The development of highly active antiretroviral therapy (HAART) has led to a dramatic reduction in the mortality among patients with HIV infection. 1–3 However, problems with patient adherence to this medical therapy have created unprecedented challenges. The central concern is that nonadherence can lead to suboptimal viral suppression and the development of a resistant virus, which may then spread among people at risk. This problem adds to the gravity of decisions about when to initiate HAART.
In this issue, Fairfield and colleagues describe the patterns of HAART use among 275 patients attending a general internal medicine practice at the Beth Israel Deaconess Medical Center. 4 They found that advanced disease (determined by CD4 count and viral load) predicted early initiation of HAART, while both injection drug use and depression predicted delay in initiating HAART. For injection drug users, the delay was approximately 9 months. Did these differences arise because physicians withheld therapy from people who were marginalized and may also have been depressed and injected drugs, or did they arise because physicians appropriately balanced the risk of nonadherence with the need for therapy?
Concern about withholding medical therapy from people who are marginalized is appropriate. Studies have shown that ethnic minorities had lower treatment rates for coronary artery disease and alcoholism and they less often received analgesia for long bone fractures, erythropoietin for renal disease, and rehabilitation after a mastectomy. 5–10 Also, African Americans with HIV infection were less likely to receive antiretroviral therapy and prophylaxis for Pneumocystis carinii pneumonia in the pre-HAART era. 11 More recently, in two studies only 14% to 17% of injection drug users were found to receive HAART, compared with 82% of other HIV-infected people. 1,12,13
Provider concern about patient nonadherence, however, may be a good reason for delaying HAART in some patients. Adherence is thought to be a central predictor of antiretroviral success because of the close relation between adherence and viral suppression. 14 Although there are fewer data about resistance, current thinking suggests that inadequate viral suppression due to nonadherence may select for a drug-resistant virus. 15,16 The development of drug resistance not only compromises the success of therapy in individual patients, but also presents a potential public health threat from the spread of multidrug-resistant HIV through unsafe sex or drug use. 17–21 Therefore, delaying HAART in people who inject drugs or who have depression may be appropriate—if these characteristics are associated with poor adherence and virologic failure, and if they are modifiable.
There is a consistent relation between depression and nonadherence to treatment for several diseases, including coronary artery disease, 22–24 multiple sclerosis, 25 asthma, 26 and end-stage renal disease. 27,28 Depression is also a consistent predictor of nonadherence to HIV therapy. 29–31 Delaying HAART until after depression is treated may be appropriate because therapy for depression is effective and safe and may lead to better adherence and thus better virologic and medical outcomes.
The relation between injection drug use and adherence is more controversial because injection drug use has been inconsistently associated with poor adherence. In New York City, Pablos-Mendez found that injection drug users were less likely to adhere to tuberculosis (TB) therapy, resulting in delayed conversion to negative cultures. 32 In that study, however, 35% of injection drug users were adherent to TB therapy. Caminero and colleagues also found injection drug use to be an independent predictor of nonadherence to directly observed therapy for TB. 33 However, others have found no differences between injection drug users and noninjection drug users regarding adherence to TB therapy. 34,35 The utility of injection drug use in predicting adherence to antiretroviral therapy is also controversial. Early studies in the pre-HAART era suggested that injection drug users were less likely to adhere to AZT therapy. 29,36,37 When they controlled for psychiatric morbidity, however, Ferrando and colleagues found no difference in adherence to AZT therapy between injection drug users and noninjection drug users. 31 In addition, recent studies since the development of HAART have found no relation between injection drug use and adherence. 30,38 These inconsistent findings may be partly due to confounding of injection drug use and psychiatric morbidity. 31
The type of substance used may have a stronger impact on adherence than the route of use. Cocaine or amphetamine injection has been associated with a more chaotic lifestyle than heroin injection, for example. The use of other substances, such as crack cocaine or alcohol, 30,39 may be more consistently associated with nonadherence than injecting drugs. Also, although treatment for substance abuse is available, it is less predictably effective (and less modifiable) than treatment for depression. Treatment efficacy varies by drug and treatment modality; methadone treatment predictably decreases the frequency of heroin injection, while behavioral therapy for injecting cocaine or using methamphetamine is less predictably effective. 40–46 Our understanding of the effects of drug use or treatment for drug use on adherence is limited. Therefore, the descriptive term “drug use” provides less reliable guidance than the diagnosis of depression in the decision to delay HAART.
Even when it is appropriate to delay HAART, it should not be delayed indefinitely. The decision to delay HAART while barriers to adherence are addressed should be balanced against the need to start HAART early enough to realize its benefits. Although the initiation of HAART is not a medical emergency (except as postexposure prophylaxis, for prevention of maternal-fetal transmission, and possibly during the acute HIV retroviral syndrome), the urgency of therapy increases with the stage of disease. The mortality benefit of HAART is most pronounced in patients with late-stage disease. 47 Patients may not realize the maximum improvements in immunologic function if therapy is delayed, and they may lose significant function if therapy is delayed until after the first opportunistic infection. Therefore, the need to start therapy increases as the disease becomes more advanced. Fairfield and colleagues observed that the stage of disease progression, which they measured by the CD4 count, a history of a prior opportunistic infection, and viral load, was a significant predictor of when therapy was started.
Although it may be appropriate to delay HAART while treating modifiable barriers to adherence, such as depression, it is not appropriate to exclude patients with poorly understood or poorly modifiable barriers to adherence, such as drug use, from ever receiving HAART. We previously argued that there was no ethical obligation to provide therapy that might be ineffective due to nonadherence. We also suggested that two-drug antiretroviral therapy might be a reasonable option when there were grave doubts about a patient's adherence. 48 But it now appears that antiretroviral therapy with three (or more) drugs not only slows disease progression but may also halt or even reverse disease progression. 49–51 We suggest that this change in knowledge, combined with the difficulty of predicting adherence, favors a more aggressive approach.
Our ability to predict adherence in individual patients is dismal, perhaps no better than chance. 30,52 Furthermore, models that predict poor candidates for therapy will inevitably fail for some patients. The goal should be to maximize success for each patient. Efforts to understand adherence are perhaps best directed toward identifying modifiable barriers, such as depression. These modifiable barriers should be addressed, to the extent possible, before starting HAART. The delay of HAART based on modifiable factors is in the best interest of both the individual patient and public health. At some point, however, delaying therapy becomes indefensible, whatever the barriers to adherence. We suggest that, given our imprecise understanding of adherence and its predictors, no person should die without a trial of therapy simply because the person was judged in advance to be nonadherent.