YES	230 (4.57%)	40 (1.16%)	190 (13.37%)
Antihypertensive				<.001
NO	3450 (88.46%)	2682 (96.29%)	768 (68.30%)
YES	461 (11.54%)	115 (3.71%)	346 (31.70%)
Statins				<.001
NO	3765 (95.62%)	2773 (98.83%)	992 (87.34%)
YES	146 (4.38%)	24 (1.17%)	122 (12.66%)

Note: There was no missing value for age, gender, race, antidiabetic, antihypertensive, or statins use. The missing rate was 0.20% for BMI, 19.8% for alcohol use, 16.6% for smoking risk, 0.5% for sedentary risk, 2.8% for diet pattern. Continuous variables were presented as weighted mean ± standard deviation (SD), and categorical variables as weighted proportions (%) with nationally representative sample weights applied, in accordance with NHANES guidelines (eg, WTSAF2YR). All statistical estimates accounted for the complex survey design through integrated sample weighting.

A total of 1082 patients with MetS were included in the survival cohort. Over a median follow-up of 109 months, 77 all-cause death events were identified, and 1005 individuals survived through the follow-up period.

Microbial diversity profile in MetS: similar alpha diversity with significantly separated beta diversity

No significant differences in α-diversity were observed between groups across ASV richness, Shannon, Simpson, or Faith’s Phylogenetic Diversity indices (all P > .05) (Appendix Figure 1). In contrast, β-diversity analysis revealed significant microbial community separation based on Bray-Curtis dissimilarity (R² = 0.003, P = .001), unweighted UniFrac (R² = 0.004, P = .001), and weighted UniFrac distances (R² = 0.004, P = .001; Figure 2A-C), indicating that MetS is associated with structural reorganization, rather than overall loss of microbial diversity.

Fig. 2.

Comparative analysis of oral microbiome profiles between participants with and without MetS. (A-C) Comparison of oral microbial beta diversity between MetS and non-MetS groups; (D) Significantly differential microbial taxa identified by LEfSe analysis (LDA score > 2.0); (E) Cladogram generated from LEfSe analysis showing microbial taxa significantly enriched from phylum (p_) to genus (g_) level in each group. (F) Bar chart of the detection rates of bacterial phyla in the oral-rinse microbiota. Phyla detected in ≥ 90% of samples were categorized as the core microbiota (purple), whereas those with detection rates below 90% were labeled as low-prevalence microbiota (green). (G) Box plots compare the relative abundance of the five most prevalent phyla between MetS and non-MetS groups. Statistical significance was determined by the Mann-Whitney U test with Benjamini-Hochberg correction (***P < .001).

Oral-rinse microbial dysbiosis characterized by actinobacteria enrichment and proteobacteria depletion in patients with MetS

To identify specific microbial taxa associated with MetS, LEfSe analysis identified differentially abundant lineages. The MetS group exhibited significant enrichment within the phylum Actinobacteria and its phylogenetic descendants (Actinobacteria/Micrococcales/Micrococcaceae/Rothia, LDA > 2, P < .05). Conversely, controls showed enrichment in the phylum Proteobacteria and its lower taxonomic units (β-Proteobacteria/Pasteurellales/Pasteurellaceae/Haemophilus, LDA > 2, P < .05; Figure 2D, E), suggesting taxonomic conservation of disease-associated shifts at the phylum level. Analysis of 9848 oral-rinse samples detected 41 bacterial phyla, with 5 phyla (Firmicutes, Bacteroidetes, Proteobacteria, Actinobacteria, Fusobacteria) exhibiting > 90% prevalence. In the stringently quality-controlled samples (n = 3911), MetS patients demonstrated significantly elevated relative abundance of Actinobacteria (median [IQR]: 0.15 [0.10-0.20] vs 0.13 [0.10-0.19]; P < .001) and significantly reduced abundance of Proteobacteria (median [IQR]: 0.10 [0.05-0.15] vs 0.11 [0.06-0.16]; P < .001) compared with controls (Figure 2F, G). No significant differences were observed for other phyla.

Multivariable-adjusted logistic regression analyses further delineated the association of microbiome phylum abundance and MetS. The highest tertile (T3) of Actinobacteria abundance was independently associated with 42% higher odds of MetS (adjusted odds ratio [aOR] = 1.42, 95% confidence interval [95% CI]: 1.06-1.91; P = .023; Appendix Table 1). Conversely, the highest tertile (T3) of Proteobacteria abundance was associated with 42% lower odds (aOR = 0.58, 95% CI: 0.43-0.79; P = .002; P_trend = .002; Appendix Table 1). RCS models confirmed a linear positive association for Actinobacteria (P_nonlinearity = .093) and a significant linear inverse association for Proteobacteria (P_nonlinearity = .456) with MetS (Appendix Figure 2). These results establish that the inverse alterations in Actinobacteria (enrichment) and Proteobacteria (depletion) represent hallmark microbial features of MetS, strongly indicating core structural dysbiosis within the MetS microbiome.

