Racial and Ethnic Disparities in Antihypertensive Medication Prescribing Patterns and Effectiveness

Abstract

Variability in drug effectiveness and provider prescribing patterns have been reported in different racial and ethnic populations. We sought to evaluate antihypertensive drug effectiveness and prescribing patterns among self-identified Hispanic/Latino (Hispanic), Non-Hispanic Black (Black), and Non-Hispanic White (White) populations that enrolled in the NIH All of Us Research Program, a US longitudinal cohort. We employed a self-controlled case study method using electronic health record and survey data from 17,718 White, Hispanic, and Black participants who were diagnosed with essential hypertension and prescribed at least one of 19 commonly used antihypertensive medications. Effectiveness was determined by calculating the reduction in systolic blood pressure measurements after 28 or more days of drug exposure. Starting systolic blood pressure and effectiveness for each medication were compared for self-reported Black, Hispanic, and White participants using adjusted linear regressions. Black and Hispanic participants were started on antihypertensive medications at significantly higher SBP than White participants in 13 and 7 out of 19 medications, respectively. More Black participants were prescribed multiple antihypertensive medications (58.46%) than White (52.35%) or Hispanic (49.9%) participants. First-line HTN medications differed by race and ethnicity. Following the 2017 American College of Cardiology and the American Heart Association High Blood Pressure Guideline release, around 64% of Black participants were prescribed a recommended first-line antihypertensive drug compared with 76% of White and 82% of Hispanic participants. Effect sizes suggested that most antihypertensive drugs were less effective in Hispanic and Black, compared with White, participants, and statistical significance was reached in 6 out of 19 drugs. These results indicate that Black and Hispanic populations may benefit from earlier intervention and screening and highlight the potential benefits of personalizing first-line medications.

Hypertension affects nearly half the United States (US) population and contributes to the development of cardiovascular disease, renal disease, and stroke.^1,2 Research has shown that hypertension is harder to control in non-Hispanic Black (Black) than non-Hispanic White (White) patients, and Black individuals have been shown to have higher rates of hypertension than White individuals, while Hispanic/Latino (Hispanic) individuals have been shown to have lower rates of hypertension than White individuals.^1,3–6

Black and Hispanic populations have been underrepresented in clinical trials in the United States despite National Institutes of Health (NIH) policy changes to mandate the analysis and reporting of race and ethnicity differences in phase III clinical trials in 2001.^7–16 Previous research has shown differences in the effectiveness of drug classes among Black and White populations. For example, calcium channel blockers (CCB) and thiazide-type diuretics were more effective antihypertensive (HTN) medications in Black individuals, while angiotensin-converting enzyme inhibitors (ACEI) and beta-adrenergic blockers (beta blockers) were more effective in White individuals.^17,18 The data supporting the differences in effectiveness in the Black population has led to different recommendations for first-line HTN medications by race in the 2014 Management of High Blood Pressure Report from the Panel Members Appointed to the Eighth Joint National Committee (JNC8) and the 2017 ACC/AHA/AAPA/ABC/ACPM/AGS/APhA/ASH/ASPC/NMA/PCNA Guideline for High Blood Pressure.^3,19 For non-Black individuals without chronic kidney disease (CKD), initial HTN medications in the 2014 and 2017 recommendations included a thiazide-type diuretic, CCB, ACEI, or angiotensin receptor blocker (ARB). For Black individuals without CKD or heart failure (HF), this recommendation was limited only to a thiazide-type diuretic or CCB. For all individuals with CKD, an ACEI or ARB was recommended, and for Black individuals with HF, a beta blocker was recommended.

Electronic health record (EHR) data have proven useful to study medication effects using large, real-world datasets.¹⁹ One such publicly available resource is the NIH All of Us Research Program (All of Us).^20,21 All of Us is a longitudinal cohort of participants that contribute a broad range of survey data, physical measurements, biospecimens, and EHR data, including drug exposures. All of Us recruits individuals from diverse backgrounds and over 45% of participants identify as racial and ethnic populations underrepresented in biomedical research.²² Previous All of Us studies have evaluated disparities in disease diagnoses, disease outcomes, and medication prescribing patterns in the context of diabetes and depression.^21,23–25

The goal of this research was to assess disparities in HTN drug effectiveness and prescribing patterns among Black, Hispanic, and White populations using All of Us. We conducted generalizable self-controlled case studies for a variety of prescribing pattern and drug effectiveness disparity models.

