UCSF

High blood pressure is a leading cause of disability-adjusted life years (DALYs) worldwide and a major risk factor for adverse health outcomes, including cardiovascular disease, kidney disease, heart failure, and all-cause mortality. Blood pressure (BP) is highly heritable, with narrow-sense heritability estimates of at least 0.2 for systolic and diastolic blood pressure and over 1,000 genome-wide association study (GWAS) loci identified. Polygenic risk scores (PRS) summarize an individual's genetic predisposition to a trait or disease based on the combined effects of genetic variants. While PRS have potential applications in disease risk prediction and population screening, their performance varies based on the context of evaluation. Factors such as age, sex, medication use, ancestry, and their interactions influence the predictive power of BP PRS, which often perform worse in underrepresented populations. This dissertation utilizes longitudinal, real-world clinical data to examine the prediction accuracy of BP PRS across sex, age, and ancestry groups, assess their associations with target organ damage outcomes such as chronic kidney disease, and evaluate BP response to first line monotherapy antihypertensive medications by ancestry and drug class. In the first chapter of my dissertation, I investigate the associations between multi-ancestry polygenic risk scores (PRSs) for blood pressure (BP) traits in a real-world EHR-based clinical cohort by sex, antihypertension medication use, and ancestry. I then looked at the effect of age on PRS association with BP traits and how variance explained by PRS changes with age across ancestry groups. Summary statistics from the meta-analysis of UKB, ICBP, and BioVU, COGENT and Biobank Japan were used to train multi-ancestry PRSs for diastolic blood pressure (DBP) and systolic blood pressure (SBP), using PRS-CSx. I then tested these PRSs in the Kaiser Permanente (KP) Research Program on Genes, Environment, and Health (RPGEH) Genetic Epidemiology Research on Adult Health and Aging (GERA) cohort, in which individuals had multiple office visit BP measures across their life-course. The primary outcomes were DBP and SBP measurements. I tested the exposures (DBP and SBP PRS) on the outcomes using linear mixed models, stratified by sex (female/ male), ancestry group (African American, East Asian, European, Latino(a)) controlling for relevant covariates. I further stratified analyses by evenly sized age-bins. In a sensitivity analysis using individuals with both treated and untreated measures, I further stratified analyses by antihypertension (AHT) medication use (whether the BP measure was treated or not). I found significant differences in BP PRS prediction accuracy, by ancestry, sex, age across ancestries and antihypertension medication use. The estimated associations and prediction accuracy measured by percent variance explained (PVE) by the PRSs for BP traits varied across strata of ancestry, sex. Generally, females had higher PVE then males. The highest PRS PVE was in Latino(a) ancestry (female) with PVE of 2.27% in SBP and 2.14% DBP, followed by European, 1.92 % SBP and 1.7% DBP, then East Asian at 1.13% SBP and 1.7% DBP, and African American ancestries at 0.42 % SBP, and 0.76% DBP. PVE peaked at midlife and tailed off, except in Afr SBP-PRS where the accuracy dropped off drastically from early age. The PVE was higher in BP outcome measures that were not treated compared to treated across all the ancestry groups. In a real-world clinical multi ancestry cohort of ~100 000 US adults (GERA) within the same healthcare system, I show significant differences in multi-ancestry BP PRS prediction accuracy (measured by PVE) by ancestry, sex, age and antihypertension treatment status within Afr, Eas, Eur, Lat ancestry groups. This highlights the challenges of using multi-ancestry BP-PRS in real-world clinical settings, even though the PRS is constructed using multiple ancestry GWAS with large samples sizes. In the second chapter of my dissertation, I investigated the association between multi-ancestry PRSs BP and the incidence of target organ damage (TOD) clinical phenotypes at any age: chronic kidney disease (CKD), and cardiovascular disease (CVD) outcomes: Ischemic stroke (IS), heart failure (HF), atrial fibrillation (Afib), cardiomyopathy (CMP), coronary artery