A/P ratio as an independent associated factor for MetS and its prognosis

Quantification of oral microbial dysbiosis through the A/P ratio revealed a significant association with MetS. Individuals in the highest A/P ratio tertile demonstrated 58% higher prevalence of MetS across fully adjusted models (adjusted OR = 1.58, 95% CI: 1.08-2.30; P = .021; Table 2). Restricted cubic spline analysis confirmed a linear exposure-response relationship (P_nonlinearity = .756, P_overall = .001; Figure 3A). Notably, the F/B ratio, an established marker in gut microbiome studies, showed no significant association with MetS (Table 2). To determine whether the results depended on the chosen quantile cutoff, a sensitivity analysis was performed using quartile-based categories of the A/P ratio. Consistent with the primary analysis, participants in the highest quartile still showed a significantly higher prevalence of MetS (aOR = 1.71, 95% CI: 1.11-2.63, P = .018), and a similar pattern was observed for mortality risk (Appendix Table 2).

Table 2.

Independent associations of A/P ratio with MetS and all-cause mortality among individuals with MetS.

Oral bacreria abundance	Model 0		Model 1		Model 2
	OR (95% CI)	P	OR (95% CI)	P	OR (95% CI)	P
A/P ratio*
continuous variable	1.18(1.08,1.29)	<.001	1.21(1.09,1.33)	.001	1.15(1.03,1.28)	.015
T1:[0.00, 0.85)	Ref		Ref		Ref
T2:[0.85, 2.12)	1.01(0.79,1.28)	.950	1.18(0.88,1.58)	.268	1.16(0.82,1.65)	.374
T3:≥2 .12	1.51(1.13,2.01)	.007	1.67(1.19,2.32)	.004	1.58(1.08,2.30)	.021
P for trend	.006		.004		.018
F/B ratio*
continuous variable	1.06(0.91,1.23)	.466	1.05(0.87,1.26)	.611	1.03(0.86,1.23)	.748
T1:[ 0.03-2.13)	Ref		Ref		Ref
T2:[ 2.13-3.42)	1.03(0.82,1.29)	.779	1.02(0.81,1.30)	.843	0.99(0.77,1.27)	.946
T3:≥ 3.42	0.97(0.76,1.24)	.827	0.96(0.69,1.33)	.795	0.94(0.65,1.36)	.741
P for trend	.826		.795		.742

Oral bacreria abundance	Model 0		Model 1		Model 2
	HR (95% CI)	P	HR (95% CI)	P	HR (95% CI)	P
A/P ratio†
continuous variable	1.29(1.09,1.53)	.003	1.25(1.08,1.45)	.003	1.29(1.02,1.63)	.036
T1:[0.00-0.93)	Ref		Ref		Ref
T2:[0.93-2.68)	0.41(0.10,1.71)	.221	0.39(0.10,1.53)	.176	0.39(0.11,1.44)	.158
T3:≥2.68	1.53(0.89,2.62)	.125	1.55(0.84,2.87)	.162	1.34(0.73,2.49)	.348
P for trend	.107		.120		.276

Note: Model 0 examined oral microbiota phylum abundance, Model 1 added sociodemographic (age, sex, race/ethnicity) and BMI, and Model 2 further included lifestyle (dietary patterns, smoking status, alcohol consumption, sedentary behavior) and antihypertensive/antidiabetic/statins medications. The F/B ratio is defined as the Firmicutes-to-Bacteroidetes ratio.

^⁎Analysis of oral microbiome abundance in relation to MetS prevalence.

^†Analysis of oral microbiome abundance in relation to all-cause mortality among individuals with MetS.

Fig. 3.

Dose-response relationships of the A/P ratio with MetS and all-cause mortality in the MetS population. (A-B) RCS showing the association of the A/P ratio with MetS and all-cause mortality in the MetS population after adjustment for covariates in Model 2. The solid curve represents adjusted odds ratios or hazard ratios with 95% confidence intervals (shaded area). (C) Kaplan–Meier survival curves depicting cumulative mortality across strata defined by the A/P ratio.

