Body’s biomarker network reorganizes dramatically at midlife in Mexican women, study finds

The network of correlations linking blood pressure, blood sugar, cholesterol and other cardiometabolic markers in the human body does not fray gradually with age. According to a new analysis of nearly 8,000 Mexican adults, it collapses in women in a single, sharply defined transition between the ages of 30 and 49, after which the architecture holds steady for two further decades. Men, by contrast, show no comparable upheaval at any point across the adult lifespan examined, a sexual dimorphism the authors describe as extending to general adult populations a pattern previously seen only in studies of strictly healthy young adults.

The study, published in Frontiers in Physiology, was led by Julian B. Uriarte-Ortiz and colleagues at the National Autonomous University of Mexico and other Mexican institutions. The team pooled data from the Mexican National Health and Nutrition Surveys (ENSANUT) of 2016, 2020 and 2021, yielding an analytic sample of 7,999 adults aged 20 to 91 years, of whom 5,058 were women and 2,941 were men. Participants with incomplete data were excluded, and only variables measured in all three surveys entered the analysis, producing 12 directly measured variables from which ten further indices were derived, for a total of 22 anthropometric, hemodynamic and biochemical measures per person.

Rather than tracking each biomarker one at a time, the researchers built correlation networks. For each combination of sex and one of five age strata, they computed Spearman correlations among all 22 variables, retained those reaching statistical significance at p < 0.05, and squared the coefficients so that only positive edge weights remained. To gauge stability, they repeated the entire procedure on 100 random subsamples of 79 percent of each stratum and averaged four topology metrics: density, the proportion of possible connections actually present; transitivity, the proportion of connected triples that close into triangles; mean distance, the average shortest path between nodes; and Louvain modularity, a measure of how strongly the network partitions into separate modules.

The results in women were striking. Between the 20–29 and 40–49 age groups, network density fell from 0.47 to 0.25, meaning the older network retained barely half the connections of the younger one. Transitivity dropped from 0.70 to 0.50, mean distance lengthened from 1.61 to 2.32, and modularity climbed from 0.40 to 0.56. In plain terms, the young female network was dense, redundant and tightly interwoven, with many alternative routes between any two biomarkers; the midlife network was sparse, compartmentalized and long-pathed. After the 40–49 stratum, all four metrics plateaued, fluctuating only within narrow bands through the oldest group, which reached age 91.

The structural anatomy of the shift was equally specific. In the youngest women, hemodynamic variables sat integrated within the main body of the network, with direct edges linking mean blood pressure to weight, the insulin-sensitivity proxy METS-IR, and insulin itself. From age 40 onward, the set formed by systolic, diastolic, mean and pulse blood pressure detached into a distinct module, connected to the glycemic and lipid core by only a reduced number of edges. Throughout all female strata, the most central nodes—by degree, eigenvector and betweenness centrality—were consistently METS-IR, insulin, weight and the iHOMA2 insulin indices, with directly measured variables such as weight and systolic pressure also recurring among the most important nodes.

Men told a different story. Male density stayed within a band of 0.34 to 0.42 across every age stratum, a range narrower than the change in female density between any two consecutive female strata during the first phase of the transition. Male transitivity, mean distance and efficiency likewise fluctuated without direction. The only directional change in men was modularity, which rose from 0.37 to 0.45 between the two youngest groups and then held steady. Notably, that early modularity step was shared by both sexes—women rose from 0.40 to 0.45 over the same interval, an almost identical magnitude—but from that point the trajectories diverged, with women continuing to 0.56 while men stopped. Men began adulthood with a less dense, less modular network than women and ended with a denser, less modular one, with the density trajectories crossing between the fourth and fifth decades.

Perhaps the most consequential finding is temporal. The decline in density and transitivity was already underway between the 20–29 and 30–39 strata, at ages preceding the menopausal transition, which in Mexican women begins several years before the median menopause age of roughly 48 years. The authors therefore conclude that the reorganization cannot be attributed to menopause alone, even though menopause is known to affect lipid regulation and vascular function. The 40–49 stratum, which overlaps the menopausal transition, recorded the most extreme value of all four metrics, and the reorganization was essentially complete by midlife.

Crucially, the discontinuity appeared only in the network view. Univariate biomarker trajectories—systolic pressure rising from 104 to 132 mmHg in women, fasting glucose from 87 to 98 mg/dL, eGFR falling from 143 to 99 mL/min/1.73 m²—changed gradually across the same ages, without any comparable break between 30 and 49 years. The topological series, in other words, located a discontinuity that the marginal series did not show. This is the central methodological claim of the paper: additive allostatic load indices, which treat each biomarker as an independent unit crossing a risk threshold, discard the covariation among markers, and it is precisely that covariation which reorganizes so dramatically in women at midlife.

The team ran several robustness checks. Because ten of the 22 variables are calculated from the others—a problem known as mathematical coupling that can inflate global metrics—they rebuilt the networks from the 12 directly measured variables alone and removed the largest clique of each network; edges among directly measured variables persisted in both analyses. They also stratified by dyslipidemia status, defined by elevated total cholesterol, elevated triglycerides or low HDL-C. Because dyslipidemia is a deterministic function of three nodes of the network itself, this analysis was explicitly exploratory, and the authors report that stratification did not fully account for the age trends, with dyslipidemic and non-dyslipidemic subgroups showing patterns compatible with selection and sampling effects rather than disease effects.

The authors are careful about interpretation. The networks describe how variables covary across individuals within each age–sex stratum, not how they are regulated within any single person; reading the between-person structure as within-person physiology requires longitudinal confirmation. The cross-sectional design also confounds age, cohort and survival effects, and the authors flag survivorship selection as a competing explanation for male stability: because men experience higher premature mortality from cardiovascular disease, diabetes and renal disease from midlife onward, each successive male stratum may represent an increasingly selected fraction of its cohort. Alternatively, the male transition may simply occur before age 20, a hypothesis the authors say could be tested directly in adolescent cohorts. Survey weights were not applied, so the estimates describe the analytic sample rather than the Mexican population, and edges were retained without correction for multiple testing. Even with those caveats, the finding stands: in women, the physiological network loses edges, loses triangles, lengthens its paths and compartmentalizes in a reorganization concentrated between ages 30 and 49, completed by midlife, and invisible to any analysis that looks at biomarkers one at a time.

Subject of Research: Age- and sex-related changes in the topology of physiological correlation networks built from allostatic load biomarkers in Mexican adults.

Article Title: Topological transition at midlife in the physiological network of Mexican women and divergent amplitude of topological change between sexes: allostatic load biomarkers across adult life

Article References: Uriarte-Ortiz, J. B., Barajas-Martínez, A., López-Mayorga, R. M., & Sampieri-Cabrera, R. (2026). Topological transition at midlife in the physiological network of Mexican women and divergent amplitude of topological change between sexes: allostatic load biomarkers across adult life. Frontiers in Physiology, 17, Article 1982421. https://doi.org/10.3389/fphys.2026.1982421

Image Credits: AI Generated

DOI: 10.3389/fphys.2026.1982421

Keywords: physiological networks, allostatic load, network topology, aging, midlife transition, sexual dimorphism, cardiometabolic biomarkers, ENSANUT, Mexico, menopause, correlation networks, network modularity