Physicians and health officials have long noted nonadherence to treatment among patients with such diseases as tuberculosis, diabetes, and hypertension. They have called patients who are unable or unwilling to follow prescribed regimens ignorant, vicious, recalcitrant, and noncompliant. Meanwhile, researchers have regularly tried to correlate nonadherence with specific patient characteristics.

We argue that history provides two major lessons for clinicians struggling with triple-drug therapy and nonadherence. First, labels such as nonadherent overemphasize the practitioner's role in determining who should receive treatment and how it should be administered. Second, the ability to predict nonadherence in individual patients remains elusive. Given these findings, we advocate revisiting the longstanding tendency to reflexively categorize patients as either adherent or nonadherent. Although it remains essential to know who is taking triple-drug therapy properly, an approach that discourages labeling and emphasizes active patient participation in therapy will avoid judgmental and inaccurate assessments of patient behavior and may help HIV-positive patients take medications more successfully.

A Historical Perspective

Health officials first became concerned about uncooperative patients with tuberculosis in the 1890s, shortly after Robert Koch proved that the disease (also known as consumption) was communicable. Officials implemented numerous policies to prevent the spread of tuberculosis, including those about coughing into handkerchiefs, proper sputum disposal, and isolation of actively infected persons [10]. Although persons from all social classes violated these directives, health departments specifically targeted single, alcoholic men from rundown "skid row" areas [11].

These men received little tolerance when they disobeyed regulations. In 1903, New York tuberculosis control pioneer Hermann Biggs opened a facility, Riverside Hospital, to isolate such men forcibly. Riverside, he wrote, was designed for "[h]omeless, friendless, dependent, dissipated and vicious consumptives ... [who] are likely to be most dangerous to the community" [12]. Such derogatory language was common. A contemporary of Biggs advocated "detention institutes for ignorant and vicious consumptives" [13]. The label of nonadherence in the early 1900s thus reflected both the actual conduct of patients and judgmental assumptions about the behavior of certain populations [14].

The introduction of curative antibiotics for tuberculosis after 1945 increased concerns about nonadherence. Streptomycin and isoniazid revolutionized tuberculosis therapy, but many patients did not take their antibiotics after they were discharged from sanatoriums. Not only did these patients spread infection, but their erratic compliance produced drug resistance. Again, attention was focused on the most obviously nonadherent patients: skid row alcoholics, who were often mentally ill [15, 16].

Health officials of this era carefully eschewed the derisive language of earlier times, calling such alcoholics recalcitrant [17-19]. Yet the term recalcitrant, defined as "stubbornly resistant to authority" [20], implied that physicians imposed regimens that patients were to obey. Not surprisingly, writers associated recalcitrance with "the alcoholic, drug addict, skid-row derelict and others with personality defects" [21-23], blaming the victim for complex behaviors and conditions that were often unmanageable [24]. Assuming that such persons were inevitably nonadherent, tuberculosis officials often confined them inappropriately; nonadherent middle-class patients were rarely detained [15].

The Era of Noncompliance

The introduction of multiple new antibiotics in the 1950s led to studies of nonadherence in patients with infections other than tuberculosis. For example, it was found that adults and children with streptococcal pharyngitis rarely completed 10-day courses of penicillin [25]. In one study [26], only 19% of patients took four pills daily for 10 days. Moreover, nonadherence to penicillin occurred both in public hospitals and in private practice [27]. This finding-that nonadherence was prevalent throughout the population-contradicted earlier assumptions that only poor patients or alcoholics disregarded medical recommendations.

Additional data were generated in the 1970s when Sackett and Haynes and their colleagues published books on compliance and noncompliance [28, 29]. Earlier commentators had used the term noncompliant and had studied the complex relation between patients' beliefs about health and their ingestion of medication [30-32], but these authors popularized the concept of noncompliance. Even though Sackett and Haynes tried to make noncompliance a nonjudgmental term, recent authors have criticized it. As a result, the terms adherent and nonadherent-which seek to emphasize patient participation in decision making-are now preferred [33].

The work of Sackett and Haynes and their colleagues spawned thousands of studies that examined the causes of nonadherence and proposed solutions for it. Suggested corrective strategies include educating patients, modifying patient behaviors, and altering the organizational setting of the clinician-patient encounter. Specific interventions include simplified drug regimens; pill-dispensing systems; user-friendly clinics; and, for tuberculosis, directly observed therapy [34-40].

The literature on nonadherence has generated many important findings. At least 40% to 50% of patients do not adhere to either short-term antibiotic regimens or long-term therapies for asthma, diabetes, or hypertension [28, 41, 42]. Nonadherence is not an all-or-none phenomenon but is rather a dynamic set of behaviors. For example, patients may adhere to treatment with some medications but not others. Even supposedly adherent patients are usually only partially adherent [43, 44]. Indeed, some studies suggest that 80% adherence yields good clinical outcomes and is a realistic goal [45]. The degree of adherence needed for triple-drug therapy is still unknown.

