The Complex World of Prescribing Behavior

Better decision making by physicians could materially improve the balance of benefits and harms in health care while saving billions of dollars. It is therefore little wonder that academics, policy makers, third-party payers, and leaders of the profession alike have been grappling for many years with the challenge of modifying physician behaviors.

Conventional wisdom on that issue has arguably evolved through 4 phases.¹ The Era of Optimism featured a belief that physicians could be transformed into critical appraisal machines, tirelessly combing the peer-reviewed literature and consistently translating the best evidence about drugs and devices into action. The Era of Innocence Lost and Regained saw a loss of faith in passive diffusion of evidence and its distillation by individual clinicians. Instead, the medical establishment fervently embraced active dissemination and collective synthesis of evidence in the form of meta-analyses, decision analyses, and practice guidelines. The Era of Industrialization followed once research studies showed that practice guidelines were not consistently guiding practice. Physician-leaders and health care administrators borrowed the ideas of industrial quality gurus; local implementation, under an alphabet soup of rubrics, including CQI (continuous quality improvement), TQM (total quality management), and Six Sigma, was the rage. The most recent phase seems to be the Era of Information Technology and Systems Engineering. Using concepts from all 3 earlier phases, today's sociomedical engineers attack systematic barriers to change, align economic and noneconomic incentives, and deploy information tools to steer clinicians and all other involved decision makers, including patients.

If nothing else, physicians have long since learned that research evidence is not the only factor that affects clinical decision making, including prescriptions of new drugs. But what happens to prescribing when evidence emerges to cast doubt on the safety of older and well-known drugs? Two reports^{2 - 3} in this issue of THE JOURNAL address that question.

The study by Hersh et al² shows dramatic declines in US use of oral hormone therapy following termination and publication of the estrogen-progestin arm of the Women's Health Initiative (WHI) trial in July 2002.⁴ Drawing on national data sources, the study estimated a 56% decline in oral estrogen-progestin prescriptions within a year of the WHI publication, with smaller but still striking declines in use of estrogens. Similar declines have already been reported from Canada.⁵

The other study in this issue by Stafford et al³ assesses the use of α-blockers in the United States. An interim analysis of the ALLHAT study (Antihypertensive and Lipid-Lowering treatment to prevent Heart Attack Trial) in 2000 showed that doxazosin was significantly inferior to chlorthalidone with respect to preventing stroke, congestive heart failure, and a composite of several cardiovascular outcomes.⁶ Prescriptions of α-blockers decreased by about 30% over 2 years—a smaller and more gradual decline than observed for hormone therapy.

The situation for these 2 drugs is similar in some respects. In both cases, harmful outcomes were demonstrated in major randomized trials, although media coverage was much more intense for the WHI findings than for the interim results of ALLHAT. Hypertension, unlike menopause, is often asymptomatic, but neither intervention saves lives acutely.

On the other hand, the trajectory of the 2 interventions differed. Hormone therapy with unopposed estrogen was associated with an increased risk of endometrial cancer in the late 1970s. A sharp decrease in prescriptions ensued but was reversed when estrogen-progestin compounds were introduced. In 1997, a pooled analysis of multiple small trials by Hemminki and McPherson challenged the link between hormone therapy and prevention of cardiovascular disease,⁷ and the HERS results in 1998⁸ and again in 2002⁹ further undercut that rationale for hormone therapy. Meanwhile, evidence of other adverse effects associated with estrogen therapy, such as an increased incidence of both cholecystectomy and appendectomy,¹⁰ continued to accumulate. The study by Hersh et al² shows that hormone therapy use increased dramatically through the latter half of the 1990s but then hit a plateau, presumably as these unsettling findings emerged and clinicians awaited the WHI results.

In contrast, α-blockers were one option among many for controlling hypertension, and by 1993 these agents had already been given second-line status in the report of the Fifth Joint National Committee on the Detection, Evaluation, and Treatment of High Blood Pressure (JNC V).¹¹ Different authors place different interpretations on the strength of the JNC V and subsequent JNC VI¹² endorsements of diuretics and β-blockers as first-line antihypertensive therapy. Indeed, physiological inferences and trial evidence (CAPPP [Captopril Prevention Project],¹³ STOP-2 [Swedish Trial in Old Patients with Hypertension-2],¹⁴ and UKPDS [UK Prospective Diabetes Study]¹⁵ ) supported a contrarian view in the hypertension community that attaining a target blood pressure with tolerable adverse effects was ultimately more important than choice of agent. α-Blockers were astutely marketed as preferable for patients with diabetes, and their salutary effects on prostatism offered an attractive secondary effect in treating older men. These factors may help explain why Stafford et al³ found that α-blocker use since 1996 was increasing slowly, not decreasing as might be expected. However, the absolute volume of α-blocker prescriptions for hypertension remained modest, particularly compared with the dominant antihypertensives (or hormone therapy) in the same period. In stock market parlance, α-blockers may already have been discounted. In addition, as astutely dissected by Stafford et al, the switch to generic status for doxazosin meant that the drug became more affordable in the same period that α-blocker use might have been expected to decrease further.

Each of these 2 studies of "bad news" about old and established drugs raises fascinating issues. Given the evidence from WHI, imagine a physician caring for a woman who has been taking hormone therapy and develops breast cancer or experiences a cardiovascular event. The population-attributable risk is small, making it unlikely that the condition was actually caused by hormone therapy, but a causal link could never be completely ruled out. That possibility is deeply unsettling from an ethical and psychological standpoint and a source of medicolegal hazard that will doubtless be played out many times in the years ahead.

