Impact of Clinical Trial Results on National Trends in α-Blocker Prescribing, 1996-2002

Context Research on factors that influence prescribing patterns and the extent of change produced by clinical trial findings is limited.

Objective To examine the changes in prescribing of α-blockers for hypertension treatment before and after the April 2000 publication of the unfavorable Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial (ALLHAT) early termination involving the study's doxazosin mesylate arm. Changes in prescribing were considered in the context of other potential concurrent influences on medication use between 1996 and 2002, including changes in α-blocker drug prices, generic conversion, drug promotion, and competition.

Design, Setting, and Patients Using 2 national pharmaceutical market research reports published by IMS HEALTH, α-blocker prescription orders reported in the National Prescription Audit—a random computerized sample of about 20 000 of 29 000 retail, independent, and mail order pharmacies and mass merchandise and discount houses—and office-based physician α-blocker prescribing patterns reported in the National Disease and Therapeutic Index—a random stratified sample of about 3500 physician offices—were tracked.

Outcome Measures Trends in physician-reported use of α-blockers and α-blocker prescribing and dispensing by US pharmacies.

Results There were steady increases in α-blocker new prescriptions, dispensed prescriptions, and physician drug use from 1996 through 1999. There was a moderate reversal in these trends following ALLHAT early termination and subsequent publications in early 2000. Between 1999 and 2002, new annual α-blocker prescription orders declined by 26% (from 5.15 million to 3.79 million), dispensed prescriptions by 22% (from 17.2 million to 13.4 million), and physician-reported drug use by 54% (from 2.26 million to 1.03 million). Other potential influences did not appear to have contributed significantly to this decline although cessation of α-blocker marketing may have hastened the decline.

Conclusions Modest yet statistically significant declines in the use of doxazosin and other α-blockers coincided with the early termination of the ALLHAT doxazosin arm. Although physicians responded to this new evidence, strategies to augment the impact of clinical trials on clinical practice are warranted.

Clinical trial results and practice guidelines are seen as the foundation of evidence that should have primary influence over physician practice, including patterns of medication prescribing. Past analyses, however, indicate limited change in physician practice that have resulted from the publication of clinical trial findings.^1-3 Analysis of results from randomized clinical trials of cardiovascular drug prescribing patterns has suggested that more incentives are needed to effect change.⁴ Highlighting such factors as relative risk reduction, numbers needed to treat, and increased quality of life may improve the impact on physician prescribing behavior.^5,6

Likewise, reports on the influence of practice guidelines on physician behavior indicate that few studies show strong adherence to guidelines,^7,8 whereas many studies suggest moderate to high levels of nonadherence with recommendations.^9-11 Reasons cited for suboptimal adherence to guidelines include lack of resources or awareness and physician disagreement with what constitutes best practice.^12-15 A variety of other factors also may influence medication prescribing practices. Pharmaceutical promotion may have a direct influence on physician prescribing practices.^16-18 Drug prices and other characteristics of the pharmaceutical market also may determine patterns of medication use.

α-Blocker use for the treatment of hypertension provides a pertinent example for gauging the influence of clinical trial results and guidelines on physician prescribing patterns in the context of other potential influences. In spring 2000, significant changes in recommendations for α-blocker use occurred as a result of early, unfavorable results from the Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial (ALLHAT). These results indicated an increased risk associated with use of the α-blocker doxazosin mesylate compared with diuretics.¹⁹ The early termination results were widely disseminated in news releases and journal publications.^20,21 With 50 million individuals in the United States diagnosed as having hypertension incurring annual treatment costs of $15.5 billion,²² these early ALLHAT results would be expected to have had broad implications on prescribing decisions.

Using nationally representative data available from IMS HEALTH, we tracked trends in α-blocker prescriptions filled by retail pharmacies and reports of α-blocker use in patient encounters with office-based physicians. We considered the impact of the early ALLHAT results and subsequent clinical recommendations in the context of other potentially influential factors, including drug promotional efforts, variations in drug price, generic conversion of α-blockers, and market competition.

