Antihypertensive Prescribing Practices: Impact of the Antihypertensive and Lipid‐Lowering Treatment to Prevent Heart Attack Trial

Abstract

This national study examines the impact of the Antihypertensive and Lipid‐Lowering Treatment to Prevent Heart Attack Trial (ALLHAT) on the prescribing of thiazide‐type diuretics and other antihypertensive medications for patients with newly diagnosed hypertension. A cross‐sectional analysis was conducted using data from a national network of electronic health records for 2 groups with newly diagnosed hypertension and started on antihypertensive medications: one a year before and the other a year after the publication of ALLHAT. The percentage of new hypertensives started on thiazides increased from 29% pre‐ALLHAT to 39% post‐ALLHAT. An increase was also seen for angiotensin receptor blockers, while prescribing for angiotensin‐converting enzyme inhibitors, calcium channel blockers, and β‐blockers declined. There was no significant change in prescriptions for α‐blockers. Prescriptions for thiazides for patients with newly diagnosed hypertension increased after the publication of ALLHAT. Data from large national trials can have a considerable impact on prescribing practices.

Hypertension is a major reason for visits to outpatient offices in the United States. ¹ Numerous studies have shown that elevated blood pressures increase the risk of cardiovascular complications, including myocardial infarction, heart failure, and stroke, ² , ³ , ⁴ , ⁵ and that treatment reduces these morbidities and associated mortality. ⁶ , ⁷ , ⁸ It appears that these benefits result primarily from the lowering of blood pressure regardless of medication used. ⁹ , ¹⁰ , ¹¹ , ¹²

Results of the Antihypertensive and Lipid‐Lowering Treatment to Prevent Heart Attack Trial (ALLHAT) were published in December 2002. ¹³ ALLHAT was a large, randomized, double‐blinded, multicenter clinical trial that evaluated the use of an angiotensin‐converting enzyme (ACE) inhibitor, calcium channel blocker (CCB), or α‐blocker‐based regimen compared with a thiazide diuretic program on major cardiovascular disease outcomes. The findings from ALLHAT showed no difference among 3 of the study medications (thiazide‐type diuretic, ACE inhibitor, and CCB) in terms of the primary outcomes of fatal or non‐fatal coronary heart disease, and no difference in mortality. The number of combined cardiovascular events was 25% lower for patients on thiazide‐type diuretics compared with those taking the α‐blocker and, for this reason, the α‐blocker arm of the study was stopped early.

For secondary outcomes, individuals on thiazides had less incident heart failure and stroke compared with patients on the ACE inhibitor and less heart failure compared with those on the CCB. Given the low cost of thiazide‐type diuretics, the ALLHAT results, plus other studies showing no difference in primary outcomes and possible secondary outcome advantages over other agents, prior recommendations that thiazide diuretics be considered for initial treatment of uncomplicated hypertension were further solidified. ¹⁴

Various studies have shown that physicians do not closely follow guidelines when prescribing antihypertensive medications. ¹⁵ , ¹⁶ , ¹⁷ Some investigators have speculated that concern regarding side effects of older antihypertensive medications or lack of knowledge and confusion about current research findings may contribute to poor guideline adherence. ¹⁸ , ¹⁹ One study showed that physicians who are aware of the current literature on treatment of hypertension are more likely to follow prescribing guidelines. ¹⁶

The impact of the ALLHAT findings on national prescribing levels is unknown. There is some indication of an impact of the ALLHAT findings within closed systems like health maintenance organizations, ²⁰ but it is unclear what the effect has been nationally. The purpose of this study was to compare the types of antihypertensive medications prescribed to patients with newly diagnosed hypertension before and after ALLHAT publication using a national electronic health record (EHR) database.

METHODS

We investigated the impact of the ALLHAT findings on the use of thiazide‐type diuretics and other medications for hypertension by examining prescribing in the year before the dissemination of the study findings and in the year following the release of the findings. The database used in this study was the Medical Quality Improvement Consortium (MQIC), which is a network of outpatient offices that use the EHR known as Logician (MedicaLogic/Medscape, Inc, Hillsboro, OR). Each practice regularly downloads blinded clinical data into a central secure repository, including limited demographic information, medications, diagnoses, laboratory data, and other clinical data such as BP and weight. The data are then cleaned and standardized by an information technology team. Once standardized, the data are moved into a reporting dataset that can be used for primary care research and quality‐of‐care projects.