MATERIALS AND METHODS

Participants

This study was conducted using participant data from the NIH All of Us Research Program, a longitudinal cohort of 239,740 participants with EHR data and drug exposures listed as of June 2023 (CDR version C2022Q4R11). Self-reported demographic information from the All of Us intake questionnaire was used to determine the race and ethnicity of participants. In this study, we analyzed participants identifying as “Black, African American, or African;” “Hispanic, Latino, or Spanish;” and “White” We omitted other racial and ethnic categories, including multiracial and multiethnic individuals, due to inadequate sample sizes. Participants granted consent for their deidentified EHRs and survey data to be accessed and used for broad-based research, and study procedures were approved by the All of Us Research program institutional review board.

Study participants were restricted to those with a hypertension diagnosis code with HTN drug exposure (Table S1). Reduction in systolic blood pressure (SBP) was used to determine the effectiveness of HTN medications by comparing SBP measurements after treatment to SBP measurements prior to the onset of treatment (SBP reduction). Further descriptions of data extraction and cleaning can be found in the Methods S1.

Study design

Figure 1 gives an overview of our phenotyping approach. We limited our analysis to oral drugs prescribed to 300 or more participants in an outpatient setting for over 28 days in All of Us (n = 19; Table S1).²⁶ We required both a calculable drug end date and a drug strength to be included in the study. To capture the first known exposure to each HTN drug in each participant, we assigned time t₀ (drug exposure start date) as the earliest date of each HTN prescription recorded in the EHR per individual. Drug exposure end dates were used where available, or were calculated using days of supply, quantity in prescription, and/or number of refills information, if available.

Figure 1.

Calculation of reduction in systolic blood pressure (SBP) using electronic health record (EHR) data. Reduction in SBP was calculated by subtracting the postmedication SBP from the premedication SBP. Premedication SBP was the closest SBP to the start of medications, but no more than 365 days beforehand. Postmedication SBP was the closest SBP to 28 days following medication use, but no more than 365 days following the start of medication. Individuals were required to have an active prescription at the postmedication SBP measurement date.

HTN medication effects on SBP have been shown to be dose-dependent.²⁷ Since most HTN medications with different dosing strengths are often prescribed using similar dosing regimens (e.g., lisinopril can be prescribed in 2.5, 5, 10, 20, 30, and 40 mg tablets, but is generally taken once daily), we extracted and used strength information as a proxy for dosage. Medication entries without strength information were omitted (12,965 out of 154,544, 8.39%). Strength information was used from the first drug exposure, and if unavailable, the minimum strength prescribed throughout the EHR for individuals was used instead. Drug exposure duration in years was calculated by determining the number of days between the start and end date divided by 365.25.

HTN medications were grouped as combination medications if they were either prescribed as a fixed-dose combination pill, or individual HTN medications both prescribed on the same day, and were analyzed separately from individually prescribed medications. Only medication entries with corresponding SBP measurements 0–365 days before the medication start date and 28–365 days after the medication start date were included (Figure 1). Only medication entries with medication end dates past the postmedication SBP measurement recorded were included.

To ensure that SBP-related results observed were in fact due to HTN drug effects, we included inhaled tiotropium as a negative control in our analyses because it is a maintenance medication and is prescribed at similar rates to HTN drugs. Tiotropium is a muscarinic antagonist typically prescribed for chronic obstructive pulmonary disorder that acts as a bronchodilator and has been shown to have no effect on systolic blood pressure.²⁸

Statistical analyses

Differences in the starting SBP at the time of the first HTN prescription among Hispanic, Black, and White groups were compared using linear regression. Models were adjusted for sex at birth, BMI, insurance status and education level, annual household income, prescription year, and age at first medication.