Our analytic goals were 2-fold: to describe patterns of α-blocker use before and after the April 2000 publication of the early ALLHAT results and to examine whether these clinical trial results or alternative influences were associated with changes in α-blocker prescribing that occurred in this time frame.

Our analysis focused on the traditional unselective α-blockers (Universal System of Classification [USC] code 31440) doxazosin (Cardura), terazosin hydrochloride (Hytrin), and prazosin hydrochloride (Minipress). These medications function as antihypertensives by blocking action against α₁-receptors that mediate the constriction of blood vessels through increased smooth muscle tone. We also separately considered the use of the non-antihypertensive, selective α-blocker tamsulosin hydrochloride (Flomax) (within USC code 24200). Like traditional α-blockers, tamsulosin is used to treat benign prostatic hyperplasia and could have been substituted for traditional α-blockers.

Data tracking the use of α-blockers were obtained from 2 data reports published by IMS HEALTH, an independent pharmaceutical market research company in Plymouth Meeting, Pa. Information was obtained on patient visits to office physicians from the National Disease and Therapeutic Index (NDTI) and on retail pharmacy activity from the National Prescription Audit Plus (NPA).

Retail Pharmacy Prescribing and Dispensing of Medication. The NPA consists of a national, random, computerized sample of approximately 20 000 of a total of 29 000 retail pharmacies, independent pharmacies, mass merchandise and discount houses, and mail order pharmacies. The universe of 29 000 stores sampled in the IMS HEALTH database accounts for more than half of all retail pharmacies in the United States. The sample is resized on a semiannual basis. Although pharmacy data are collected every day, we used data reported in monthly aggregates, with the major unit of measure being the prescription. Prescription data consist of 2 types: prescription orders received by pharmacies and actual medications dispensed by pharmacies and received by consumers.

The NPA Therapeutic Category Report shows monthly national data on both the inflow of prescriptions written by physicians and the retail outflow of filled prescriptions moving from pharmacies to consumers. For each dose of each brand and generic drug name, data are reported on the number of prescriptions, the number of new and refill dispensed prescriptions, total dollar charges of all dispensed prescriptions, and drug manufacturer. New prescriptions are those ordered by physicians; most commonly the written physical orders are received by pharmacies to be processed and then dispensed to patients. We tracked quarterly national prescribing and dispensing patterns for α-blockers at retail pharmacies from 1996 through 2002, following individual trends for brand name and generic versions of doxazosin, prazosin, and terazosin.

We identified the number of both new prescriptions and total dispensed prescriptions as the likely best indicators of physicians' attitudes toward the use of α-blockers. In contrast, the dispensing of refills reflects the accumulation of current and past prescribing practices. Because drug use reported by office physicians indicates the medications that were newly prescribed or continued at the conclusion of a visit, decreasing drug use would suggest physician decisions to discontinue α-blocker use.

Prescribing by Office-Based Physicians. The NDTI is a continuing survey of a random stratified sample of approximately 3500 office-based physicians in the United States that is subject to attrition with replacement. It provides nationally representative information on drug use patterns for patients treated by office-based physicians. The surveyed physicians were selected by random stratified sampling by specialty using master lists of the American Medical Association and the American Osteopathic Association (both in Chicago, Ill). Participating physicians were surveyed for 2 consecutive workdays per quarter and provided information on each patient encounter during this period. Physician contact with patients largely consists of office visits although encounters with patients during telephone calls and hospital and nursing home visits also are included. The participation rate among physicians for reporting their prescribing patterns was at 84% in 1998, which was similar for other years. Physicians report information on diagnoses for each patient and then report all new or continuing drugs for each diagnosis. The NDTI does not capture patient adherence to medications and reflects a physician's best knowledge of medications taken by the patient. In NDTI, each reported use of a drug constitutes a drug use. Both brand and generic names are recorded in the survey. We used quarterly prescribing data from 1996 through 2002 to assess the likelihood that α-blockers were mentioned during office visits by patients reported to have hypertension.

