A dramatic reduction in HIV-related morbidity and mortality has been recognized in countries where ART has been made widely available (Anna et al., 2002). However it is also recognized that extremely high levels of adherence to ART (at least 95%) are needed to ensure optimal benefits, and that this may often be complex in terms of the pill burden, dietary restrictions and dosing frequency. Where adherence is sub-optimal,

HIV rapidly selects for resistance (Papella et al., 1998), in part due to rapid and error-prone replication (Perelson et al., 1996) but also often aided by the low genetic barrier of several ARVs to resistance (Kuritzes et al., 1996). Though effective adherence levels have not been fully defined for ART, levels of adherence below 95% have been associated with poor virological and immunological responses (Paterson et al., 2000; Orrell et al., 2003). Other data suggest that 100% adherence levels achieve even greater benefits (Fischl et al., 2000). Estimates of average rates of adherence to ART range from 50% to 70% in many different social and cultural settings, and the risks associated with sub-optimal adherence are extensive at both individual and societal levels (Chesney et al., 2000; Bangsberg et al., 2000; Liu et al., 2001; Nemes et al., 2004; Saferen et al., 2005).

Concerns about low adherence have been cited by those who question the feasibility of rapid scaling up of ART programmes in resource-poor settings (Stevens et al., 2004; Gill et al., 2005). Harries et al. (2001) argued that adherence problems would constitute a perceived significant barrier to the delivery of ART in sub-Saharan Africa. They warned that unregulated access to ARVs in sub-Saharan Africa could lead to the rapid emergence of drug-resistant viral strains and individual treatment failure, curtailing future treatment options and leading to the transmission of drug-resistant strains of HIV. The authors also maintained that, at present: there are few health care providers skilled in the provision of ART and in the management of patients who are on treatment; the existing health infrastructure is incapable of monitoring viral load, immune status, or the side-effects of ARVs; medicine procurement and distribution systems are weak; and interruptions in the medicine supply chain are likely. In addition, they highlighted current concern about the theft of medicines from health institutions for sale in the market, shops and private clinics, and across national borders.

2.3.1 Measurement of adherence

There is no gold standard by which to measure adherence to medication. Many studies employ a number of methods, either alone or in combination to measure adherence. The most common include: electronic drug monitoring (EDM) devices, pill counts, biochemical markers, pharmacy refill records and various self-reporting tools such as questionnaires and visual analogue. According to Gill et al. (2005) the hierarchy of adherence measures ranks physician and self-assessment report the least accurate, pill count intermediate and EDM the most accurate adherence marker. However, no single measure is appropriate for all settings or outcomes. It has been found that the use of more than one measure of adherence allows the strength of one method to compensate for the weakness of the other and to more accurately capture the information needed to determine adherence levels (Vitolins et al., 2000).

Studies in African settings have indicated optimal adherence rates (i.e., the proportion of patients who adhered to their ART schedule at least 95% of the time) ranging from 54% to 98% depending on the measure used: Botswana (Weiser et al., 2003: 54%); Nigeria (Daniels, 2004: 79%); South Africa (Ferris et al., 2004: 77%; Darder et al., 2004: 80%); Uganda (Byakika-Tusime, 2003: 67%; Munganzi, 2004: 98%); and Rwanda (Omes, 2004: 87%).

2.3.2 Factors affecting adherence

A number of factors have been associated with adherence to ART and are commonly divided into five intersecting categories (Reiter et al., 2003). These categories are: patient variables, treatment regimens, disease characteristics, patient-provider relationship, and clinical setting.

Patient variables

Patient variables include sociodemographic factors (age, gender, race, income, education, literacy, housing status, HIV risk factors) and psychosocial factors (mental health, substance abuse, sociocultural issues and support, knowledge and attitude about HIV and its treatment) (Carrieri et al., 2002; Nemes et al., 2003; Murphy et al., 2004; Machtinger and Bangsberg, 2005).