Cox regression analysis revealed a significant association between the A/P ratio and MetS prognosis. A significant association was observed between the A/P ratio and mortality, with each 1-unit increment corresponding to a 29% higher risk of death (HR = 1.29, 95% CI: 1.02-1.63, P = .036; Table 2). In contrast, categorization into tertiles showed no significant association. RCS modeling confirmed a linear dose–response relationship with all-cause mortality among individuals with MetS (P_nonlinearity = .193; P_{overall association} = .007; Figure 3B). Consistently, Kaplan–Meier analysis demonstrated markedly improved survival in the low A/P ratio group (log-rank P < .0001; Figure 3C), reinforcing A/P ratio’s prognostic relevance.

Subgroup analyses and sensitivity verification

To evaluate the robustness of A/P ratio across subpopulations and identify effect modifiers, we performed stratified analyses and interaction testing. Notably, the A/P ratio consistently demonstrated directionally concordant positive associations with both MetS and all-cause mortality across all prespecified subgroups stratified by age, BMI, and metabolic medication use (antidiabetic, lipid-lowering, antihypertensive; Figure 4). Importantly, no significant interactions were detected for any metabolic medication class (antidiabetics P_interaction = .452; lipid-lowering agents: P_interaction = .318; antihypertensives: P_interaction = .601), indicating current metabolic pharmacotherapy does not mitigate the pathogenic effect of elevated A/P ratios. Conversely, significant interactions emerged for age (P_interaction = .041) and BMI (P_interaction = .003). RCS modeling further delineated that the significant positive association between A/P ratio and MetS was exclusively evident in participants aged ≤ 64 years and those with BMI ≥ 25 kg/m² (P_overall < .05, Appendix Figure 3).

Fig. 4.

Subgroup analyses of the association between A/P ratio with MetS and all-cause mortality among individuals with MetS. Note: A/P ratio was categorized into tertiles (T1, T2, T3), with the lowest tertile (T1) as reference. P-values indicate between-group differences and interaction effects. Horizontal lines indicate 95% CI.

Sensitivity analyses substantiated the stability of the primary association between the A/P ratio and MetS. After restricting the cohort to participants aged ≥ 20 years and incorporating additional socioeconomic adjustments including education, marital status, and family income, individuals in the highest A/P ratio tertile maintained a significantly higher MetS prevalence (aOR = 1.65, 95% CI: 1.10-2.48, P = .021, Appendix Table 3). Mortality results were consistent with the primary analysis: each 1-unit increase in the A/P ratio corresponded to a 23% higher risk of all-cause mortality among individuals with MetS (HR = 1.23, 95% CI: 1.00-1.50, P = .047).

In a complete-case cohort (n = 1829) excluding subjects with imputed data, the association remained statistically significant. Individuals in the highest A/P ratio tertile exhibiting an 83% higher prevalence of MetS (adjusted OR = 1.83, 95% CI: 1.06-3.16, P = .032; Appendix Table 4). Importantly, the association remained virtually unchanged after additional adjustment for periodontal status and mouthwash use characterized by extensive missing data in the primary cohort, confirming the stability of the association against both imputation-related bias and residual confounding from oral health factors.

Discussion

MetS, a cluster of conditions increasing the risk of cardiovascular disease and type 2 diabetes, arises from complex multisystem pathophysiology. Using large-scale cross-sectional data from NHANES, we demonstrate for the first time in a population-based cohort that oral-rinse microbiota – marked by elevated Actinobacteria, reduced Proteobacteria – is independently associated with MetS. This finding advances beyond previous research focused on isolated microbial changes by quantifying reciprocal phylum-level alterations and establishing the A/P ratio as a core biomarker of metabolic health.

Consistent with established findings, individuals with MetS showed clear separation in β-diversity without significant change in α-diversity, indicating a structural reshuffling – rather than a loss of microbial diversity. We identified a conserved dysbiotic profile characterized by enrichment of Actinobacteria (including genera such as Rothia) and depletion of Proteobacteria, consistent with a recent study.¹³ In contrast, a Brazilian cohort (n = 66) identified significantly increased Bacteroidetes abundance in the oral microbiome of MetS patients.¹⁴ Critically, our analysis revealed no significant correlation between Bacteroidetes abundance and MetS in the main effect model. Regional nutritional habits likely drive this mismatch. Specifically, distinct local diets may selectively amplify Bacteroidetes-linked carbohydrate-active enzyme (CAZyme) activity, thereby introducing region-specific phenotypic biases.^26,27 By utilizing the massive, multi-ethnic NHANES cohort (n = 3911) and employing multivariable-adjusted models, our study offers a stronger and more generalizable characterization of oral dysbiosis in MetS.

