The research drew on the Health and Retirement Study, a landmark longitudinal survey that has followed more than 37,000 American adults aged 50 and older since 1992. For this analysis, the team focused on 3,467 individuals who had reported stroke status at the twelfth survey wave, consented to blood draws in the 2016 Venous Blood Study, and completed cognitive assessments through wave 15. The cohort had a median age of 71 years, 59 percent were women, and 295 participants had experienced a stroke. Rather than treating stroke as a single moment in time, the researchers framed lifetime stroke burden as their exposure, then asked how it propagated forward into cognition measured in 2018 and again in 2020.

The methodological centerpiece of the study is the four-way decomposition, a modern causal inference framework rooted in non-parametric structural equation models. Instead of merely asking whether stroke correlates with worse cognition, the decomposition splits the total effect of stroke on cognition into distinct components: a controlled direct effect that bypasses intermediate biology, pure indirect effects that flow through a mediator, and two interaction terms that capture how stroke and the mediator amplify each other. In plain terms, the approach distinguishes between pathways where the mediator carries the stroke signal forward and pathways where the stroke only does damage because it meets a susceptible biological environment. Standard errors were computed from 1,000 bootstrapped samples, and a Bonferroni-corrected threshold of 0.002 guarded against spurious findings across dozens of comparisons.

The mediators themselves represent some of the most advanced tools in contemporary geroscience. Using DNA methylation data, the chemical marks deposited on DNA that shift predictably with age, the team quantified biological aging with eight epigenetic clocks, including the Horvath pan-tissue clock built from 353 CpG sites, the Hannum clock based on 71 sites, the Levine PhenoAge clock, and GrimAge, a composite measure trained in the Framingham Heart Study that estimates epigenetic surrogates of plasma proteins such as GDF-15, cystatin C, leptin and plasminogen activator inhibitor 1, along with cumulative smoking exposure. On top of the epigenetic layer, the researchers measured a panel of inflammatory cytokines, including interleukin-6, interleukin-1 receptor antagonist, interleukin-10, tumor necrosis factor receptor 1 and transforming growth factor-beta, alongside immune cell counts covering basophils, eosinophils, lymphocytes, monocytes, neutrophils and total white blood cells.

The first striking finding is that stroke survivors were biologically older than their calendar years. Their median GrimAge was 74.5 years compared with 70 years among participants who had never had a stroke, an acceleration consistent with the idea that cerebrovascular disease is not merely a disease of aged vessels but a manifestation of accelerated systemic aging. When the team then traced how stroke affected cognition at wave 14, the short-term outcome, they found that biological age carried a substantial indirect burden: the pure natural indirect effect through biological aging was −0.47 points on the total cognition score, with a P value below 0.001. A nearly identical signal appeared for intermediate-term cognition measured in 2020, with a pure natural indirect effect of −0.48. Mediation through biological aging accounted for roughly 28 percent of the stroke effect on short-term cognition and 23 percent on intermediate-term cognition.

Among the immune and inflammatory markers, one molecule towered over the rest. Transforming growth factor-beta, a cytokine with near-omnipresent signaling roles in cell growth, differentiation, apoptosis, wound healing, angiogenesis and tissue remodeling, displayed a powerful interactive effect rather than a purely mediative one. The reference interaction effect for TGF-beta on short-term cognition was −2.09, with a 95 percent confidence interval spanning −3.41 to −0.78 and P below 0.001, meaning that the harm stroke inflicted on cognition was strongly modified by the level of this cytokine when the mediator sat at its natural, unexposed value. The effect on intermediate-term cognition was attenuated at −1.72 and did not survive the strict Bonferroni correction, suggesting the interactive damage is most pronounced in the months shortly after the vascular insult.

Weaker but statistically detectable mediation emerged through the immune compartment as well. White blood cell count carried a small indirect effect on short-term cognition of −0.02 with P equal to 0.03, and TNF receptor 1 showed a comparable marginal signal. More intriguingly, when biological aging, inflammatory cytokines and immune markers were modeled together as a joint mediator set, the combined indirect effect reached −0.42 on short-term cognition, indicating that these systems act in concert rather than in isolation. The authors point to the concept of inflamm-aging, the chronic low-grade inflammation that accumulates with age, as a plausible link: TGF-beta is inversely related to epigenetic aging clocks, it halts the cell cycle at the G1 phase, and epigenetic regulators can influence its activation and downstream signaling, creating feedback loops between aging biology and inflammatory control.

