Your Mouth Bacteria Can Predict Biological Age and Disease Risk
A new oral microbiome aging clock predicts mortality, frailty, and cancer risk beyond conventional factors using non-invasive saliva samples.
Resumen
Researchers analyzed oral microbiome data from nearly 6,000 Americans across two NHANES cohorts and developed a machine learning model that predicts biological age from 64 age-associated bacterial genera. The resulting Oral Microbiome Aging Acceleration (OMAA) Score — the gap between microbiome-predicted age and chronological age — independently predicted all-cause mortality, frailty, and impaired kidney function, while meaningfully improving risk prediction for cancer and heart attack beyond standard clinical factors. Validated in an external multi-country cohort, the OMAA Score was largely unaffected by diet or medications, suggesting it captures intrinsic biological aging rather than lifestyle noise.
Resumen detallado
Population aging is accelerating globally, yet chronological age poorly captures an individual's true physiological state. Biological age clocks — derived from epigenetics, blood biomarkers, or gut microbiome composition — have emerged as more informative metrics. This study makes the case that the oral microbiome, sampled non-invasively via oral rinse, can serve as an equally powerful and far more accessible window into biological aging.
The team analyzed 16S rRNA sequencing data from oral rinse samples in two sequential NHANES cohorts totaling 4,675 adults aged 30–70. Using generalized additive models adjusted for sex and race/ethnicity, they identified 64 bacterial genera whose abundance shifted significantly and reproducibly with age. Among the most informative: Rothia (enriched in older adults and previously linked to frailty) and Scardovia (depleted with age, potentially reflecting altered oral carbohydrate metabolism). Alpha-diversity consistently declined with age across all four metrics, while beta-diversity analysis confirmed significant compositional separation between younger and older participants.
A random forest model trained on these 64 genera predicted chronological age with a Spearman correlation of 0.44 in the discovery cohort and 0.35 in the internal validation cohort. Applied to an independent external dataset of 1,293 participants from multiple countries using 37 overlapping genera, the model still achieved a significant correlation (ρ = 0.22, P = 2.17×10⁻¹⁵), demonstrating cross-population generalizability. The OMAA Score — defined as the residual between microbiome-predicted age and actual chronological age — was then derived and tested against clinical outcomes.
Higher OMAA Scores independently predicted all-cause mortality (HR = 1.05, P = 0.024) and frailty (OR = 1.05, P = 0.008) after adjustment for chronological age and standard covariates. Kidney function, measured by eGFR, declined significantly with higher OMAA scores (β = −0.066, P = 5.22×10⁻⁴). Crucially, adding the OMAA Score to conventional risk factor models significantly improved discrimination for cancer (AUC 0.70 vs. 0.67, P = 0.009) and heart attack (AUC 0.79 vs. 0.76, P = 0.016). Notably, dietary patterns showed minimal association with OMAA scores, and medication use had only weak links, suggesting the score reflects deeper biological aging processes rather than modifiable lifestyle behaviors measured at a single time point.
The study positions the OMAA Score as a scalable, non-invasive tool for population-level risk stratification. Its reliance on a simple oral rinse rather than blood draw or stool sample lowers barriers to deployment in routine clinical and public health settings. Limitations include cross-sectional design (causality cannot be established), reliance on 16S rRNA sequencing rather than shotgun metagenomics (limiting functional resolution), and the fact that external validation required reducing the model to 37 genera, attenuating performance.
Hallazgos clave
- 64 oral bacterial genera shift significantly and reproducibly with age across two large NHANES cohorts (N=4,675).
- Random forest model predicts chronological age from oral microbiome with ρ=0.44 internally and ρ=0.22 in a multi-country external cohort.
- OMAA Score independently predicts all-cause mortality (HR=1.05) and frailty (OR=1.05) beyond chronological age.
- Adding OMAA Score improves cancer and heart attack risk prediction (AUC gains of +0.03 each, P<0.02).
- Higher OMAA Score correlates with lower eGFR, indicating impaired kidney function independent of other factors.
Metodología
Two cross-sectional NHANES cohorts (2009–2010 and 2011–2012; N=4,675 adults aged 30–70) provided oral rinse 16S rRNA sequencing data. A random forest model trained on 64 age-associated genera (identified via GAMs) predicted chronological age; the OMAA Score residual was then linked to mortality (Cox models), frailty (logistic regression), kidney function (linear regression), and disease risk (AUC comparison) with external validation in N=1,293 multi-country participants.
Limitaciones del estudio
The cross-sectional design prevents causal inference about whether oral microbiome changes drive aging outcomes or merely reflect them. External validation required reducing the model from 64 to 37 genera, weakening cross-population performance (ρ dropping to 0.22). 16S rRNA sequencing provides taxonomic but not functional resolution, and cohort-specific sequencing protocols may introduce batch effects.
¿Te ha gustado este resumen?
Recibe la última investigación sobre longevidad en tu bandeja de entrada cada semana.
Introduce tu correo electrónico para suscribirte: