High blood pressure (BP) is the leading risk factor for cardiovascular disease (CVD) mortality. Despite the high heritability of BP levels, clinical applications of BP genetics to prevent hypertension and its long-term adverse effects, such as CVD mortality, are lacking.

BP traits are highly polygenic, with 2103 independent signals identified in the latest large-scale genome-wide association study for systolic (SBP), diastolic (DBP), and pulse pressure [1]. Polygenic risk scores (PRS) are currently the most informative way of quantifying genetic burden for BP traits and hypertension. Indeed, BP PRSs are associated with BP differences starting from childhood and predict hypertension prevalence across ancestries [2, 3]. However, predictive power alone does not guarantee clinical value. Lambert et al. outline three main properties of PRSs relevant when considering integrating PRSs into clinical risk prediction models (Fig. 1) [4]. While these properties are actively studied in the context of BP PRSs predicting hypertension risk, hard endpoints such as all-cause and CVD mortality have received less attention.

Main properties of PRSs relevant for integration into clinical risk prediction models: (i) independence from traditional risk factors, (ii) statistical interactions with traditional risk factors, and (iii) added predictive power over traditional risk factors, as outlined by Lambert et al. [4]. PRS polygenic risk score. Created with BioRender.com

All-cause and cause-specific mortality are clinically meaningful gold standard endpoints in epidemiological research. In a trans-biobank study of 676,000 individuals, Sakaue et al. estimated associations between 45 biomarker PRSs and all-cause mortality. The authors found that SBP and DBP PRS were among the few PRSs associated with all-cause mortality across ancestries [5]. SBP PRS was also associated with CVD mortality, and the association between SBP PRS and all-cause mortality was driven by type 2 diabetes, cerebral infarction, and dyslipidemia. Meisner et al. took a similar approach, estimating associations between 12 major mortality risk factor PRSs and all-cause and cause-specific mortality in 337,000 UK Biobank participants. SBP and DBP PRSs were associated with all-cause and CVD mortality, but only in men [6]. In both studies, hazard ratios per 1 standard deviation increase in BP PRS for all-cause and CVD mortality were modest, ranging from 1.04 to 1.10.

Interactions between BP PRSs and clinical CVD risk factors could bring novel insights into CVD prevention. For all-cause mortality, Sakaue et al. did not find statistically significant interactions between SBP PRS and alcohol consumption, smoking, or regular exercise in 170,000 Biobank Japan participants [5]. A recent paper examined cross-sectional BP PRS interactions with environmental risk factors for coronary artery disease prevalence in a Korean sample [7]. The authors reported interactions between BP PRSs and obesity, abdominal obesity, triglycerides, high-density lipoprotein cholesterol, and smoking. However, because the study was cross-sectional, it could not assess the long-term implications of these interactions.

The present study by Fujii et al. tackles all properties i-iii outlined by Lambert et al. for all-cause and CVD mortality in a Japanese prospective cohort comprising 9296 participants [8]. For (i), the authors demonstrate remarkably strong associations between BP PRSs and CVD-related mortality after adjusting for lifestyle factors. For (ii), the authors employed two different approaches to examine interactions between BP PRSs and smoking, alcohol use, and sodium intake for CVD mortality: (1) estimating interaction terms in Cox proportional hazards models and (2) comparing attributable risks between lower (0–80%) and higher (>80%) SBP PRS groups. While PRS stratification yielded interesting results, neither approach demonstrated statistical significance. For (iii), the authors compared the predictive power of clinical risk models for CVD mortality with and without BP PRSs, again finding no statistically significant differences.

The ability of BP PRSs to independently predict lifetime BP trajectories extends to all-cause and CVD mortality. It remains to be seen whether BP PRSs have clinically meaningful interactions with traditional CVD risk factors on all-cause or CVD mortality. BP PRSs alone may have limited clinical utility in improving clinical risk prediction models for mortality endpoints. Nevertheless, BP PRSs will likely find their first clinical applications in individuals with few traditional CVD risk factors.