INTRODUCTION

Hypertension costs the USA an estimated $131 billion yearly,1 and is the primary or contributing cause of over half a million US deaths each year.2 Over one-fifth of adults in the USA take an antihypertensive medication, which makes them the most used type of medications in the USA.3, 4 Antihypertensives are also among the best-studied of all drugs, with at least 464,000 people randomized into antihypertensive clinical trials by 2003.5 They are also the topic of multiple high-profile clinical practice guidelines and performance measures.6, 7 Which antihypertensive drugs are used, why they are used, and ensuring that effective clinical research exists to use the most effective drugs possible could have a large public health impact. But which drugs are used and for what reason is not always known.

A recent study assessing trends overtime in blood pressure management in the outpatient setting saw increases in multidrug therapy for hypertension with angiotensin-converting enzyme inhibitors, calcium channel blockers, and beta-blockers as the most frequently prescribed class of blood pressure medications in 2016.8 Nevertheless, there remain relatively few studies of the determinants of the use of specific classes of antihypertensive medications and even fewer examining which drugs are used within specific classes.3, 9,10,11,12 Few trials have directly compared drugs within classes. As a result, guidelines and previous studies have mostly referred to medications as classes rather than unique individual compounds. However, indirect comparisons have provided evidence that there could be important intra-class differences.13,14,15,16 Therefore, understanding the trends of between-class and within-class clinical use of antihypertensives could help guide both implementation and research going forward.

Given this context, the primary goal of this study is to gain a better understanding of the between- and within-class trends in antihypertensive medication use over the last 20 years. Due to the complexity of individual decision-making, this descriptive study compiles a hypothesis generated list of observations based on an understanding of the literature. We hypothesized that between-class trends occur in relation to cost of medications, whereas within-class trends of antihypertensive agents occur in relation to availability and generic conversion.

METHODS

Study Participants

A repeated cross-sectional study of the Medical Expenditure Panel Survey (MEPS) was used to describe trends in antihypertensive medication use over time.17 MEPS is sponsored by the Agency of Healthcare Research and Quality (AHRQ) and is representative of the non-institutionalized population of the USA. Each year of the survey is comprised of two overlapping cohorts, which are interviewed five times over the 2 years. The survey includes information regarding demographic, socioeconomic, and prescription medication utilization. All adult individuals included in the survey between 1997 and 2017 were included in the analysis.

Study Variables

Antihypertensive medications were identified through a combination of therapeutic classification and prescription medication name. The survey uses numerous methods to achieve accurate reporting of prescription medications including contacting pharmacies.18,19,20 Previous studies have validated the accuracy of chronic medication reporting in MEPS.21 Included classes in the study were angiotensin-converting enzyme inhibitor (ACE-I), thiazide diuretic, angiotensin receptor blocker (ARB), beta-blocker, loop diuretic, dihydropyridine calcium channel blockers (CCB), non-dihydropyridine CCB, aldosterone receptor antagonists, hydralazine, and clonidine. Other drug classes were not studied given low levels of use and/or numerous therapeutic uses across numerous categories. Within these classes, individual medications were identified by name. Combination medications were reported in applicable categories. An individual was identified as a medication user if he or she had any use of the medication during a given year. Generic availability of medications was identified as the first approved Abbreviated New Drug Application (ANDA) through the Food and Drug Administration (FDA).

Analysis

Complex survey weights were used in all analyses to make these analyses representative of the non-institutionalized population of the USA. The figures report 95% confidence intervals.

The OhioHealth Institutional Review Board ruled the study exempt. All analyses used STATA15 (College Station, TX).

RESULTS

Between 1997 and 2017, 493,596 individuals were identified. We identified 43 relevant antihypertensive medications among 10 classes (Table 1). All but 4 of these drugs were approved before our study dataset began in 1997, but 26 achieved generic status between 1997 and 2016. The newest drugs were 3 angiotensin receptor blockers and nebivolol.

Table 1 Antihypertensive Medications Brand and Generic Approval Dates
Full size table

The proportion of individuals taking antihypertensive medication during a year increased from 1997 to the early 2010s (Fig. 1). The proportion taking 2, 3, and 4 medications all followed a similar pattern. However, from 2015 to 2017, the proportion of the population taking 2 or more medications declined, while the number taking one increased.

Figure 1
figure 1

Proportion of the adult population on antihypertensives who are on a particular number of antihypertensive medication classes between 1997 and 2017. A user was defined by the report of any medication in a medication class. Data from the 1997 to 2017 Medical Expenditure Panel Survey were used. Brackets represent 95% confidence intervals.