Research has consistently failed to identify demographic variables that predict whether patients will follow the advice of physicians. In other words, nonadherence occurs regardless of class, ethnicity, or extent of education [9, 28, 41, 42]. Certain patient characteristics-such as homelessness, mental illness, substance abuse, and previous nonadherence-may help to predict future nonadherence [46-48]. Yet none of these factors, with the possible exception of psychiatric illness [8, 42], is reliably associated with nonadherence across studies [36]. Most important, predictions about specific risk groups do not apply to all persons in such groups. For example, in one recent study that found an association between injection drug use and nonadherence to tuberculosis therapy [48], 40% of all drug users completed treatment-without directly observed therapy. Thus, it is difficult to predict nonadherence in individual patients at a rate better than that achieved with chance alone [9, 36, 41, 49-51].

Nonadherence and HIV Disease

Although triple-drug therapy has highlighted the subject of nonadherence, this issue is not new to HIV disease [9]. For example, nonadherence to zidovudine monotherapy has been well described. As is true with other diseases, demographic factors have not predicted the individual behavior of HIV-infected patients. Some researchers have concluded that certain patient characteristics, such as injection drug use, are associated with greater nonadherence [7, 52-54], but others have found that HIV-positive drug users are reliable if given social supports [55-57].

The triple-drug HIV cocktail greatly magnifies earlier concerns about nonadherence. When all of their medications are included, HIV-positive patients receiving therapy with zidovudine, lamivudine, and a protease inhibitor may take 20 or more pills each day. These drugs, moreover, must be taken two or three times daily and at strict intervals. Some must be taken while patients are fasting; others must be taken with a high-fat, high-protein meal. Because the pills may cause side effects, frequent modifications of therapy may be necessary [58, 59].

Triple-drug therapy also raises concerns about drug resistance, both for individual persons and for the community. In the case of tuberculosis, erratic adherence led to the emergence of multidrug-resistant strains [46, 60]. Early data suggest that patients taking the triple-drug cocktail sporadically or at a reduced dose will probably develop drug resistance [61-66]. Moreover, resistance to one drug may produce cross-resistance within the same class, jeopardizing future treatment options for individual patients. Because multidrug-resistant strains may be transmissible, the public health consequences of nonadherence are potentially alarming [9].

Physicians prescribing triple-drug therapy have drawn on well-established adherence strategies, such as educational support and enhanced social services. They are also using newer approaches, including financial incentives, outreach workers, pharmacist counselors, and beeper and telephone reminders [35, 67-70]. Modified programs of directly observed therapy have also been suggested [4]. Meanwhile, researchers are developing new drugs and simpler regimens [71]. Better pills are not a panacea for nonadherence, but they will almost certainly enhance the ability of patients to take medications regularly.

The approach to nonadherence has thus become considerably more sophisticated. Yet debates over the triple-drug cocktail have raised old issues. As reported in The New York Times in 1997 [2], certain clinicians, justifiably concerned about drug resistance, are withholding triple-drug therapy from selected patients. As their predecessors did, practitioners are assessing patients' lifestyles and behaviors when deciding who should receive therapy. And, as in the past, discussions about nonadherence have focused on disadvantaged populations with complicated social problems. In turn, critics have cautioned against denying the powerful new treatments to the poor, the homeless, and substance users. How can history illuminate and help resolve this debate?

Lessons from the Past

For decades, clinicians and public health officials have used labels to describe patients who disregard medical advice. To be sure, nonadherent is less judgmental than ignorant, vicious, recalcitrant, or even noncompliant. Nevertheless, by suggesting that patients who miss doses of medication are violating physicians' "orders" [72, 73], such terms potentially stigmatize persons who cannot adhere. Moreover, given that half of all persons regularly miss medication doses and that even "adherent" persons miss some doses, such labels are inexact.

History also cautions against over-reliance on weakly predictive models that compare nonadherence rates among various populations. Such models are appealing because they reflect concern for the disadvantaged and promote targeted interventions for particular groups, such as the homeless and injection drug users [53, 74-76]. Indeed, we believe that continuing efforts to help patients complete treatment must pay close attention to patients' sociocultural circumstances [77]. Yet we must also acknowledge the ramifications of identifying characteristics and behaviors assumed to be associated with nonadherence. This practice has often led clinicians to designate specific populations as necessarily uncooperative while ignoring nonadherence by other persons [78]. In addition, even when studies have shown nonadherence among certain groups, the prediction of future adherence by individual patients has "been found wanting in almost every study in which it has been tested" [42].