ALLHAT Publication. ALLHAT was a large-scale, 9-year randomized clinical trial conducted from February 1994 to March 2002.²³ One of its goals was to evaluate the possible superiority of α-blockers, calcium channel blockers, and angiotensin-converting enzyme (ACE) inhibitors over less expensive, generic diuretics in lowering blood pressure and preventing coronary heart disease and other cardiovascular events in high-risk hypertensive persons 55 years and older.²⁴ In January 2000, a decision was made to terminate the doxazosin arm of the ALLHAT study after finding that the risk of combined cardiovascular events with the α-blocker doxazosin was more than 25% higher than that among patients treated with chlorthalidone, a thiazide diuretic.²⁵ The doxazosin arm findings were reported in a press release issued by the National Heart, Lung, and Blood Institute in March 2000, followed by publication in JAMA on April 15, 2000. Although we concern ourselves herein with the early α-blocker ALLHAT results, final results that suggest modestly superior outcomes with use of chlorthalidone over lisinopril and amlodipine were published in JAMA on December 18, 2002.²²

National Recommendations and Guidelines. In March 2000, following the ALLHAT early termination, clinical recommendations were released suggesting that α-blockers may not be the best choice for initial therapy of hypertension.²⁰ Previous Joint National Committee (JNC V) guidelines in 1993 had suggested the initial use of diuretics and β-blockers but with possible consideration of α-blockers, calcium channel blockers, and ACE inhibitors as other first-line agents.²⁶ In 1997, JNC VI guidelines emphasized the use of diuretics and β-blockers over α-blockers and other medications except as dictated by compelling comorbidities.²⁷

Drug Prices. The NPA retail pharmacy data provide total drug prices for antihypertensive medications. The mean retail drug cost per prescription was derived by dividing total dollar cost to pharmacies by the total number of prescriptions dispensed. We specifically followed drug prices per prescription from 1996 through 2002 for individual α-blocker drugs. We examined temporal fluctuations in drug prices of α-blockers in relation to prescribing patterns to gauge their potential influence on physician choices in drug prescribing.

Drug Promotion. The IMS HEALTH Office Promotion Report (OPR) is an ongoing survey of a national sample of approximately 3862 US office-based physicians. Survey participants are chosen from a random stratified sample of physicians from across 25 specialties in 43 states. The OPR data are reported in monthly aggregates and are stratified by pharmaceutical company, medication class, and drug name.

The OPR provides information on physician contact with pharmaceutical sales representatives, including estimates of corporate expenditures in promotional activities related to individual products by these sales representatives. The dollar value is derived by multiplying the reported minutes of detailing associated with each product by a current average cost per minute estimate. We examined quarterly total dollar expenditure for α-blocker office promotion from 1996 through 2002 to monitor trends in drug promotion.

Other Market Factors. Other commercial factors were explored as possible influences on physician prescribing behavior. The most notable market event affecting α-blockers occurred when generic doxazosin became available in October 2000. Prazosin had been available as a generic in September 1988 and terazosin in March 1998.

We also considered patterns of α-blocker use for the treatment of indications other than hypertension, particularly benign prostatic hyperplasia. In examining the use of tamsulosin in the treatment of benign prostatic hyperplasia, we found that increases in tamsulosin use predated the decline in traditional α-blocker use. Furthermore, the use of traditional α-blockers for genitourinary disorders increased slightly from 40% in 1996 to 42% in 2002, suggesting that any decline in α-blocker use could not be attributed to direct substitution by tamsulosin. We have not presented further details of this analysis.

Finally, we also considered broader trends in antihypertensive medication prescribing to gauge whether α-blockers may have been subject to increasing competition from other drugs in 2000. We present information on changes in the use of other drug classes, as well as individual medications.

The Cox and Stuart test for trend was used to test for significance in changes observed for α-blocker prescribing patterns over time.²⁸ We analyzed the temporal sequence of NPA prescribing and dispensing data and applied a 1-tailed test. The null hypothesis states that there was no significant downward trend in α-blocker use. We used Microsoft Office Excel (Redmond, Wash) software to statistically analyze the IMS Health data reported herein. Results were evaluated with P = .05 as the threshold for statistical significance. Estimated sampling errors also were available for both NPA²⁹ and NDTI data,³⁰ and we used them to derive 95% confidence intervals (CIs) for our reported estimates.