Sociodemographic and psychological issues have great potential to impact on adherence. For instance, family support and religious beliefs about illness and medication may influence motivation and adherence (Becker, 1990; Haynes, et al., 1996; Chesney, 1997). The issue of disclosure has also been found to have serious implications for adherence (Ormazu, 2000; Klitzman et al., 2004; Zea et al., 2005). For example, the use of medication may inadvertently reveal a person's HIV status; poverty may prevent individuals from following treatment-related dietary advice; drug and alcohol abuse may impair judgment and the ability to adopt and maintain routine medication use; and family responsibilities may require adults to place the health care needs of others before their own. Mental health problems such as depression have been associated with low adherence in HIV-positive adults and adolescents as have other psychological variables such as perception of one's ability to follow a medication regimen, or self-efficacy (Singh, 1996; Eldred, 1998; Murphy, 2001; Tuldra, 2002). Beliefs about health and illness, in particular about the necessity of medication to ward off illness and concerns about potential medicine-related adverse events have been found influential in both HIV and other disease areas (Horne, 2001).

Although side-effects have been cited by some studies in developed countries as predictors of adherence, experience of symptoms and views about medications may be complex and may vary according to the type of regimen (Chesney, 2000; Carr and Cooper, 2000; Ammassari et al., 2001; Carr, 2002; Murphy et al., 2004). Symptoms may stimulate the use of medications by acting as a reminder or reinforcing beliefs about the necessity for treatment. However, patients' expectations of symptom relief are also likely to have an important effect. This could be problematic if expectations are unrealistic, or where treatment is given for asymptomatic disease, as occurs with HIV infection (Horne, 2001). In addition, patients' concerns about the potential harm of ART may be entirely rational. Horne and colleagues have proposed that for some individuals missed doses may be a logical attempt to moderate this risk by taking fewer medications (Horne, 2001). Patients who understand the rationale for ART and treatment failure report higher adherence levels than those without this information (Anderson, 1999; Horne, 2001). Efforts both to reinforce information provided verbally with written information to take home and to check that information has been correctly understood are likely to be beneficial, as patients commonly misunderstand their health care provider's instructions. One study found that 13% of patients prescribed ART were not taking their medication correctly, despite believing that they were (Bangsberg, 2001).

However, studies investigating the role patient variables play as predictors of adherence have produced largely inconsistent results. The tendency to ascribe low adherence to (often deprived) social groups is a well-established trend in the general literature (Horne, 1998). However, as later experience with antibiotics would demonstrate, low adherence is not restricted to certain social classes but is widespread and unpredictable (Lerner, 1998). Moreover, adherence rates vary not just between individuals but within the same individual over time (Carrieri, 2002). Adherence is therefore best thought of as a variable behaviour rather than as a stable characteristic of an individual. Most people will exhibit low adherence some of the time (Horne, 1998).

Treatment regimens

These include the number of pills prescribed, the complexity of the regimen (dosing frequency and food instruction), the specific type of ARV and medication side-effects. The complexity of the regimen and side-effects caused by it are clearly associated with sub-optimal adherence (Machtinger and Bangsberg, 2005).

Patient-provider relationship

This includes the patient^’s overall satisfaction and trust in the provider and clinic staff; the patient's opinion of the provider's competency; the provider's willingness to include the patient in the decision-making process; the affective tone of the relationship (e.g. warmth, openness, cooperation); the compatibility of race/ethnicity between patient and provider; and the adequacy of referral.

Clinic setting

This includes the availability of transport, general environment, flexibility of appointments, perceived confidentiality, and satisfaction with past experience with the health care system. Chesney (2003) found that dissatisfaction with the health services is a predictor of non-adherence.

Disease characteristics

This includes: the stage and duration of HIV infection, associated opportunistic infections, and HIV-related symptoms. The severity of the illness could impact negatively or positively on adherence to ART.