Mechanistically, the findings make biological sense. Ischemic injury releases danger-associated molecular patterns that activate Toll-like receptors, recruit innate and adaptive immune cells and ignite the complement system. TGF-beta, in turn, recruits macrophages that generate additional cytokines such as interleukin-1, tumor necrosis factor-alpha and fibroblast growth factor, fueling a positive feedback loop at the injury site, while reactive oxygen species amplify TGF-beta activity through thrombospondin-1. Because the cytokine governs tissue and vascular remodeling, the authors argue that regulating TGF-beta after cerebrovascular injury could become a therapeutic strategy, timed to the window when its interaction with the stroke insult does the most cognitive harm. Epigenetic mechanisms may also connect to classical Alzheimer pathology, since DNA methylation variability has been linked to amyloid-beta and neurofibrillary tangles in brain tissue.

The study is not without limitations, which the authors confront directly. Residual confounding can never be fully eliminated in observational data, stroke was ascertained by self-report of a physician diagnosis, the follow-up window for cognition was relatively short, and the median age of 71 means the results cannot be extrapolated to younger populations. Severe disability may also have prevented some stroke survivors from participating in cognitive testing. Yet the evidence base is reassuring in key respects: self-reported stroke in the Health and Retirement Study correlates strongly with hospital records, and sensitivity analyses adjusting for race, education, baseline cognition and log-transformed biomarkers produced comparable or attenuated but consistent estimates.

The significance of the work lies in its reframing of what happens after a stroke. Rather than viewing post-stroke cognitive decline as a fixed consequence of brain injury, the unified causal framework portrays it as a dynamic interplay between an accelerated aging process that can be measured in a blood sample and an inflammatory milieu that can, in principle, be therapeutically modulated. Because both the exposure and the proposed mediators are potentially manipulable through non-invasive means, DNA methylation clocks combined with routine immunoinflammatory blood markers could one day serve as prognostic dashboards, identifying which stroke survivors are aging fastest and which immune checkpoints are most likely to erode their cognition. In a population where stroke remains a leading cause of acquired dementia, turning biological aging and TGF-beta from silent accomplices into monitored, modifiable targets could reshape how medicine protects the aging brain.

Subject of Research: The causal roles of DNA methylation-based biological aging, inflammation and immunity in post-stroke cognition

Article Title: The role of DNA methylation-based biological aging, inflammation and immunity in post-stroke cognition: A unified causal framework

Article References: Waziry, R., Miles, C. H., Williams, O., & Tiemeier, H. (2026). The role of DNA methylation-based biological aging, inflammation and immunity in post-stroke cognition: A unified causal framework. PLOS Aging and Health, 1(3), e0000033. https://doi.org/10.1371/journal.page.0000033

Image Credits: AI Generated

DOI: 10.1371/journal.page.0000033

Keywords: stroke, cognition, biological aging, DNA methylation, epigenetic clocks, GrimAge, TGF-beta, inflammation, immunity, causal mediation, Health and Retirement Study, cognitive decline

The study, published in Nature Human Behaviour, analyzed data from 3,171 adolescents participating in the Tokyo Teen Cohort, a population-based longitudinal study in which 46.9 percent of participants were girls. The cohort was assessed at three time points, ages twelve, fourteen, and sixteen, allowing researchers to track how experiences unfolding in early and mid-adolescence shaped mental health outcomes later in the teenage years. At age sixteen, the disparities were stark. Girls showed a prevalence of depressive symptoms of 15.9 percent compared with 8.2 percent among boys, nearly a doubling of risk. Suicidal ideation followed a similar pattern, affecting 29.8 percent of girls versus 18.1 percent of boys, meaning nearly one in three sixteen-year-old girls in the cohort reported thoughts of ending their own life.

What distinguishes this investigation from much of the prior literature on sex differences in adolescent mental health is its method of hypothesis generation. Rather than relying solely on theory or prior statistical associations, the research team coproduced the study with young people themselves, working alongside the Bridging Divides Young Persons Advisory Group. This collaborative process, in which adolescents helped shape the conceptualization and design of the research, suggested loneliness and desire for thinness as the most plausible candidate pathways linking female sex to internalizing problems. The approach reflects a growing movement in child and adolescent mental health research toward genuine participation of young people in setting research priorities, a shift that recent commentaries in leading medical journals have argued is long overdue.

To test whether these candidate pathways actually carried a causal contribution, the researchers employed the mediational g-formula, a statistical technique from the field of causal inference that goes well beyond traditional regression-based mediation analysis. The g-formula, developed for settings with time-varying exposures and multiple mediators, allows investigators to estimate what would happen under hypothetical interventions, for example, if loneliness and desire for thinness at age fourteen were eliminated or set to the levels observed among boys. This counterfactual framing is critical because it treats the mediators not merely as correlated variables but as potential intervention targets, and it accommodates the complex web of confounders and intermediate variables that accumulate across adolescence.