Before the publication of the ALLHAT doxazosin results, overall α-blocker use slowly increased according to each available data source. As described herein, between 1996 and 1999 the observed increase ranged from 5% to 15%, depending on the metric used to assess prescribing patterns. Following the ALLHAT results, α-blocker use declined by as much as 54% from 1999 through 2002. In addition to the impact of the early ALLHAT results, we also considered other potential influences, including drug promotional efforts, drug prices, generic conversion of α-blockers, and market competition. With the possible exception of diminished promotional efforts, these other factors do not appear to have contributed significantly to the observed decline in α-blocker use.

US retail pharmacies dispensed 3.53 million (95% CI, 3.52 million to 3.54 million) α-blocker prescriptions to consumers in the first quarter of 1996; the total dispensing volume increased consistently through 1999 to peak at 4.44 million (95% CI, 4.43 million to 4.45 million) in the fourth quarter of 1999 (Figure 1). Following early ALLHAT termination, this increasing trend reversed beginning in the first quarter of 2000 (4.21 million; 95% CI, 4.20 million to 4.22 million) and declined continuously thereafter to 3.46 million (95% CI, 3.45 million to 3.47 million) in the first quarter of 2002 and to 3.26 million (95% CI, 3.25 million to 3.27 million) by the fourth quarter of 2002. The trends observed for total dispensed prescriptions include both newly dispensed and refills for α-blockers. The downward trend recorded for total dispensed α-blockers was statistically significant (P = .006) using the Cox and Stuart test for trend. The quantity of newly dispensed α-blocker prescriptions increased by 16% from 1.16 million (95% CI, 1.15 million to 1.17 million) in the first quarter of 1996 to 1.35 million (95% CI, 1.34 million to 1.36 million) in the fourth quarter of 1999. By the second quarter of 2000, it declined by 8% to 1.23 million (95% CI, 1.22 million to 1.24 million). Newly dispensed prescriptions declined further to 0.91 million (95% CI, 0.90 million to 0.92 million) in the fourth quarter of 2002, marking a 33% decline from 1999 levels. The decline in newly dispensed α-blockers was also measured to be statistically significant (P = .006). Dispensed refill values likewise show a statistically significant decline in the α-blocker trend line (P = .03). Following a similar trend, the number of dispensed refill α-blockers increased by 30% from 2.37 million (95% CI, 2.36 million to 2.38 million) in the first quarter of 1996 to 3.09 million (95% CI, 3.08 million to 3.10 million) in the fourth quarter of 1999. Values decreased by 24% from peak 1999 levels to 2.35 million (95% CI, 2.34 million to 2.36 million) in the fourth quarter of 2002. On an annual basis, dispensing grew from 15.0 million in 1996 to 17.2 million in 1999 before declining by 22% to 13.4 million in 2002. Note that the volume of total dispensed prescriptions exceeds the number of prescriptions received by pharmacies because of dispensing that occurs as previously received prescriptions are refilled.

The dispensing of generic and brand name prazosin remained relatively infrequent and in steady decline from 1996 through 2002. Despite an early increase in 1996, brand name terazosin declined continuously throughout the entire study span, a 15% decrease between the fourth quarter of 1996 and the second quarter of 1999 and an overall 85% decrease by the fourth quarter of 2002. Dispensed brand name doxazosin increased from 1.1 million (95% CI, 1.09 million to 1.11 million) in the first quarter of 1996 to 2.28 million (95% CI, 2.27 million to 2.29 million) in the fourth quarter of 1999, slightly declined from the first quarter of 2000 (2.21 million; 95% CI, 2.20 million to 2.22 million) to the third quarter of 2000 (2.17 million; 95% CI, 2.16 million to 2.18 million) but then decreased significantly to 0.67 million (95% CI, 0.66 million to 0.68 million) in the first quarter of 2001 and to 0.16 million (95% CI, 0.15 million to 0.17 million) in the fourth quarter of 2002.

