Researchers analyzing data from 27,510 hypertensive participants in the UK Biobank Pharma Proteomics Project have developed separate protein-based risk scores for men and women that substantially improve the prediction of major adverse cardiovascular events, or MACE, over a ten-year horizon. The work, published in Biology of Sex Differences, harnessed large-scale plasma proteomic profiling to capture biological information that conventional clinical risk models simply do not see.

The team, led by researchers at Jilin University in China, applied a sex-stratified machine learning strategy known as Least Absolute Shrinkage and Selection Operator, or LASSO, regression to sift through thousands of circulating proteins. This technique penalizes complexity and selects only the most informative features, producing lean statistical models built from the proteins most strongly associated with future cardiovascular events. The result was a panel of 28 priority proteins for men and 23 for women, with only five proteins shared between the sexes: MMP12, NT-proBNP, NEFL, BCAN, and ADAMTS13.

That limited overlap is one of the most intriguing findings of the study. MMP12 is a matrix metalloproteinase involved in tissue remodeling and inflammation, while NT-proBNP is a well-established marker of cardiac strain. NEFL, a neurofilament light chain protein, signals neuronal injury, and ADAMTS13 is an enzyme that regulates blood clotting by cleaving von Willebrand factor. BCAN, brevican, is less familiar in cardiovascular circles, being primarily associated with the extracellular matrix of the brain. The fact that these five proteins anchor both models while the remaining two dozen or so differ suggests that the molecular pathways driving cardiovascular risk in hypertensive men and women are only partially shared.

The predictive power of the scores was considerable. In the validation set, participants in the highest protein risk group faced dramatically elevated odds of a major cardiovascular event compared with those in the lowest group. For men, the adjusted hazard ratio was 2.68, with a 95 percent confidence interval of 1.97 to 3.64. For women, it was even higher at 3.12, with a confidence interval of 2.20 to 4.43. In practical terms, a hypertensive person whose blood protein profile placed them in the top risk tier was roughly three times as likely to suffer a heart attack, stroke, or cardiovascular death within a decade as someone in the bottom tier.

Crucially, the protein scores added value on top of SCORE2, the standard European clinical tool for estimating ten-year cardiovascular risk. When the protein scores were integrated into the SCORE2 framework, the C-statistic, a measure of how well a model discriminates between people who do and do not experience events, rose from 0.627 to 0.667 for men and from 0.657 to 0.695 for women. Category-based net reclassification improvement, which quantifies how many people are correctly shifted into more accurate risk categories, reached 22.4 percent for men and 14.0 percent for women. These are meaningful gains in a field where incremental improvements in prediction can translate into lives saved through earlier intervention.

Recognizing that a 28-protein panel may be impractical for routine clinical testing, the researchers also constructed simplified scores using only the ten most heavily weighted proteins for each sex. In the simplified models, just two proteins overlapped between men and women, underscoring how divergent the sex-specific signatures are. Remarkably, these pared-down scores maintained predictive performance comparable to the full models, suggesting that a compact blood test measuring a handful of proteins could one day flag high-risk hypertensive patients who appear deceptively healthy by conventional measures.

Pathway enrichment analyses of the male and female protein panels revealed that the biological processes represented in the two models partially overlap, while others receive different relative emphasis. The authors are careful to note that these findings support sex-stratified prediction strategies rather than proving fundamentally distinct disease mechanisms between the sexes. The differences in protein selection may reflect genuine biological divergence, but they may also stem from differences in how proteins correlate with risk in each population. Either way, the study adds to mounting evidence that cardiovascular medicine, which has historically been built on data from men, should routinely stratify its models by sex.

The clinical implications are tantalizing. Hypertension is among the most prevalent and serious cardiovascular risk factors worldwide, and current risk prediction tools rely almost exclusively on conventional clinical variables such as age, cholesterol, smoking status, and systolic blood pressure. These tools cannot see the inflammatory, thrombotic, and cardiac stress processes that proteins like MMP12, ADAMTS13, and NT-proBNP reveal. A proteomic layer of risk assessment could identify which hypertensive patients need more aggressive therapy, closer monitoring, or novel interventions long before symptoms appear.