Differences in HTN medication effectiveness (reduction in SBP) among White, Hispanic, and Black groups were compared using linear regression. Analyses included an unadjusted model and a model adjusted for sex at birth, body mass index (BMI), days between SBP measurements, medication strength, insurance status, educational attainment, annual household income, prescription start year, age at first medication, and starting SBP. To ease the comparison of effect size differences and normalize for disparities in starting SBP on medication effectiveness, the adjusted model and the predict function in R were used to predict medication effectiveness if all participants had started medication at 140 mmHg. This function calculated the estimated reduction in SBP using the adjusted linear regression previously created with the original starting SBP values replaced with 140 mmHg. All other variables were kept the same.

For all analyses, we used White participants as the reference group to compare with Hispanic and Black populations since it identified the largest number of participants with each medication. We performed regressions for the negative control tiotropium, 19 HTN medications, as well as an aggregated measure for the first HTN drug prescription. The Bonferroni-corrected threshold of P < 0.0025 (a = 0.05, 20 tests) was used to assign statistical significance.

We calculated the number of unique HTN treatments taken per individual throughout their EHR history both by individual medication and by medication class in case of differences in within-class medication switching. We tested for differences in the number of HTN medications prescribed in Black and Hispanic, compared with White, populations using Poisson regression with an offset term added for the number of years from the first to the last medication taken per individual. We also calculated differences in proportions of first- and second-line HTN drugs prescribed in Hispanic, Black, and White populations using a Pearson’s chi-squared test. Additionally, we determined the percentage of first-line HTN prescriptions within Black, Hispanic, and White populations that adhered to the 2017 ACC/AHA clinical guidelines (Table S11), both within the entire span of EHR records, and limited to after 2018.

All statistical analyses were performed in R version 4.2.2. We visualized data with the ggplot2 and networkD3 package and the sunburst function from the plotly express package in Python version 3.7.12.^29–32 All statistical tests assumed a two-tailed distribution and were performed within the All of Us Researcher Workbench. Our research workspace is available upon request to any registered All of Us researcher and can be found in the project directory.

RESULTS

Participant demographics

The study population included 17,718 individuals: 9,863 White, 5,067 Black, and 2,788 Hispanic participants (Table 1). Medications included in this analysis (n = 19) ranged in sample size from 348 (labetalol) to 6,450 (amlodipine) participants (Table S2). Black and Hispanic participants reported lower household incomes than White participants; the median annual income selected on the income questionnaire by Black and Hispanic participants was $10,000–25,000, compared with $75,000–$100,000 for White participants (P < 1E-10; Table 1). More Hispanic (42.2%) and Black (24.0%) participants were missing income survey data than White participants (12.4%; P < 1E-10). More Black and Hispanic participants reported using Medicaid compared with White participants (45.8%, 55.3%, and 12.1%, respectively; P < 1E-10).

Table 1.

Study participant demographics. Demographics are summarized for all participants and stratified by self-reported race and ethnicity