The results were striking. The effect transmitted through loneliness and desire for thinness at age fourteen, relative to the total effect of female sex, was estimated at 34.7 percent for depressive symptoms at age sixteen and 46.3 percent for suicidal ideation at age sixteen. In other words, roughly a third of the sex gap in depression and nearly half of the sex gap in suicidal thoughts in this cohort could be statistically attributed to these two intermediates. The researchers visualized their hypothesized causal structure in a directed acyclic graph, adjusting for confounders measured at age twelve, including family and individual characteristics that could independently shape both the mediators and the outcomes. Analyses were conducted in R, with the parametric g-formula implemented through custom code that the team has made publicly available.

The choice of mediators is theoretically meaningful. Loneliness has emerged in recent years as one of the most robust psychological predictors of adolescent psychopathology, with prior work by members of the same team linking chronic loneliness to subsequent psychotic experiences, depression, anxiety, and diminished well-being. Girls may be disproportionately exposed to social disconnection during the middle teenage years, a period when peer relationships intensify in importance and social comparison becomes pervasive. Desire for thinness, meanwhile, taps into body image concerns that are well documented to intensify among adolescent girls, shaped by cultural ideals, weight teasing, and media exposure. Previous research has associated weight perception and weight-related mistreatment with poorer mental health among adolescents, and the present findings suggest that this internalized pressure may be a genuine engine of the depression and suicidality gap rather than a mere correlate.

The study’s framing of sex as the exposure deserves careful attention. In causal inference methodology, there has been extensive debate about whether characteristics like sex, race, or other social categories can be treated as causes of health outcomes, since they are not manipulable interventions in the conventional sense. The authors navigated this methodological terrain deliberately, situating their analysis within a tradition that interprets such disparities as reflecting the consequences of social positioning rather than immutable biology. By estimating the proportion of the total effect that flows through modifiable intermediates, the analysis implicitly asks a policy-relevant question: how much of the gap could close if the social conditions that girls face were changed? The answer, roughly 35 percent for depressive symptoms and 46 percent for suicidal ideation, suggests that a large fraction of the disparity is potentially addressable.

Several caveats temper the interpretation. The data come from a single cohort in Tokyo, and although the Tokyo Teen Cohort is one of the most comprehensive adolescent studies in the world, the generalizability of the specific proportions to other cultural contexts remains to be established. Residual and unmeasured confounding is an ever-present threat in observational mediation analyses, and the authors acknowledge the broader methodological literature on how such bias can distort causal estimates. The mediators were measured at a single age, fourteen, whereas loneliness and body image concerns fluctuate across development; earlier work from the same cohort has shown that distinct loneliness trajectories carry different risks. Moreover, the analysis identifies statistical pathways consistent with mediation, but it cannot prove that intervening on loneliness or thinness desire would fully close the gap, since other unmeasured pathways may operate in parallel.

Even with these limitations, the implications are considerable. The findings suggest that interventions targeting social connection and body image during early adolescence, before the age-fourteen window examined here, could meaningfully reduce the excess burden of internalizing problems carried by girls. Schools, clinicians, and policymakers now have evidence-based targets that emerged not from adult assumptions but from the lived insights of young people themselves. As the global mental health community confronts what researchers have called a silent crisis in girls’ mental health, this study demonstrates that the mechanisms behind the crisis are identifiable, measurable, and, crucially, modifiable. The work was supported by the Japan Society for the Promotion of Science, the Japan Science and Technology Agency, the Wellcome Trust, and the ESRC Centre for Society and Mental Health at King’s College London, and its analytical code is openly shared to enable replication and extension by other research teams worldwide.

Subject of Research: Mediation of sex differences in adolescent depressive symptoms and suicidal ideation through loneliness and desire for thinness in a longitudinal cohort

Article Title: Loneliness and desire for thinness explain sex differences in adolescent internalizing problems

Article References: Loneliness and desire for thinness explain sex differences in adolescent internalizing problems. (n.d.). https://doi.org/10.1038/s41562-026-02578-2

Image Credits: AI Generated

DOI: 10.1038/s41562-026-02578-2

Keywords: adolescent mental health, loneliness, desire for thinness, depression, suicidal ideation, sex differences, Tokyo Teen Cohort, causal mediation, mediational g-formula, internalizing problems, body image, coproduction