A sharp difference between vulnerability paths
People whose health vulnerability increased between two UK Biobank assessments had a markedly higher observed rate of a first cardiovascular disease event than those whose vulnerability stayed low, a longitudinal cohort study found. In the fully adjusted analysis, the hazard ratio was 1.79, with a 95% confidence interval of 1.43 to 2.22. A hazard ratio compares the rate of events over time, so this does not mean that 79% of participants developed cardiovascular disease.
People with persistently elevated vulnerability also had higher observed risk than the persistently low group, with a hazard ratio of 1.28 and a 95% confidence interval of 1.19 to 1.38. The decreasing-vulnerability group had a hazard ratio of 1.22, with an interval of 0.82 to 1.82, a result that was not statistically significant at P = 0.320.
How vulnerability was tracked
Researchers screened 502,369 UK adults and included 16,072 participants who had two health-vulnerability assessments and no cardiovascular disease at baseline, after exclusions. The analysis used two UK Biobank assessment waves, and the paper reports ethics approval and written informed consent.
Health vulnerability was represented by an index that summed 15 indicators. Higher scores meant greater vulnerability and worse health. For a cumulative measure, the researchers averaged each participant's two scores and multiplied that average by the interval in years, calling the result cumulative health vulnerability, or CHV.
The primary outcome was first incident cardiovascular disease, defined as a composite of ischemic heart disease, atrial fibrillation, hypertensive disease and other cardiovascular diseases. Follow-up began at baseline and ended at diagnosis, death, loss to follow-up or 2 November 2022. Over a median 8.9 years, 3,033 first events were recorded, equal to 18.9% of the analysis group.
To identify patterns, the researchers fitted models with one through five classes and selected a three-class solution at both assessments. A latent transition analysis then produced nine possible transition patterns, which were grouped into four longitudinal vulnerability patterns. The risk comparisons used Kaplan-Meier curves and three Cox proportional-hazards models: unadjusted, adjusted for age and sex, and fully adjusted for education, smoking, drinking, body mass index, race, family history, diabetes and blood lipids.
The cumulative measure told a similar story
CHV showed a similar pattern. Participants in the highest CHV quartile, with scores of at least 44.2, had a hazard ratio of 1.47 for incident CVD compared with those in the lowest quartile, whose scores were below 19.0. The 95% confidence interval was 1.32 to 1.63, and the trend across higher CHV levels was statistically significant at P < 0.001.
A flexible curve analysis found a positive linear association between CHV and CVD risk, with no significant evidence that the relationship was nonlinear. The reported P value was less than 0.001 for linearity and 0.363 for nonlinearity.
The pattern differed across some groups
The relationship was not identical across age and smoking groups. The interaction with age was statistically significant at P = 0.036: among participants younger than 60, the increasing-vulnerability group had a hazard ratio of 2.12, with a 95% confidence interval of 1.61 to 2.78, while among those aged 60 or older, the persistently elevated group had a hazard ratio of 1.14, with an interval of 1.02 to 1.28. Smoking also interacted with the pattern at P = 0.002; among current smokers, the increasing-vulnerability group had a hazard ratio of 3.04, with an interval of 1.58 to 5.87.
After adjustment, however, no significant associations were reported for the individual cardiovascular disease subtypes: atrial fibrillation, ischemic heart disease, hypertensive heart disease or other cardiovascular diseases.
An association, not a cause
The researchers ran additional checks to see whether the pattern depended on how cardiovascular disease or follow-up was defined. After hypertensive diseases were excluded, hazard ratios were 2.14 for increasing vulnerability and 1.18 for persistent elevated vulnerability. In a 2014 landmark analysis that excluded 3,190 early cases, persistent elevated vulnerability remained associated with risk, with a hazard ratio of 1.18. A regrouping of the transitions produced hazard ratios of 1.62 for dynamic vulnerability and 1.28 for persistent elevated vulnerability.
A competing-risk analysis, which accounts for other events that can affect the chance of observing the outcome, produced hazard ratios of 1.77 for the increasing group and 1.30 for the persistently elevated group, with confidence intervals of 1.42 to 2.22 and 1.21 to 1.41. The authors also reported that stratified checks found no violation of the proportional-hazards assumption.
The findings are associations from a two-wave cohort, not evidence that vulnerability changes caused cardiovascular disease. The analysis cannot show that monitoring or reducing vulnerability would prevent it. Because the nine modeled transitions were grouped into four patterns, the categories provide a broad summary of change rather than a detailed account of every health trajectory.
Paper data and sources
Original title: Longitudinal associations of dynamic transitions of multidimensional health vulnerability with cardiovascular disease risk: evidence from UK Biobank
Authors: Wu J, Chen L, Sun S et al.
Journal/Repository: Frontiers in public health
Status: Peer-reviewed
First online: 2026-10-07
DOI: Not available
Original paper · Full text