Variable	Black	Hispanic	White	Total	P value (Black)	P value (Hispanic)
N	5,067	2,788	9,863	17,718
Female sex at birth (%)	67.67	66.39	55.10	60.48	<1E-10	<1E-10
Age at HTN medication (median years, IQR)	55.07 (47.13–62.00)	56.81 (47.96–64.88)	62.32 (53.49–69.25)	59.22 (50.29–66.92)	<1E-10	<1E-10
Starting SBP (median mmHg, IQR)	138 (122–152)	136 (123–149)	132 (120–146)	135 (120–148)	<1E-10	<1E-10
Reduction in SBP (mean mmHg, 95% CI)	4.74 (4.17–5.31)	4.48 (3.75–5.21)	5.11 (4.74–5.48)	4.91 (4.62–5.20)	0.28	0.14
Days between SBP measures (median, IQR)	63 (39–117)	63 (38–118)	70 (42–136)	67 (40–127)	<1E-10	1.45E-08
Year of first HTN prescription (median, IQR)	2016 (2014–2018)	2017 (2014–2019)	2016 (2013–2018)	2016 (2013–2018)	9.97E-06	<1E-10
Prescribed multiple HTN drugs (%)	58.46	49.90	52.35	53.62
Prescribed multiple HTN drug classes (%)	56.97	48.17	50.62	51.95
BMI (median, IQR)	32.4 (27.6–38.2)	30.8 (27.0–35.3)	29.8 (25.8–35.0)	30.6 (26.4–35.9)	<1E-10	9.30E-05
Income (median questionnaire selection)	10–25 k	10–25 k	75–100 k	35–50 k	<1E-10	<1E-10
Highest education attained
Less than high school (%)	14.43	31.85	2.44	10.50	<1E-10	<1E-10
High school/GED (%)	29.70	23.53	11.58	18.64	<1E-10	<1E-10
Some college (%)	30.29	22.17	26.58	26.95	1.70E-06	2.46E-06
College graduate (%)	12.83	12.91	27.14	20.81	<1E-10	<1E-10
Advanced degree (%)	9.24	5.85	31.47	21.08	<1E-10	<1E-10
Skip (%)	3.51	3.69	0.78	2.02	<1E-10	<1E-10
Medical insurance type
Employer/Union (%)	27.08	19.40	44.09	35.34	<1E-10	<1E-10
Purchased (%)	4.78	3.91	14.12	9.84	<1E-10	<1E-10
Medicaid (%)	45.77	55.27	12.07	28.50	<1E-10	<1E-10
Medicare (%)	36.33	35.11	56.42	47.32	<1E-10	<1E-10
Other (%)	4.92	4.85	6.34	5.70	<0.001	<0.001
None (%)	0.63	<0.72	<0.20	0.30	3.06E-06	0.11
Skip (%)	1.34	3.91	0.49	1.27	6.53E-08	6.93E-34

Statistical significance is based on unadjusted linear or logistic regressions, and significance determined by Bonferroni-adjusted P value (P < 0.05/30).

“<” nomenclature in percentage of demographics used in accordance with All of Us requirements to protect participant privacy and reduce risk for re-identification of groups with <20 participants.

Racial and ethnic differences in starting blood pressure

The starting SBP within Hispanic and Black populations was higher than in White populations (Figure 2a; Table S4). For 13 out of 19 HTN medications, Black participants were prescribed medications at a higher starting SBP than White participants. This was also true for 7 out of 19 HTN medications in Hispanic participants compared with White participants. Black and Hispanic participants were started on their first HTN medication at similar SBPs at a mean of 136.0 (95% CI = 135.3–136.7) mmHg and 136.3 (95% CI = 135.6–137.0) mmHg, respectively, while White participants were started on HTN medications at a mean SBP of 130.9 (95% CI = 130.5–131.4; P < 1E-10) mmHg. Our negative control tiotropium showed similar starting SBP in all three populations.

Figure 2.

Starting systolic blood pressures at onset of the first HTN medication (a), unadjusted effectiveness of HTN medications (b), and adjusted effectiveness of HTN medications normalized to a starting systolic blood pressure (SBP) of 140 mmHg (c) in self-reported Black, Hispanic, and White populations. Mean and 95% confidence intervals are displayed for each analysis. Each vertical panel represents a medication analyzed, with the grouped sample size per medication listed after the medication name in parentheses. Asterisks (*) indicate Bonferroni-adjusted significance as determined by linear regression in Black and Hispanic populations compared with White participants. The last panel on the right displays results from tiotropium, which was used as a negative control.

We analyzed age at drug exposure within populations to determine whether the difference in starting SBP was due to delayed blood pressure screening for Hispanic and Black, compared with White, populations. Our results showed that Black and Hispanic participants were a median of 7.2 and 5.5 years younger at the time of their first HTN medication (medians of 55.1 and 56.8, respectively) than White participants (median 62.3, P < 1E-10 for both; Table 1).