Important caveats remain. The scores were developed and validated within the UK Biobank, a cohort that is not fully representative of global populations, and the researchers emphasize that external validation in independent and more diverse cohorts is required before the scores can be considered clinically transportable. The study also relied on internal validation, and the modest though significant gains in discrimination must be weighed against the cost and complexity of proteomic testing. Still, as large-scale protein measurement becomes faster and cheaper, the prospect of a routine blood panel that reads the molecular weather forecast for the cardiovascular system is moving closer to reality. For the millions of people with hypertension whose risk is invisible to today’s models, that forecast could not come soon enough.

Subject of Research: Sex-stratified proteomic risk scores for predicting major adverse cardiovascular events in hypertensive patients

Article Title: Sex-specific proteomic risk scores reveal distinct cardiovascular risk profiles and improve prediction of major adverse cardiovascular events in hypertension

Article References: Ding, J., Li, J., Xu, W., Zhang, W., Gao, Y., & Cheng, L. (2026). Sex-specific proteomic risk scores reveal distinct cardiovascular risk profiles and improve prediction of major adverse cardiovascular events in hypertension. Biology of Sex Differences. https://doi.org/10.1186/s13293-026-00984-9

Image Credits: AI Generated

DOI: 10.1186/s13293-026-00984-9

Keywords: hypertension, proteomics, cardiovascular risk, UK Biobank, SCORE2, biomarkers, sex differences, LASSO regression, MACE, risk prediction, NT-proBNP, precision medicine

The study, led by Chiara de Waure of the University of Perugia and colleagues spanning institutions in Perugia, Rome, Palermo and Catania, drew its participants from the INNOPREV randomized controlled trial, a project funded by the Italian Ministry of Health under the National Recovery and Resilience Plan. The trial, registered at ClinicalTrials.gov as NCT05883878, is designed to test innovative, personalized strategies for cardiovascular disease prevention in high-risk adults. Before any randomization took place, the researchers asked participants aged 40 to 69 a deceptively simple question: how much do you believe cardiovascular disease threatens you personally? The answer came through a validated instrument called the Health Beliefs Related to Cardiovascular Disease Scale, or HBCVD, a 25-item questionnaire built on the Health Belief Model, one of the most influential psychological frameworks in preventive medicine.

The Health Belief Model holds that whether people adopt protective behaviors depends on four interconnected perceptions: their susceptibility to a condition, the severity it would have if it struck, the benefits of taking preventive action, and the barriers standing in the way of that action. The HBCVD scale translates these constructs into four subscales, each scored out of 100. When the researchers analyzed the 988 returned questionnaires from the 1,019 enrolled participants, they found a mean overall score of 61.10, with a standard deviation of 6.61. In other words, on average these adults, all of whom had been clinically assessed as being at moderate-to-high cardiovascular risk, perceived their risk as barely more than moderate.

To understand what shaped these perceptions, the team compared HBCVD scores across a battery of socio-demographic and clinical characteristics, including sex, age, marital status, education, employment, smoking status, body mass index, family history of cardiovascular conditions, the SCORE2 risk algorithm, and Life’s Essential 8, the American Heart Association’s composite measure of cardiovascular health covering diet, physical activity, nicotine exposure, sleep, body weight, blood lipids, blood glucose and blood pressure. Because the analysis involved many simultaneous comparisons, the researchers applied the Benjamini-Hochberg procedure to control the false discovery rate, a statistical safeguard that reduces the risk of spurious findings when testing many hypotheses at once. Effect sizes were reported as rank-biserial correlations or epsilon-squared values, allowing readers to judge not just statistical significance but practical magnitude.

After this correction, two factors stood out as consistently associated with overall risk perception: body mass index category and Life’s Essential 8 score. Participants with higher BMI perceived greater risk, an association confirmed in multivariable linear regression models fitted with heteroskedasticity-consistent standard errors. Compared with participants of normal weight, those who were overweight scored 1.13 points higher on the overall HBCVD scale, while those with obesity scored 2.94 points higher, the latter difference highly statistically significant. The direction makes intuitive sense: people carrying excess weight may receive more frequent warnings from clinicians, or may themselves sense the physical burden, and this awareness translates into stronger perceived vulnerability.