disease (CAD) and myocardial infarction (MI) in a real-world clinical cohort across African (Afr), East Asian (Eas), European (Eur), and Latino(a) (Lat) ancestry groups. I constructed multi-ancestry PRSs for SBP and DBP using summary statistics from genome-wide association studies (GWAS) of the UKB, ICBP, BioVU meta-analysis, the COGENT collaboration, and Biobank Japan. I then used these PRSs to test their association with TOD incidence in the Kaiser Permanente (KP) Research Program on Genes, Environment, and Health (RPGEH) Genetic Epidemiology Research on Adult Health and Aging (GERA) cohort. The primary outcome was incidence of each TOD, with age as the time scale in the analysis. Individuals entered the risk set at their age of entry into the cohort and were followed until the age at which they experienced the TOD or were censored, excluding those with prevalent disease at their age of entry. I used Cox proportional hazard models, stratified by ancestry group (Afr, Eas, Eur, Lat) and adjusted for sex, genetic principal components, diabetes, and respective PRS of the target organ damage phenotype obtained from the polygenic score catalogue (PGS catalogue, which is an open database of published polygenic scores that provides consistently annotated metadata for each score). An increase of 1 SD of the SBP PRS resulted in a 22%, 21%, 13%, 11% higher risk in incidence of CKD at any age in Lat, Afr, Eur, Eas ancestries respectively. This was statistically significant associated. However, the BP PRS association with the incidence of the CVD related TOD phenotypes was weaker, and not consistently statistically significant across ancestries. The effect of SBP PRS was found to be largely independent of PRS of the target organ damage phenotypes and diabetes status at start of follow up for CKD as the association only slightly deteriorated when added to the models. However, for ischemic stroke, the HRadj became statistically insignificant after addition of stroke TOD PRS. The correlation coefficient between SBP PRS and stroke TOD PRS was observed to be high across all ancestries. BP-PRS is associated with increased risk in incident CKD TOD at any age across ancestries, independent of TOD PRS and diabetes status at start of follow up. This association was less apparent in CVD related TOD outcomes. Finally, in the final chapter of my dissertation, I investigated ancestry differences in individual level response to monotherapy blood pressure medication treatment in a real-world clinical cohort with longitudinal EHR data by drug class type and dose in individuals of African, East Asian, European and Latino(a) ancestry with hypertension. I used prescription data from the Kaiser Permanente Research Program on Genes, Environment, and Health (RPGEH) Genetic Epidemiology Research on Adult Health and Aging (GERA) cohort, linking it with BP measurements where individuals had multiple internal medicine office visit BP measures across their life-course. We investigated drug classes—Angiotensin-Converting enzyme (ACE) inhibitors, hydrochlorothiazide (HCTZ), and dihydropyridine calcium channel blockers (CCBs)—and calculated the average change in systolic (SBP) and diastolic blood pressure (DBP) due to treatment by monotherapy AHT. I used ANOVA and Tukey’s Post hoc tests to test the differences in average change by ancestry group. I found that estimated average BP reductions varied depending on ancestry, drug class. Generally, estimated SBP reductions were more modest compared to those seen in clinical trials. The largest average reduction in SBP was seen with HCTZ in individuals of African ancestry, while the smallest was with ACE inhibitors in the same group. African ancestry individuals showed larger SBP reductions with HCTZ, but statistically significant smaller reductions with ACE inhibitors compared to other ancestry groups. The findings suggest that estimated BP reduction from antihypertensive medications vary by ancestry, drug class, in real-world clinical cohort settings. This challenges the common assumption in GWAS studies that BP medication effects are uniform (e.g., a 15 mm Hg reduction in SBP) across all ancestry populations and drug types, highlighting the need for more tailored approaches in clinical and genetic research e.g. in genome-wide association studies (GWAS) studies correcting for BP medication use.