The estimated number of α-blocker new prescription orders received by pharmacies increased by 16% from 1.16 million (95% CI, 1.11 million to 1.21 million) prescriptions in the first quarter of 1996 to a peak of 1.35 million (95% CI, 1.31 million to 1.39 million) in the fourth quarter of 1999 (Figure 2). α-Blocker prescriptions then decreased to 1.15 million (95% CI, 1.11 million to 1.19 million) by the third quarter of 2000, with a further decline to 0.91 million (95% CI, 0.87 million to 0.95 million) by the end of 2002, reflecting a 33% decline from peak 1999 levels. On an annual basis, new α-blocker prescriptions increased from 4.90 million in 1996 to 5.15 million in 1999 before declining by 26% to 3.79 million in 2002. Statistical analysis using the Cox and Stuart test for trend showed a significant downward trend for newly prescribed α-blockers after the ALLHAT results (P = .006). The decline in total α-blocker use resulted mostly from overall decreased prescriptions of both brand name and generic doxazosin from 0.70 million (95% CI, 0.67 million to 0.73 million) in the first quarter of 2000 to 0.49 million (95% CI, 0.46 million to 0.52 million) in the fourth quarter of 2002. There was a transient increase in total doxazosin prescriptions between the third quarter of 2000 (0.62 million; 95% CI, 0.59 million to 0.65 million) and the fourth quarter of 2000 (0.76 million; 95% CI, 0.73 million to 0.79 million) that accompanied the availability of generic doxazosin (Figure 2).

Transition from brand name to generic doxazosin occurred in October 2000, only 6 months after the publication of the ALLHAT results. As noted, brand name doxazosin dispensing decreased by 86% from 2.17 million (95% CI, 2.16 million to 2.18 million) in the third quarter of 2000 to 0.30 million (95% CI, 0.30 million to 0.30 million) in the fourth quarter of 2001 following the availability of generic doxazosin (Figure 3). The availability of less expensive generic doxazosin (priced at $29 per prescription in the fourth quarter of 2001 vs $41 for brand name doxazosin), however, offset much of this decline. Generic doxazosin dispensed was 1.71 million in the fourth quarter of 2001 so that the net reduction in total doxazosin dispensing (generic plus brand name) was consistent with trends established early in 2000 before doxazosin's generic conversion.

Annual α-blocker drug use at visits to office-based physicians in the NDTI increased by 9% from 2.08 million (95% CI, 1.70 million to 2.46 million) in 1996 to 2.26 million (95% CI, 1.89 million to 2.63 million) in 1999. During this time, the proportion of drug use associated with the treatment of hypertension and related cardiovascular disease remained stable at 58%. As a result, α-blocker use for hypertension treatment increased from 1.20 million (95% CI, 0.93 million to 2.47 million) in 1996 to 1.28 million (95% CI, 1.02 million to 1.54 million) in 1999. Following the ALLHAT doxazosin results, quarterly α-blocker drug use declined from 0.52 million (95% CI, 0.37 million to 0.67 million) in the fourth quarter of 1999 to 0.42 million (95% CI, 0.29 million to 0.55 million) in the fourth quarter of 2000, then gradually declined further to 0.22 million (95% CI, 0.13 million to 0.31 million) by the fourth quarter of 2002 (Figure 4). Overall, annual α-blocker use declined by 54% between 1999 (2.26 million; 95% CI, 1.89 million to 2.63 million) and 2002 (1.03 million; 95% CI, 0.79 million to 1.27 million). α-Blocker use for hypertension decreased by 66% from 0.32 million (95% CI, 0.21 million to 0.43 million) in the fourth quarter of 1999 to 0.11 million (95% CI, 0.05 million to 0.17 million) in the fourth quarter of 2002.

Estimated expenditures for the office promotion of α-blockers increased by 47% from $6.3 million in the first quarter of 1996 to a peak of $9.3 million in the third quarter of 1998 (Figure 5). Following this peak, there was a slow decline in promotional expenditures to $7.3 million in the third quarter of 1999. Office promotion expenditure then declined more rapidly to $5.5 million in the first quarter of 2000 and $4.0 million in the third quarter of 2000. By the first quarter of 2001, no α-blockers were being promoted on the market. Office promotion expenditures for brand name doxazosin initially increased from $3.0 million in the first quarter of 1996 to a peak of $5.5 million in the second quarter of 2000 followed by a rapid decline to no expenditures by the first quarter of 2001.