Racial and ethnic differences in HTN drug effectiveness

In unadjusted models, the average reduction in SBP across all HTN drugs was similar (~5 mmHg) in Hispanic, Black, and White populations (Figure 2b). In this analysis, only lisinopril was significantly less effective in Black (compared with White) participants (Figure 2b; Black mean reduction in SBP, 95% CI = 4.05, 3.05–5.05; White = 6.92, 6.29–7.55; Hispanic = 6.21, 4.99–7.43; p = 1.2E-06).

Higher starting SBPs were significantly correlated with greater reductions in SBP in response to HTN medication (Pearson r² = 0.50, 95% CI = 0.49–0.51, P < 2.2e-16). We observed significant differences in drug effects after adjusting for starting SBP and other covariates in our analysis. After adjusting for covariates and predicting reduction in SBP assuming a starting SBP of 140 mmHg, the first HTN drug prescribed, and all 19 HTN individual drugs, were trending to be less effective in Black compared with White participants, as well as 15 out of 19 medications in Hispanic compared with White participants (Figure 2c, Table S5). Six out of 19 medications reached statistical significance in effectiveness in both Black and Hispanic, compared with White, populations in adjusted models with normalized starting SBP (Table S6). Four out of the six medications that showed statistically significant differences in adjusted effectiveness were also the four HTN drugs with the highest sample sizes in our dataset: amlodipine (n = 6,450), lisinopril (n = 6,109), metoprolol (n = 4,704), and losartan (n = 4,231). Individual medications that were significantly less effective after normalizing for starting SBP in Hispanic and Black populations, compared with White, also included hydrochlorothiazide (HCTZ)/losartan (n = 453) and chlorthalidone (n = 1,159).

Although confidence interval overlap precludes making definitive conclusions about the most effective medications within Hispanic, White, and Black populations, there were trends toward the most effective HTN medications within populations after adjusting for starting SBP. Chlorthalidone had the greatest predicted reduction in SBP in White participants (predicted effectiveness from 140 mmHg starting SBP = 11.7, 95% CI = 10.7–12.6 mmHg; Table S6), while spironolactone had the greatest predicted reduction in SBP in Hispanic participants (mean = 10.1, 95% CI = 6.6–13.7 mmHg) and HCTZ/lisinopril had the greatest predicted reduction in SBP in Black participants (mean = 9.0, 95% CI = 6.3–11.7 mmHg). The combination drug HCTZ/losartan (diuretic/ARB) had the greatest range of predicted reduction in SBP between populations (Hispanic mean = −1.5; 95% CI = −8.1 to 5.0 mmHg; Black mean = 2.6, 95% CI = −1.3 to 6.5 mmHg; White mean = 11.3, 95% CI = 7.7–14.9 mmHg).

We found that six out of the seven most effective medications after adjusting for starting SBP in Black participants included diuretics (including one fixed-dose combination diuretic/ACEI) and CCBs, while the fifth most effective medication was clonidine, a central alpha2-adrenergic agonist. In Hispanic participants, four of the top seven medications included diuretics, while the others included clonidine (central alpha2-adrenergic agonist), lisinopril (ACEI), and atenolol (beta blocker). In White participants, four out of the top seven medications included diuretics (including one compound diuretic/ARB and one compound diuretic/ACEI), lisinopril (ACEI), clonidine (central alpha2-adrenergic agonist), and amlodipine (CBB).

Adjusting for socioeconomic (including insurance type, income, and college degree completion), lifestyle (including cigarette smoking, drug use, and alcohol use), and additional socioeconomic variable (including home ownership and employment status) covariates in our models did not alter the inferences we drew from initial (adjusted) associations (Tables S5, S8, S9). Medication strength was similar across Hispanic, Black, and White populations (Table S3). Since we found that the mean starting SBP was <140 mmHg for all populations, we repeated the analysis requiring a SBP >140 mmHg prior to HTN medication start. This analysis corroborated differences in starting SBP between populations and trends in medication effectiveness (mean, 95% CI White starting SBP = 153.24, 152.87–153.61 mmHg; Hispanic = 154.66, 153.97–155.35 mmHg; Black = 155.66, 155.13–156.19 mmHg), but with much-reduced samples sizes (Figure S1).