At the time of this study, MQIC comprised more than 3000 physicians and other providers (85% primary care) from more than 50 institutions in 28 states across the country. The database included more than 2 million patients.

We analyzed antihypertensive medications prescribed for 2 groups of individuals 20–80 years of age with a new diagnosis of hypertension. The first group represented individuals diagnosed and treated between December 1, 2001, and November 30, 2002, which is the year before the publication of ALLHAT. The second group represented a similar cohort from January 1, 2003, through December 31, 2003, which is the year following ALLHAT publication. For inclusion, each subject had to have a new diagnosis of hypertension within the respective year, at least 1 antihypertensive medication started after the hypertension diagnosis, and no other antihypertensive medications documented in the record before the diagnosis. In particular, we documented the initiation of treatment by medication category including thiazide‐type diuretics, β‐blockers, ACE inhibitors, CCBs, angiotensin receptor blockers (ARBs), α‐blockers, and others. Combination medications were counted as 2 separate medications and analyzed as part of each medication's respective category (eg, for a subject started on an ACE inhibitor/thiazide combination, we counted them separately as an ACE inhibitor and a thiazide). Therefore, the number of individual medications started in this population is greater than the total number of subjects. Lastly, to help ensure that this was a new diagnosis of hypertension, each patient had to have at least 1 documented visit in the EHR before the start of the hypertension diagnosis.

This study was approved by the Christiana Care (Wilmington, DE) internal review board.

Data Analyses

Statistical analyses were performed using SAS software (SAS Institute Inc, Cary, NC). Demographic information including patient sex, and average age was computed for each group. P values were generated using chi‐square for sex and a t test for age. We used logistic regression analysis to compute odds ratios (ORs) for each medication category, adjusting for age and sex. The group in the year before ALLHAT publication was the reference group. We were unable to control for race/ethnicity, because these data were inconsistently available in the MQIC database.

RESULTS

In our primary analysis, there were 5950 patients in the pre‐ALLHAT group and 7706 patients in the post‐ALLHAT group. The demographic information on age and sex is summarized in Table I. Of note, the 2 groups were similar in mean age but differed somewhat in sex.

Table I.

Demographics of Patients Newly Diagnosed With Hypertension Pre‐ and Post‐ALLHAT

	Pre‐ALLHAT (n=5950)	Post‐ALLHAT (n=7706)	P
Sex, %			.018*
Women	53.7	55.7
Men	46.3	44.3
Mean age (SD), y	53.3 (13.52)	52.89 (13.64)	.08†
ALLHAT indicates Antihypertensive and Lipid‐Lowering Treatment to Prevent Heart Attack Trial. *Chi‐square. †t test.

In the year before ALLHAT publication, 29.4% of patients diagnosed with hypertension were started on a thiazide diuretic, which increased to 39.1% in the year after ALLHAT. Results of the logistic regressions with age and sex as covariates are presented in Table II. For thiazide diuretics, for example, the odds of having a thiazide‐type diuretic prescribed for the post‐ALLHAT group was 53% higher (OR, 1.53; 95% confidence interval [CI], 1.43–1.65) than for the group before ALLHAT publication. Use of the newer ARBs also increased (adjusted OR [AOR], 1.17; 95% CI, 1.06–1.30). Conversely, prescribing rates for all other medication categories decreased in the post‐ALLHAT group, including ACE inhibitors (AOR, 0.81; 95% CI, 0.76–0.87), CCBs (AOR, 0.80; 95% CI, 0.72–0.88), and β‐blockers (AOR, 0.91; 95% CI, 0.84–0.98). Prescribing for α‐blockers was low for both groups; there was no change seen between them (AOR, 0.74; 95% CI, 0.54–1.03).