In this context, a pointed question is already being asked. Were health care opinion leaders and the pharmaceutical industry locked into an alliance that accelerated the medicalization of menopause? Hormone therapy was indeed a heavily promoted and profitable product line. This is wonderful terrain for neo-Marxists and feminist deconstructionists, but the problem may be even deeper-seated. Perhaps physicians today are a little less likely to prescribe benzodiazepines to transiently anxious or distressed patients, but the therapeutic culture is still omnipresent. Loss of interest in a sexual partner is treated with sildenafil or one of its congeners. Psychotic-level and debilitating depression goes underdiagnosed and undertreated while antidepressants are prescribed to persons with transient dysphoria. Everyone must "know their number" and consider lipid-lowering drugs. Moreover, just as the incredibly wide uptake of the WHI results by the media turned patients into "evidence vectors," direct-to-consumer advertising gives a whole new meaning to the concept of mass medicalization. The ever-expanding role for therapy, particularly prescription drugs, is generally supported by at least some lines of evidence and embraced by millions of professionals and patients eager to avoid death, disease, dysfunction, disability, and distress. The current therapeutic culture, in short, has its roots in the religion of modernity: faith in science and redemption through technology.

Considering the findings of Stafford et al³ for antihypertensive drugs, what is perhaps most interesting is not the moderate decline in use of α-blockers but rather the limited and declining use of thiazides. Now that the final results of ALLHAT have been published,¹⁶ it can be argued vigorously that low to moderate doses of thiazide diuretics are both more effective and less costly than calcium channel blockers such as amlodipine or angiotensin-converting enzyme inhibitors such as lisinopril. It will be fascinating and informative to study the trajectory of thiazide prescriptions in industrialized nations since the dissemination of the ALLHAT results. In direct contrast to harms associated with α-blockers and hormone therapy, this evidence should breathe new life into an old drug. In a Research Letter in this issue of THE JOURNAL, Austin et al¹⁷ show that after publication of ALLHAT results, there was a sharp increase in the use of thiazides and a dropoff in the use of angiotensin-converting inhibitors and angiotensin receptor blockers(ARBs) enzyme for elderly patients in Ontario. The authors note that their findings suggest that many physicians may indeed respond to "research evidence irrespective of the cost or brand-name status of the drug."¹⁷ But while these preliminary results are encouraging, well-studied compounds that have long since become available in generic forms are pharmaceutical orphans. If industry invests in promoting them for new indications, they are likely to do more for the bottom line of their generic rivals than their own corporate balance-sheets.

Another of these middle-aged orphans in the pharmaceutical armamentarium is spironolactone. In part because of the prevalence of gynecomastia as an adverse effect, but also owing to the usual life cycle of patented prescription drugs, spironolactone essentially had fallen off the therapeutic radar screen during the 1990s. The RALES trial (Randomized Aldactone Evaluation Study) results in 1999 showed surprising benefits from adding this old drug to modern heart failure regimens.¹⁸ Majumder et al¹⁹ examined prescriptions of spironolactone in Canada and the United States before and after publication of the RALES results. Despite lack of investment in drug promotion, use of the compound increased by about 2% per month during the year following publication. But this growth paled by comparison with the increase in use of heavily promoted ramipril following the HOPE study (Heart Outcomes Prevention Evaluation).²⁰

The ALLHAT investigators are trying to facilitate the translation of their findings into practice, but no team of academic clinicians can hope to match the marketing know-how and deep pockets of the pharmaceutical industry. What is needed are stronger mechanisms to avoid premature retirement of aging orphan drugs that are still useful or may have new indications. A particular shortfall arises in research where, once compounds move off-patent, industry is no longer likely to support studies that assess their comparative effectiveness or test new indications. A variation on this problem arises with vitamins or common nutraceuticals for which patent protection is absent, with the result that investment in marketing takes precedence over scientific studies of efficacy.

Set against the life cycles of older compounds, prescribing trends for new compounds may seem relatively straightforward. It is still difficult to understand, however, why some new drugs catch on so quickly, while others do not. The interactions of old and new may also have their own peculiar logic. For example, ARBs are the fastest growing class of antihypertensive agents in Canada and the United States.²¹ Angiotensin receptor blockers are new, physiologically attractive in their specificity, apparently have a good safety profile, and are supported by strong marketing campaigns. The LIFE trial (Losartan Intervention For Endpoint reduction in hypertension) suggests that ARBs have outcome advantages over β-blockers²² ; but in LIFE, a substantial fraction of patients also received thiazides. Given the LIFE and ALLHAT results, will physicians infer that diuretics and ARBs are the new first-choice drugs for hypertension?

The answer is difficult to predict. Indeed, if patterns of human behavior were simple to understand, change, and predict, the world would be a very different and much less interesting place. In the complex realm of prescribing behaviors, physicians now process ever-increasing amounts of evolving evidence in contexts with their own unique micro- and macro-sociology, where the economic incentives are not infrequently perverse. Prescribers must individualize therapy and consider combinations of drugs for which randomized evidence may be limited. The pharmaceutical industry markets its wares relentlessly to physicians and consumers alike and exerts its own steering effects on basic and clinical research. Third-party payers are pressing for drug cost containment, patients' expectations have increased rapidly, and medicolegal pitfalls are omnipresent. Moreover, pharmacogenomics heralds a phase in which prescribing may be more elegantly tailored to the patient's biology but more dependent on technological support than ever before. At this critical juncture in the evolution of pharmaceuticals and health care, the time has come for much more systematic study of the myriad factors that affect what physicians prescribe for their patients.