Racial and ethnic differences in the number of HTN drugs prescribed

A greater proportion of Black individuals were prescribed multiple HTN medication classes (57.0%) compared with Hispanic (48.2%) and White (50.6%) populations (P = 2.5E-9; Figure 3; Table 1; Table S7). This was also true when analyzing the number of unique HTN medications prescribed per individual (P = 8.9E-17).

Figure 3.

Cumulative percentage plot of the average number of HTN medication classes prescribed in self-reported Black, White, and Hispanic participants. The number of drug classes prescribed are plotted by self-reported race and ethnicity. Each dot shows the cumulative percentage per group that is prescribed up to the number of medications indicated on the x-axis. The percentage of participants prescribed multiple HTN drug classes was 56.97% in Black, 50.62% in White, and 48.17% in Hispanic populations.

Racial and ethnic differences in HTN drug prescribing patterns

Within our entire dataset, 60.5% of Black participants were prescribed a 2017 AHA/ACC guideline-recommended first-line HTN treatment (diuretic or CCB; ACEI or ARB if preexisting CKD; beta blocker if preexisting HF; Table S11), compared with 73.8% of White participants and 81.3% of Hispanic participants (diuretic, CCB, ACEI, or ARB; Figure S6).⁶ After selecting only participants who initiated pharmacological HTN treatment after 2018 (following the release of the ACC/AHA guidelines), we found that 63.5% of Black participants were prescribed a recommended first-line treatment, compared with 75.8% of White participants and 81.8% of Hispanic participants (Figure 4).

Figure 4.

First and second antihypertensive medication classes prescribed since 2018 in self-reported Hispanic, Black, and White populations. Sankey plots showing prescription patterns for the first and second HTN medication classes prescribed since 2018 in self-reported Black, White, and Hispanic individuals. The first column of each plot shows the first medication classes prescribed to individuals of each group—with bar size proportional to the fraction of the population first prescribed each class. The second column displays the proportion of second medication classes prescribed to individuals, with the white bar representing the proportion of individuals on a given medication class who were not ever prescribed a second medication class.

The patterns of first medications prescribed starting in 2018 for HTN medicines varied by racial and ethnic population (P < 1e-08; Figure 4). Starting in 2018, ACEI were the most prescribed first medication in White and Hispanic populations, while CCBs were the most prescribed in Black participants, which ranked third and second in White and Hispanic participants, respectively.

DISCUSSION

This study evaluated HTN medication effectiveness and prescribing pattern differences among self-identified White, Hispanic, and Black populations in All of Us. In comparison with White participants, Black and Hispanic participants were started on HTN drugs at higher SBPs, had a smaller reduction in SBP in response to HTN drugs, and were prescribed different first- and second-line therapies. Black participants were also prescribed more HTN medications than both Hispanic and White participants. Increasing efforts to follow clinical guidelines for first-line recommendations of HTN drugs may maximize the effectiveness of initial treatments and may help to attenuate the trend for more HTN prescriptions in the Black population. Some of these findings likely represent treatment disparities and thus are opportunities for improved treatment of diverse populations.

Black and Hispanic participants were started on HTN medications at higher SBPs than White participants, despite being started on HTN medications 7.3 and 5.5 years earlier, respectively. These findings support previous research and suggest an earlier need for HTN screening and therapy in Black and Hispanic populations.³³

Unadjusted models suggested that HTN medications resulted in similar reductions of SBP in Black, Hispanic, and White participants. However, when adjusted for covariates, six HTN drugs were less effective in both Black and Hispanic populations compared with White, and almost all trended less effective in Black and Hispanic populations. Given the overall trends, we hypothesize that most HTN medications examined would reach statistical significance as being less effective in Black and Hispanic compared with White participants if analyzed in a larger population.