DISCUSSION

Using national data from a network of practices using the same EHR, this study indicates that there was a change in prescribing patterns from pre‐ to post‐ALLHAT. There was a significant increase in thiazides for patients with newly diagnosed hypertension after ALLHAT, consistent with current recommendations. ACE inhibitors and CCBs showed the largest decrease in the post‐ALLHAT group; this was expected, as these medications were part of the ALLHAT study. The nonsignificant decrease in use of α‐blockers was likely due to low power from the small sample size of individuals receiving them. Given the other findings of this study, a significant decrease in α‐blocker use would also be expected, since that arm of the ALLHAT study was terminated early due to significant adverse events compared with thiazides. In fact, one study has documented a decrease in the prescribing of α‐blockers after the early termination of that arm. ²¹

The increase in prescribing for ARBs during the study period is not completely unexpected. At the time of the study, ARBs were a relatively new class of antihypertensive medication coming into use. Extensive pharmaceutical marketing at the time may have played a role in this increase. Also, the ARB class of medications was not part of the ALLHAT study and therefore head‐to‐head comparisons of cardiovascular outcomes with the other classes of medications were not available to influence prescribing practices in either direction. Therefore, prescribing of ARBs may have increased independent of ALLHAT or may have increased to a lesser extent than if ALLHAT had not been published.

Results using our national dataset stand in contrast to past studies suggesting that physicians do not closely adhere to recommended treatment guidelines for hypertension treatment. The results are nonetheless consistent with those reported by Xie et al ²⁰ using regional health maintenance organization data pre‐ and post‐ALLHAT. This study demonstrated that the ALLHAT evidence may have had a positive impact on prescribing patterns in a national sample. Of note, treatment guidelines from the Seventh Report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure (JNC 7) ²² were published 7 months before the end of our study period and could have partially influenced the outcome. It remains appropriate, however, to use ALLHAT publication as the dependent variable, because that study influenced the subsequent JNC 7 recommendations.

Strengths of this study include the use of data from a national sample of users of an EHR that give direct information from the point of care and are reflective of physician prescribing patterns. Previous studies show that EHR data are useful in determining physician practice patterns and quality of care for various outcomes on a national scale. ²³ , ²⁴ , ²⁵ The EHR may have been partially instrumental in this case in affecting prescribing patterns relatively rapidly after ALLHAT publication. While the EHR and MQIC consortium do not provide specific recommendations for preferred treatment or prompts within the medical record, the EHR gives faster access to these if users decide to create their own individually or on an institutional level. Secondly, this study compared similar groups before and after ALLHAT, suggesting some influence from the ALLHAT results. Lastly, looking at patients with a new diagnosis of hypertension was useful since this reflects initiation of therapy and does not require differentiation between patients who may have experienced intolerance or lack of response to previous therapy.

Limitations

Some limitations of our study are noted. First, we were unable to control for provider type. Providers may be any users of the EHR system, and not just those who primarily treat adult hypertension (family physicians, internists, cardiologist, or nephrologists). Second, we cannot ensure that hypertension was a new diagnosis for all patients within our study, although steps were taken to control for this, as previously noted. Given that most patients were taking only 1 antihypertensive medication, however, it is reasonable to believe that the large majority, in fact, had a new diagnosis of hypertension.

Third, we were unable with our dataset to control for compelling indications requiring specific classes of antihypertensive medications. JNC 7 ²² lists heart failure, post‐myocardial infarction, high coronary disease risk, diabetes, chronic kidney disease, and recurrent stroke prevention as compelling indications for specific antihypertensive medication use. It is unlikely, however, that compelling indications made much difference in our study, for several reasons. Of the compelling indications, all but 2 include thiazide diuretics as part of the recommended treatment regimen. ²² Chronic kidney disease and post‐myocardial infarction treatment do not include thiazides. Most of these require care from specialists, however, and in our study, 85% of providers were primary care providers. Also, study subjects had a new diagnosis of hypertension, so it is unlikely that many of the patients had any of these compelling indications for other medications.

CONCLUSIONS

This study demonstrates that significant clinical trial findings can translate into useful changes in clinical practice. Moreover, physicians appear willing to adopt recommendations to use older medications instead of newer agents when the evidence is strong.

Acknowledgments and disclosure: The authors would like to acknowledge Wan Yin Li, Department of Family and Community Medicine, Christiana Care, Wilmington, DE, and Mark A. Carnemolla, Department of Family Medicine, Medical University of South Carolina, Charleston, SC, for their help with statistical analysis. Support for this project was provided in part by grants No. 5D55HP002220200, 1D14HP00161, and 1D12HP00023 from the Health Resources and Services Administration. A modified version of this paper was presented as a poster at the Society of Teachers of Family Medicine Conference, May 2005, in New Orleans, LA, and at the North American Primary Care Research Group Annual Meeting, October 2005, Quebec City, Quebec, Canada.