Trends in drug prices were evaluated to gauge whether observed trends in α-blocker use were consistent with price-driven market responses. The mean price per α-blocker prescription increased from $41 in 1996 to $45 in 2000 before decreasing to $40 in 2002. This pattern reflects 2 countervailing trends in the consistently increasing prices of brand name α-blockers and in the availability of lower-priced generic forms of terazosin in March 1998 and doxazosin in October 2000. Without these generic conversions, overall drug prices would have continued to increase in 2001 and 2002. For example, the price per prescription of brand name terazosin increased 77% from $48 in the first quarter of 1996 to $85 in the fourth quarter of 2002, whereas brand name doxazosin increased 29% from $34 in the first quarter of 1996 to $44 in the fourth quarter of 2002. These rates of increase in the price of brand name products were substantially in excess of the increase in the overall producer price index for pharmaceuticals, the medical consumer price index, or overall inflation in the US economy. Lower prices were associated with generic terazosin ($52 in 2002) and generic doxazosin ($28 in 2002). The ratio of brand name to generic prices therefore increased over time for both terazosin (1.15 in 1999 to 1.60 in 2002) and doxazosin (1.23 in 2000 to 1.54 in 2002) as a result of increasing brand name prices and small decreases in generic prices.

We examined whether the decline in α-blocker use corresponded to the growth in one or more drug classes or individual drugs. This analysis suggested that angiotensin receptor blockers (ARBs) might have partially replaced α-blockers even though ARBs experienced substantial growth before the decline in α-blocker use. Between 1996 and 2002, the use of ARBs rose steadily from 3% of patients with hypertension in 1996 to 18% by 2002, with the largest increases between 1999 (8.5%) and 2000 (14.0%). Among ARBs, valsartan use gradually increased from 0% in 1996 to 7% in 2002, whereas losartan potassium use increased from 3% in 1996 to 6% in 2002. The use of ACE inhibitors gradually increased from 33% among patients with hypertension in 1996 to 40% in 2002, without any differences in this pattern before and after the first quarter of 2000. Lisinopril, the ACE inhibitor studied in ALLHAT, did not increase substantially between 1996 (12%) and 2002 (13%). The use of benazepril increased from 4% in 1996 to 8% in 2002 (including its use as a combination product with amlodipine). Overall, calcium channel blocker use declined from 34% in 1996 to 26% in 2002. However, amlodipine, the calcium channel blocker studied in ALLHAT, increased from 9% in 1996 to 17% in 2002, with the greatest increase between 2000 (12%) and 2001 (16%). Diuretic use declined slightly from 22% of patients in 1996 to 19% in 2002. Although use of chlorthalidone, the diuretic studied in ALLHAT, remained rare (<1% of patients), hydrochlorothiazide use increased slightly from 6% in 1996 to 8% in 2002. β-Blocker use remained essentially stable at 19.0% among patients with hypertension between 1996 and 2002, as did the use of the leading β-blocker atenolol (9% in 2002). Note that more than 1 antihypertensive medication may be reported per patient because of the use of combination products and multiple antihypertensive products.

Our study describes α-blocker prescribing patterns before and after the publication of early ALLHAT results in April 2000 and examines other possible simultaneous influences on physician prescribing patterns for α-blockers during this period. The doxazosin arm of the ALLHAT trial was discontinued in January 2000, with the subsequent release of clinical recommendations against the use of α-blockers as first-line agents for hypertension treatment.

Our description of patterns in α-blocker use immediately before and after the publication of the early ALLHAT results suggests a reversal of preexisting trends. Before the ALLHAT early termination results were released, α-blocker prescribing and dispensing increased steadily between the first quarter of 1996 through the fourth quarter of 1999. Beginning in the second quarter of 2000, synchronous with the ALLHAT results, overall α-blocker drug use began to decline. We noted consistent patterns for newly written prescriptions, prescriptions dispensed by pharmacies, and drug use patterns reported by office physicians. The most significant declines were noted in drug use patterns, perhaps suggesting that visits to physicians provided an opportunity to discontinue the use of α-blockers that had been previously prescribed.