Previous work has shown that the reduction in SBP upon initiation of a HTN medication is positively correlated with the starting SBP.³⁴ Similarly, we found an association between SBP reduction and starting SBP. Indeed, the higher starting blood pressures in Black and Hispanic populations were a driver of decreased effectiveness of HTN medications in these populations compared with White populations. Model adjustment for available indicators of socioeconomic status and modifiable lifestyle risk factors did not account for observed associations between race and ethnicity and HTN drug effectiveness. However, it is important to note that these results do not demonstrate that biologic factors are driving these observed differences in effectiveness. For example, unmeasured environmental, lifestyle, diet, or other social determinants of health may also affect drug effectiveness.

Our analysis of prescription patterns showed that since 2018, 63.5% of Black participants were prescribed a 2017 ACC/AHA guideline-recommended first-line HTN treatment, compared with 75.8% of White participants and 81.8% of Hispanic participants. Given prior studies of efficacy and the relative SBP effects of different classes of HTN medications in this study, it is possible that greater adherence to 2017 AHA/ACC Guideline-recommended first-line treatments could result in fewer total medications prescribed, especially in the Black population.

CCBs and diuretics were among the most effective medications in Black participants in our dataset after adjusting for starting SBP, which corroborates previous studies and the 2017 AHA/ACC Guideline recommendations for diuretics and CCBs as first-line prescriptions in the Black population.^17,35 In the Hispanic population, the medications with the greatest predicted effectiveness included diuretics, the central alpha2-adrenergic agonist clonidine, the ACEI lisinopril, and the beta-blocker atenolol. Notably, these top medications were different from both Black and White populations. However, overlapping confidence intervals caution the interpretation of these results.

Currently, 2017 AHA/ACC Guidelines do not recommend different first-line treatments in Hispanic patients. Our results suggest that it may be worthwhile to create a separate set of recommendations for HTN management in the Hispanic population, though further study would be needed. Since many patients started on one agent will later require additional agents to reach the blood pressure target, optimizing medication effectiveness in a manner which considers self-reported race and ethnicity may minimize the number of medications prescribed.^6,35,36

A previous study reported that personalizing HTN medication decisions between four commonly used HTN drugs (lisinopril, losartan, HCTZ, and amlodipine) can lead to a 4 mmHg greater reduction in SBP, compared with a nonpersonalized approach.³⁷ Research has demonstrated that a 10 mmHg reduction in SBP reduced the risk of major cardiovascular events by 20%, coronary heart disease by 17%, stroke by 27%, HF by 28%, and all-cause mortality by 13%, while a 5 mmHg reduction in SBP reduced the risk of major cardiovascular events by 10%, suggesting that personalized approaches that maximize SBP lowering will be beneficial to public health.^38,39 Reductions of mean SBP across both African American and White populations as small as 1 mmHg are associated with reductions in HF events, indicating that even modest SBP population-wide shifts can reduce CVD burden.⁴⁰

Our findings may aid personalized HTN medication decision making and help to mitigate inequities in hypertension-related morbidity and mortality among Black and Hispanic populations.⁴¹ The methods presented here also have a broad potential applicability among various drugs and outcome measures to assess disparities in various demographics. These methods can be used to bolster the push toward precision medicine by optimizing treatment based on individuals’ demographics to maximize benefits.

Limitations caution the interpretation of these results. We did not analyze other racial and ethnic populations and less common HTN medications due to limitations of sample size. The starting systolic blood pressures across all groups were lower than expected, although repeating the analysis while requiring both hypertension diagnosis and SBP >140 mmHg before HTN medication start found similar trends in starting SBP and effectiveness between populations. Our medication effectiveness analyses were not time-restricted, meaning that we did not only analyze individuals who started taking HTN drugs after the time of their HTN diagnosis. Therefore, we may be seeing the effects of individuals taking HTN drugs for other indications (e.g., spironolactone for ascites due to cirrhosis).^42,43 Due to the high number of indications for which HTN medications may be prescribed, we also caution interpretation of prescribing pattern results, which may be confounded by prescriptions for other indications.