Full size image

Use of drugs within classes changed substantially over time (Fig. 2). Use of ARBs and dihydropyridine CCBs increased substantially during the study period. The proportion of the population who used ACE-Is increased until 2010 after which rates remained stable before a small decline in 2016–2017. Beta-blocker use was similar to that of ACE-Is with an earlier decline starting in 2012. Thiazide diuretic use increased from 1997 to 2007, leveled off until 2014, and declined from 2015 to 2017. Non-dihydropyridine CCB use declined throughout the study period. Loop diuretic, spironolactone, and clonidine rates of use were largely unchanged, while the proportion of people using hydralazine increased slightly.

Figure 2
figure 2

Proportion of Adult Population Using Antihypertensive from Medication Classes During a Year Using the 1997–2017 Medical Expenditure Panel Survey. Brackets represent 95% confidence intervals. ACE, angiotensin-converting enzyme; CCB, calcium channel blocker.

Full size image

Within classes, there was a striking tendency toward one drug developing one dominance in the class, but when the drug became dominant varied greatly (Fig. 3). The thiazide diuretic hydrochlorothiazide and the loop diuretic furosemide were dominant before 1997 began and became more so by 2017. The ACE-I lisinopril, ARB losartan, beta-blocker metoprolol, and CCB amlodipine all became dominant. Within-class dominance sometimes followed generic conversion, but sometimes predated it. The dominant drug was the first introduced in 3 of 6 major classes.

Figure 3
figure 3

Information on the Proportion of Adult Population using Different Medications by Medication Class by Year. The 1997–2017 Medical Expenditure Panel Survey was used. The Y-axis represents the proportion of the adult population. Large diamonds represent the year a medication had an approved generic (also see Table 1). Brackets represent 95% confidence intervals. ACE, angiotensin-converting enzyme.

Full size image

DISCUSSION

Our findings show a mismatch between within-class and between-class use choices of antihypertensive medications. Within classes, we found that most antihypertensive medication classes eventually equilibrate to one dominant medication. There was no clear evidence that the medication that became dominant did so for reasons of formal studies demonstrating superiority within the class. The dominant medication was not consistently associated with early entry or early generic availability. Drugs with some evidence of comparative advantage (i.e., chlorthalidone, torsemide)22,23,24,25,26 did not gain dramatical change in use, while atenolol did appear to decrease in use with less robust evidence.27,28,29

However, between-class medication use may be partly associated with key studies like SPRINT, and clinical practice guidelines.30 As guidelines until 2014 recommended generally intensifying treatment, overall use followed. With the decline in treatment intensity recommended in the 2014 clinical practice guidelines,31 there was a small change in clinical practice. Treatments did not appear to change dramatically after either ALLHAT or SPRINT trials.32 Medications with more negative harm profiles had declining use over time.33,34,35 Beta-blockers and CCBs had more movement over the course of the study, which might have been related to mechanistic differences and alternative uses of individual medications.

These data do raise the concern that if small-to-moderate differences in effect or adverse events within classes exist (i.e., phenformin in biguanides or troglitazone in thiazolidinediones), current practice may not recognize those differences. For instance, if one of the numerous ACE-Is lowered overall mortality or disease-oriented outcomes by 10% more than other drugs of that class, it could improve the health or adherence to medications of a population, though would require very large clinical trials to be effective, likely four times the size of the SPRINT trial, assuming otherwise similar populations. Determining when medications and medication classes will justify these studies would be challenging and require foresight. Although a few comparative trials are underway (e.g., ClinicalTrials.gov NCT03978884, NCT03928145, NCT02185417), these obstacles are formidable. Alternatives to classical clinical trials offer some hope, but challenges remain with implementing these types of trials.36

Although it is beyond the scope of this study to determine causal drivers of the trends observed in this paper, there are a few potential hypotheses as it relates to between-class prescribing. Previous studies have noted geographic and specialty variation in adherence to clinical practice guidelines for hypertension,37, 38 as well as differences in prescribing for younger and older patients suggesting consideration of risks/benefits of different medications.39, 40 Between-classes variation may also be due to tolerability of side effect profiles41 and generic availability, as well as physicians-in-training adopting practice style of senior physicians.

A few limitations of this study are worth noting. First, the high-level (ecological) observational nature of our data makes strong explanatory conclusions inappropriate. We cannot rule out residual confounding, demographic or health changes, costs, or other unmeasured causes of change in use, including advertising and promotion.42,43,44,45 Additional limitations include possible under-reporting of medications, a lack of biometric measurements, and potential utilization of certain medications for non-blood pressure–related purposes.

CONCLUSION

We found that over the last 20 years, antihypertensive drug classes have tended to equilibrate on one in-class medication. Additionally, there have been shifts in the types and amount of medications used in the population. Future research on comparative effectiveness for within-class medications early in the life cycle that are probable to have widespread use might improve health. In the absence of within-class randomized clinical trials, future funding for building the infrastructure for randomized evaluations in the ambulatory setting might be an avenue to do these trials, though these study designs are not without limitations.30 Without this renewed focus on within-class drug efficacy, prescribing patterns for antihypertensive medications within-class and between-class may continue to be done in the presence of considerable uncertainty.