Most current research focuses on absolute abundance changes of individual microbial taxa, overlooking the potential role of inter-taxa ratios as independent exposure variables. This unidimensional analytical approach may underestimate the complex interactions within microbial communities and their collective impact on host metabolic homeostasis. Therefore, we specifically investigated key phylum-level ratios as candidate exposure variables in association analyses. In contrast to the well-established F/B ratio in gut-related metabolic studies, we found no association between oral F/B ratio and MetS – consistent with prior oral microbiome research. Instead, the A/P ratio emerged as a definitive dysbiosis index. Notably, traditional alpha-diversity indices (eg, Shannon, Simpson) failed to discriminate MetS status in this cohort. Given the cost-effectiveness and accessibility of oral-rinse collection, this ratio could be deployed as an early warning signal for metabolic dysregulation in large-scale population screenings.

Oral actinomycetes serve as early colonizers in dental plaque, capable of polysaccharide degradation and nitrate/nitrite reduction – key processes that modulate host nitric oxide (NO) signaling and contribute to blood pressure regulation and systemic metabolic homeostasis.28, 29, 30, 31 However, their overgrowth is linked to periodontitis and may exacerbate metabolic disorders such as obesity and diabetes through chronic inflammation.³² In contrast, certain genera within the phylum Proteobacteria (eg, Haemophilus), which also participate in nitrogen metabolism, play crucial roles in energy harvest and immune modulation.^30,33 Their depletion has been associated with an elevated risk of MetS and cardiovascular disease, suggesting a role as potential biomarkers of metabolic health.³⁴ Importantly, neither phylum is purely beneficial or detrimental. Their functional equilibrium, reflected in the A/P ratio, is critical for metabolic health.

Stratified analyses revealed a pronounced association between an elevated A/P ratio and MetS among participants under 65 years of age and those with a body mass index ≥ 25 kg/m², suggesting heightened metabolic susceptibility in these subgroups. This effect may be partly attributable to age and BMI related alterations in the oral microbiome. Age-related alterations include a general reduction in microbial diversity and a specific decrease in Actinobacteria (including Rothia), resulting in an age-dependent decline in the A/P ratio that may mask the contribution of microbial dysbiosis to MetS pathogenesis in older adults.³⁵ Current evidence indicates a significant association between BMI and alterations in the composition and diversity of the oral microbiome. Studies have revealed distinct microbial profiles in overweight and obese individuals compared to those with normal weight, including marked shifts in the relative abundance of core bacterial phyla, such as Firmicutes, Bacteroidota, and Proteobacteria.^36,37 Notably, the association between the A/P ratio and MetS persisted independently of conventional metabolic drug use, indicating that it may represent a causal or contributing factor independent of pharmacological intervention. These findings underscore the necessity of complementing standard metabolic management with targeted modulation of the oral microbiome, particularly the A/P ratio, to achieve more comprehensive risk reduction in MetS.

Several limitations should be noted. The cross-sectional design precludes causal inference. Prospective studies are needed to confirm the temporal relationship between A/P imbalance and MetS onset. Furthermore, 16S rRNA sequencing limits functional insights. Future integrated metagenomic and metabolomic analyses will be essential to elucidate mechanistic pathways, such as nitrate reductase deficiency in Proteobacteria or glycoside hydrolase activity in Actinobacteria, within the oral-gut systemic axis.

Conclusion

In this study, utilizing large-scale population data from NHANES, we identified the A/P ratio as an independent microbial biomarker significantly associated with MetS and all-cause mortality in the MetS population. These findings refine the ecological understanding of oral-systemic metabolic crosstalk and provide novel evidence for conserved phylum-level microbial signatures in MetS. The present study is subject to inherent limitations, including its cross-sectional design and the taxonomic resolution of 16S rRNA sequencing, both of which preclude direct causal inference. Future studies should leverage integrated multi-omics and prospective designs to elucidate mechanistic pathways and evaluate microbiota-targeted interventions within the oral-gut metabolic axis.

Ethics approval and consent to participate

This study complies with the ethical principles of the Declaration of Helsinki. Analysis was performed on publicly accessible, anonymized NHANES data that adheres to international ethical standards. All NHANES protocols were approved by the National Center for Health Statistics Institutional Review Board (IRB), with written informed consent obtained from all participants prior to data collection.

Availability of data and materials

Processed data supporting this study's findings are available from the corresponding author upon reasonable request. Original NHANES datasets are publicly accessible at: http://www.cdc.gov/nhanes.

Funding

This research received no external funding.

Author contributions

Jie Zhang: Contributed to design, acquisition, analysis, or interpretation of data; drafting and critical revision of the manuscript. Junwen Huang: Contributed to acquisition, analysis of data; drafting and critical revision of the manuscript. Yan Li: Contributed to interpretation of data and critical revision of the manuscript. Xiaoshan Wu & Feng Guo: Contributed to conception, design, and critical revision. All authors gave their final approval and agree to be accountable for all aspects of the work.

Conflict of interest

None disclosed.