This study is also limited by the accuracy and breadth of EHR data. We analyzed drug exposure records instead of participant-verified medications and date ranges. As such, we may not be capturing all HTN medications used by each participant or may not have complete EHR records reflecting the accurate start and end dates of each HTN medication used. Additionally, more than half of the participants in our analyses were prescribed multiple HTN medications. Those included in analyses for each drug could confound effectiveness measurements for subsequent drugs. The effect size and starting blood pressure of various drugs may also be influenced by the order of their prescription, particularly those that are not first-line drugs.

Additionally, differences in our results due to adherence to medication or early medication cessation (e.g., due to adverse side effects) would not be captured by our approach, as it is not possible to extract this information using EHR data. The disparities identified in the present study are important regardless of the underlying cause, and this study may warrant future follow-up studies to further identify the reasons for these disparities, which will help to implement actionable solutions. We also cannot account for social effects that may influence EHR data, such as “white coat hypertension,” which occurs when blood pressure readings are increased at a doctor’s office, typically due to stress associated with healthcare settings.⁴⁴ White coat hypertension has been shown to influence SBP readings measured in medical settings, and may differentially affect Hispanic, White, and Black populations.⁴⁵

To compare effectiveness between populations analyzed and among various medications, we adjusted our analysis for covariates and predicted effectiveness given a starting SBP of 140 mmHg. While we adjusted our analysis to provide a normalized indicator of each drug’s effectiveness, effectiveness data should be interpreted with caution given that it is transformed.

This work demonstrated disparities among Hispanic and Black, compared with White, populations in blood pressures at the start of HTN drugs, drug effectiveness, and drug prescribing patterns. White participants were started on HTN drugs at significantly lower SBPs than Hispanic and Black participants. Over their course of treatment, Hispanic and White participants were prescribed fewer medications, and a greater percentage of guideline-recommended first-line medications, when compared to Black participants. Our data demonstrated a greater drug effectiveness in White, compared with Black and Hispanic, participants. This study demonstrates the potential of longitudinal EHR data to rapidly generate medication effectiveness data across diverse populations and provides a scaffold for a systematic approach to determine drug effectiveness in EHR and may extend to other medication effectiveness studies.

Supplementary Material

Supplementary information accompanies this paper on the Clinical Pharmacology & Therapeutics website (www.cpt-journal.com).

Study Highlights.

WHAT IS THE CURRENT KNOWLEDGE ON THE TOPIC?

Non-Hispanic Black (Black) and Hispanic/Latino (Hispanic) populations have been underrepresented in clinical trials in the US despite NIH policy changes to mandate the analysis and reporting of race and ethnicity differences in phase III clinical trials in 2001. Previous research showed differences in the effectiveness of drug classes among Black and non-Hispanic White (White) populations. Calcium channel blockers (CCB) and thiazide-type diuretics were more effective antihypertensive (HTN) medications in Black individuals, while angiotensin-converting enzyme inhibitors (ACEI) or beta-adrenergic blockers (beta blockers) were more effective in White individuals. The breadth of data supporting the differences in effectiveness in the Black population has led to different recommendations for first-line HTN medications by the 2017 ACC/AHA/AAPA/ABC/ACPM/AGS/APhA/ASH/ASPC/NMA/PCNA Guideline for High Blood Pressure.

WHAT QUESTION DID THIS STUDY ADDRESS?

Are there disparities in the prescription patterns and effectiveness of antihypertensive medications in White, Black, and Hispanic populations?

WHAT DOES THIS STUDY ADD TO OUR KNOWLEDGE?

We determined the effectiveness and prescribing patterns of 19 common antihypertensive medications in self-identified White, Black, and Hispanic populations. Black and Hispanic participants were started on medications at higher systolic blood pressures, and many of the antihypertensive drugs were less effective in these populations. Black participants were prescribed more antihypertensive medications and fewer Black participants were started on recommended first-line medications.

HOW MIGHT THIS CHANGE CLINICAL PHARMACOLOGY OR TRANSLATIONAL SCIENCE?

This real-world evidence highlights the potentials for earlier intervention and greater adherence to first-line medication recommendations for Black and Hispanic patients with hypertension.