We examined the declining trends in α-blocker use in the context of other influential factors, including generic conversion of doxazosin, drug promotion, drug prices, and other market factors. Among these factors, significant reduction in promotion coincides with decline in α-blocker use although the most significant declines in promotion occurred after α-blocker use had begun to decline. The complexity and variety of these potential influences, however, make definitive inference difficult.

Beyond the ALLHAT early termination results, generic conversion of brand name doxazosin may have temporarily delayed the decline in α-blocker use. An increase in α-blocker prescriptions was noted between the fourth quarter of 2000 and the first quarter of 2001, marking an increase in α-blocker use, according to this metric, in the otherwise decreasing trend line. The decrease in total α-blocker prescriptions following early ALLHAT results preceded any significant decline in doxazosin promotion expenditures. Brand name doxazosin promotion expenditures decreased in the fourth quarter of 2000. This pattern is consistent with past economic studies³¹ that link declines in promotional expenditures to impending loss of patent status. The sharp decline in doxazosin promotion may have further exacerbated the decrease in α-blocker use observed from mid 2000 to 2002. The ratio of prices for brand name doxazosin to generic increased over time after the initial availability of generic doxazosin. This pattern of price change is consistent with that described in economic studies of generic entry and pricing.³² The α-blocker prescription rates also declined despite the availability of less expensive generic versions between 2000 and 2002. In contrast, the average price per prescription for all antihypertensive medications increased from $43.85 in 1993 to $44.83 in 2001. To the extent that prescribing is sensitive to such comparative prices, the reduced cost of selected generic α-blockers could have offset what might have been an even faster decline in their use.^33,34

Increased market competition from other antihypertensive drugs, particularly ARBs, also may have influenced the prescribing of α-blockers. Angiotensin receptor blockers, as well as ACE inhibitors and the calcium channel blocker amlodipine, gained market share contemporaneously with the decline in α-blocker use. As a relatively novel addition to the antihypertensive market, ARBs may have been a leading candidate for competitive substitution. Despite the inverse association between ARB use and α-blocker use, however, our data does not allow assessment of whether patients were switched from one drug to another.

Past literature suggests that physician response to clinical recommendations is more rapid when clinical trials reveal negative rather than positive results.³⁵ Although our findings are consistent with past studies, physician response was neither as marked nor as rapid as might have been expected.

Several limitations of our analysis should be highlighted. The NDTI data are based on a sample of physicians who may not necessarily be representative of all physicians. These physicians report their best knowledge of patients' medications during a sample of visits. Although a large sample of retail pharmacies is available in NPA, we did not analyze information on drug dispensing from other sources. These limitations are mitigated by the cross-validation provided by using both NDTI and NPA data. The comparable trends we observed across these 2 databases argue for the validity of the prescribing patterns recorded in this study. This study builds on previous research on individual factors that affect physician prescribing decisions. There is an absence of research focusing on multiple, simultaneous influences on national prescribing patterns. Despite our consideration of multiple factors in this study, we acknowledge that all conceivable factors potentially affecting drug prescribing have not been included. In addition, all of the complex interactions between these factors have not been explicitly considered in this descriptive analysis.

Because there are multiple simultaneous influences, it is difficult to establish a primary influencing factor on the significant decline in physician prescribing of α-blockers. Nevertheless, our findings are clearly consistent with ALLHAT early termination results having a significant impact on α-blocker use. Declining pharmaceutical industry promotion also may have contributed further to decreased α-blocker use. The lack of an abrupt and more pronounced decline in prescribing shortly after the ALLHAT results, however, suggests slow and potentially incomplete diffusion of information from this clinical trial. Combined with past work that questions the potency of clinical trial results on physician practice, our analysis suggests the need for additional strategies to augment the dissemination of results from clinical trials.