Rising Biological Wear-and-Tear Score Tracks Social Disadvantage and Predicts Early Death

For decades, scientists have known that poverty, low education and economic insecurity leave fingerprints on the human body. What has remained stubbornly unclear is precisely how social disadvantage becomes written into our physiology, and whether that biological inscription can be measured as it unfolds over time. A new study drawing on the UK Biobank, one of the world’s largest prospective cohort resources, now offers a rare longitudinal window onto this process, tracking changes in a composite measure of multisystem biological health and asking whether those changes help explain why disadvantaged people die earlier and fall ill more often.

The research, published in eClinicalMedicine, builds on the concept of allostatic load, the idea that repeated or chronic stressors progressively over-solicit the body’s regulatory systems, producing a kind of physiological wear-and-tear. Previous work by the same team developed a Biological Health Score, or BHS, that extends traditional allostatic load measures by incorporating physiological systems not directly tied to the stress response, such as liver and kidney function. In earlier analyses, the score showed strong social gradients and predicted mortality and disease incidence. But nearly all such studies relied on a single snapshot in time, leaving open whether the trajectory of biological dysregulation, rather than its level at any one moment, carries independent prognostic information.

To address that gap, the researchers turned to the UK Biobank’s repeat assessment visit. Of the 502,536 volunteers enrolled between 2006 and 2010, some 20,344 returned for a second assessment between December 2012 and June 2013, donating blood that was re-analysed for the same panel of biomarkers. After exclusions, 20,048 participants, 10,266 women and 9,782 men with a mean age of 57.1 years at the first assessment, formed the analytical sample. The team selected 15 biomarkers spanning five physiological systems: metabolic markers including HbA1c, HDL and LDL cholesterol and triglycerides; cardiovascular measures of blood pressure and pulse rate; inflammatory markers including C-reactive protein, insulin-like growth factor 1 and glycoprotein acetyls; liver enzymes ALT, AST and GGT; and kidney markers creatinine and cystatin C.

Each biomarker was log-transformed and standardised, and for protective markers such as HDL and IGF-1 the sign was reversed so that any increase in the score corresponded to greater biological risk. System-specific sub-scores were computed as the mean of their constituent standardised biomarkers, and the overall BHS was the mean of the five sub-scores, ensuring that each physiological system contributed equally regardless of how many biomarkers it contained. Crucially, the same methodology was applied to the standardised differences between the second and first measurements, yielding a change score that captured the direction and magnitude of biological drift over the roughly 4.3 years between visits.

The results confirmed that the score behaves as a cumulative measure should. The total BHS increased between the two time points, more strongly in women than in men, and every sub-score rose except the cardiovascular one. The largest biomarker increases were seen for cystatin C, creatinine and HbA1c, while HDL cholesterol, diastolic blood pressure and pulse rate declined. Notably, change was consistently smaller among participants with higher baseline scores, a statistical ceiling effect that the authors say underscores the value of modelling trajectories rather than static levels. The flat cardiovascular sub-score, they suggest, may partly reflect the widespread use of antihypertensive medication, which pharmacologically controls blood pressure and thereby blunts the sensitivity of those markers to underlying deterioration.

The social patterning of these changes was striking. Using a stability selection procedure based on LASSO regression across 100 random subsamples, the team identified confounders and mediators from a pool of 270 candidate variables, then modelled BHS change against six social exposures: age at completion of education, qualification level, household income, financial difficulties, ability to confide in someone, and loneliness. Educational and economic disadvantage emerged as the dominant predictors. Each additional year of education was associated with a smaller increase in the score in both sexes, while low qualifications and reported financial difficulties were linked to larger increases in both women and men. Low income predicted greater biological deterioration in men but not in women. By contrast, socio-affective factors, loneliness and rarely having someone to confide in, showed little association with BHS change once confounders were accounted for.

Did these biological trajectories matter for health? Follow-up through national registry linkage extended to December 2022, giving nine to ten years of outcome data after the second assessment. In age-scaled Cox models adjusted for baseline BHS and the interval between measurements, a one-unit increase in the BHS change score was associated with substantially higher risks. For all-cause mortality, the hazard ratio was 1.46 in men and 1.26 in women; for atherosclerotic cardiovascular disease, 1.33 in men and 1.24 in women; and for cancer incidence, 1.29 in women and 1.16 in men. Because these associations were independent of the baseline score, they suggest that two people with identical biological profiles at one point in time may nonetheless diverge in risk depending on the direction of their subsequent physiological drift.

The associations proved robust to adjustment. When the models were sequentially adjusted for history, social and environmental factors, behaviours, diet, anthropometrics and baseline health, the effects attenuated somewhat, most visibly in women, where fully adjusted estimates were no longer statistically significant. In men, however, the associations with mortality and cardiovascular disease persisted even after full adjustment, with hazard ratios of 1.38 and 1.25 respectively. Sub-score analyses hinted at system-specific pathways: mortality links appeared driven by cardiovascular and inflammatory markers in women and by inflammatory, liver and renal markers in men, while cancer associations centred on inflammatory and renal systems.

The most conceptually ambitious part of the study was its mediation analysis. Using the parametric g-formula implemented through Monte Carlo simulation, a counterfactual framework that can handle exposure-induced mediator-outcome confounding, the researchers decomposed the total effect of each social exposure on each outcome into direct effects and effects routed through BHS change. The total effects themselves were sobering: low income was associated with a 51 percent higher mortality hazard in men, financial difficulties with a 45 percent higher hazard, and each year less of education with a 4 to 5 percent higher hazard of mortality or cardiovascular disease. Yet the proportion of these effects explained by short-term BHS change was modest. Statistically significant mediated proportions appeared only for economic disadvantage and mortality in men, roughly 4.5 percent for low income and 6.9 percent for financial difficulties. No significant mediation was observed in women.

The authors interpret this cautiously. Short-term changes in multisystem dysregulation, they conclude, capture only part of the longer-term biological embodiment of social experience, and multiple other pathways, from health behaviours to psychosocial stress to access to care, clearly contribute. Limitations also temper the findings: the UK Biobank’s well-documented healthy-volunteer bias is amplified among participants who return for repeat assessments, the broad cancer definition may have diluted site-specific signals given long latency periods, and residual confounding and reverse causality cannot be excluded. Still, the central message is compelling. Repeated multisystem biological measurements appear to capture an evolving physiological vulnerability that is socially patterned and predictive of death and disease beyond what any single snapshot reveals. As the researchers argue, extending such analyses to multiple measurements across the life course could open a genuine window onto the dynamics of how social disadvantage becomes biology, and potentially identify the points at which intervention might still rewrite that trajectory.

Subject of Research: Longitudinal changes in a multisystem Biological Health Score as a mediator between social disadvantage and mortality and disease in the UK Biobank cohort

Article Title: Contribution of short-term changes in a composite health score on pathways linking social disadvantage and health: evidence from the UK biobank prospective cohort study

Article References: Contribution of short-term changes in a composite health score on pathways linking social disadvantage and health: evidence from the UK biobank prospective cohort study. (n.d.). https://doi.org/10.1016/j.eclinm.2026.104221

Image Credits: AI Generated

DOI: 10.1016/j.eclinm.2026.104221

Keywords: UK Biobank, allostatic load, Biological Health Score, social disadvantage, health inequalities, biomarkers, mortality, cardiovascular disease, cancer incidence, mediation analysis, physiological dysregulation, prospective cohort study