The association with cardiovascular health ran in the opposite direction, and this is where the findings become genuinely thought-provoking. Participants with moderate cardiovascular health scored 2.63 points lower on risk perception than those with low cardiovascular health, and those with high cardiovascular health scored 3.80 points lower. On the surface this seems logical, since healthier people genuinely face lower risk. But the effect sizes were small, and the pattern raises a subtle concern for prevention science: people who already behave well may feel insulated from cardiovascular disease, and that sense of safety could erode their motivation to maintain healthy habits over the long term. Prevention programs that ignore this psychological dynamic risk losing participants precisely when those habits have begun to pay off.

When the researchers dissected the four subscales separately, each told a different story. Perceived susceptibility showed the strongest and most consistent relationships with objective risk factors, tracking closely with BMI, family history, and Life’s Essential 8 score. Perceived benefits, by contrast, varied mainly with socio-demographic characteristics, suggesting that beliefs about the value of prevention are shaped more by social context than by clinical status. Perceived barriers were linked both to education level and to cardiovascular health, hinting that people with less education may face more practical obstacles to preventive action, and that even health-conscious individuals encounter friction in sustaining it. Perceived severity differed significantly only by sex and age class, indicating that judgments about how bad a cardiovascular event would be are relatively uniform across the population, anchored more in personal demographics than in medical risk profiles.

The methodological choices in the study deserve attention. Because all variables were measured at a single pre-randomization time point, the authors are careful to note that no causal inference is possible; the analysis shows associations, not mechanisms. The cross-sectional design means that, for example, obesity might drive higher risk perception, or conversely that anxious people might gain weight, or that both could reflect some third factor. Yet even as a snapshot, the data are valuable precisely because they come from a trial population that was thoroughly phenotyped, with clinically validated risk scores and standardized health assessments, rather than from a self-selected survey sample. The modest effect sizes, with epsilon-squared values around 0.02, underscore that risk perception is a diffuse phenomenon shaped by many small influences rather than one dominant driver.

Why does all this matter? Because risk perception is the psychological gateway to prevention. Clinical guidelines can calculate a person’s ten-year cardiovascular risk to several decimal places, but that calculation changes nothing unless the person believes it applies to them. The concept of unrealistic optimism, well documented in health psychology, describes people’s tendency to believe their own risk is lower than that of their peers, and the moderate scores observed here, in a population selected specifically for elevated risk, are consistent with that bias. If the very people flagged by SCORE2 as needing intervention do not feel vulnerable, adherence to medications, lifestyle counseling and follow-up appointments may all suffer. The INNOPREV findings give clinicians and public health planners a map of where those perception gaps concentrate.

The authors argue that their results can support the development of tailored preventive strategies to enhance risk awareness and promote healthy behaviors, and the trial’s randomized phase will test exactly that idea. In an era when personalized prevention is moving from slogan to policy, this study is a reminder that personalization must begin with the mind as well as the body. A prevention program calibrated to a person’s cholesterol but not to their beliefs is working with half the information. As Italy and other countries grapple with the enormous burden of cardiovascular disease, the quiet insight from this analysis may prove to be among the most actionable: before we can persuade people to protect their hearts, we first have to convince them their hearts are worth protecting, and that task starts with understanding how they currently see the risk.

Subject of Research: Cardiovascular risk perception among Italian adults at moderate-to-high cardiovascular risk enrolled in the INNOPREV prevention trial

Article Title: Perceived risk of cardiovascular diseases in the Italian population: a baseline cross-sectional analysis of participants enrolled in the INNOPREV trial

Article References: de Waure, C., Valentini, I., Arcelli, M., Volpi, F., Scarsi, N., Gobbetti, C., Lolli, O., Russo, L., Proto, L., Galarducci, R., Pasciuto, T., Severino, A., Tuttolomondo, A., De Bella, D., Daidone, M., Morello, G., Muli, R. L., Previti, A., Pieri, A., … Boccia, S. (2026). Perceived risk of cardiovascular diseases in the Italian population: a baseline cross-sectional analysis of participants enrolled in the INNOPREV trial. BMC Public Health. https://doi.org/10.1186/s12889-026-29286-w

Image Credits: AI Generated

DOI: 10.1186/s12889-026-29286-w

Keywords: cardiovascular disease, risk perception, INNOPREV trial, Health Belief Model, HBCVD scale, Life's Essential 8, body mass index, SCORE2, preventive behaviours, Italy, public health, BMC Public Health