Rather than trying to predict, identify, and correct nonadherent behavior in certain patients, physicians might instead focus on helping all patients take their medications more successfully. Such a strategy would deemphasize potentially stigmatizing labels, such as nonadherent, in favor of language that stresses the patient's role in devising and implementing individualized treatment plans [79]. The physician-patient interaction would thus become more of a negotiation between informed patients and caregivers. Family and friends of the patient would be encouraged to participate. A small but growing body of literature suggests that such clinician-patient "alliances" may improve treatment outcomes [80-83].

It might be argued that empowering all patients is too much of a risk with triple-drug therapy. Efforts to enhance the autonomy of patients with noninfectious diseases, such as hypertension or diabetes, present few drawbacks, but the possibility of producing a cohort of partially adherent HIV-positive persons is undesirable. Such persons would increase the potential for drug resistance both in themselves and in the community.

We agree that partial adherence must be avoided. Yet the model that we propose does not promote a system in which clinicians, deprived of authority, simply prescribe triple-drug therapy for all willing patients. The goal of treatment plans is not to cede decision making to patients but rather to promote nonjudgmental dialogue and negotiation between patients and physicians. Physicians will still need to retain "veto power" for patients in whom the HIV cocktail is not medically appropriate [84] or for situations in which patients cannot or will not create a feasible, testable plan [5]. In addition, treatment should be immediately reevaluated if the results of monitoring strategies-such as viral load testing [85, 86]-suggest that patients may be missing too many doses. However, persons who are denied therapy and persons whose therapy is discontinued should be encouraged to create new treatment plans as their lives evolve and as simpler, better-tolerated therapies are developed.

A comparison of HIV disease with tuberculosis, another disease with public health ramifications, is instructive. For years, directly observed therapy was reserved for "manifestly unreliable" patients and was seen as a punishment [87]. More recently, health departments have reframed directly observed therapy as a service available to all tuberculous persons. As a result, directly observed therapy has become a way to expedite rather than force completion of treatment [88, 89]. The use of treatment plans would reflect a similar philosophy.

But the tuberculosis model is only partially applicable to HIV disease. Because of the greater communicability of tuberculosis, health departments have always been able to compel adherence. For HIV and AIDS, however, patient-centered approaches emphasizing autonomy and voluntary cooperation have always dominated clinical decision making, even though public health issues have been at stake [90]. Triple-drug therapy should be no exception to this rule.

Devising Treatment Plans

Patients and clinicians devising treatment plans for triple-drug therapy should build on established interventions [8, 9, 34-4052, 67-71, 91]. First, patients should be given access to 1) culturally appropriate educational pamphlets, videos, and CD-ROMs describing how pills should be taken and the possible side effects of the pills [92]; 2) professionals who can help them identify necessary physical, psychological, and social supports, such as peer group advisors, transportation, and child care; and 3) technologies, including printed medication charts, beepers, and pill boxes, that may facilitate reliable drug ingestion. Second, patients should receive help in 1) identifying lifestyle characteristics, such as homelessness, irregular work schedules, or frequent travel, that may interfere with the treatment plan and may need to be changed before therapy is initiated; 2) arranging for housing, food stamps, or drug rehabilitation if such assistance is seen as a precondition to initiation of treatment; and 3) designing a "draft plan" that links regimens to established daily routines, such as watching a specific television program, brushing teeth, and checking the mail. Third, patients should be offered 1) an experimental "dry run" with dummy pills to test the viability of proposed treatment plans; 2) regular feedback on viral load and T-cell counts after therapy begins; and 3) tools with which to assess the treatment plan, such as medication diaries and electronic timing devices. Fourth, patients should be assured of continued access to these types of supports for as long as treatment is medically indicated. It should be stressed that provision of adequate funding for all HIV-positive persons who begin and continue to require triple-drug therapy is fundamental to this effort of facilitating treatment.

What we propose is not revolutionary. Many physicians already avoid such terms as noncompliant and involve HIV-positive patients in decision making [5]. Nevertheless, we believe that the manner in which clinicians approach the issue of treatment completion needs to be revisited. The laudable historical concern about the problems that certain disadvantaged persons may have with adherence has often perpetuated the notion that such persons are, by definition, unable or unwilling to take medications appropriately. Shifting the focus from judgments about nonadherence to the development of patient-centered treatment plans can facilitate the larger goal of providing powerful new antiviral agents to all willing patients who are likely to benefit from them. Although the ability of such plans to help patients continue triple-drug therapy requires empirical study, this new approach will at least discourage the historical tendency to reflexively label certain populations as nonadherent.

Dr. Gulick: Cornell University School of Medicine, Cornell Clinical Trials Unit, 119 West 24th Street, New York, NY 10011.

Ms. Dubler: Montefiore Medical Center, Albert Einstein College of Medicine, 111 East 210th Street, Bronx, NY 10467-2490.