Poor adherence to HAART can lead to HIV disease progression, evolution of drug resistance, and subsequent immunological and clinical failure.[2], [3] Adherence to antiretroviral therapy is essential for maximal suppression of viral replication and is believed to be a critical determinant of long-term survival among HIV-infected individuals.[4], [5] Non-adherence to these medications can greatly lower their effectiveness resulting in poor therapeutic outcomes.[2], [3] More than 95% of the prescribed antiretroviral treatment doses should be taken for optimal response; lesser adherence is often associated with virologic failure.[6] Recent studies demonstrate that HIV-infected individuals who adhere to HAART have HIV viral load reduced to levels so low that HIV transmission is dramatically reduced or eliminated.[7] Thus, the benefits of adherence to HAART make it critical to identify factors associated with non adherence that could lead to increased HIV incidence, treatment failure and drug resistance.

The last visit within 6 months prior to HAART initiation tended to be a little over 2 months (median =65.5 days) prior to the date of initiating antiretroviral treatment. The date of the visit closest to 12 months post HAART initiation (the post-HAART visit) tended to be almost one year (median =347.5 days) following the start date of antiretroviral treatment. All women, 354/354, reported 100% adherence to HAART at the post-HAART visit. As adherence was 100%, further analyses of predictors of non-adherence were not undertaken.

Adherence in this study is higher than reported elsewhere including from other African and non-African countries.[8], [16], [17] For instance, in a study assessing factors associated with medication adherence in HIV-infected adults in Botswana, overall adherence was 81%.[16] In Nigeria, a similar study showed adherence rates to HAART of approximately 75%.[17] Our findings, however, are consistent with an earlier small study of adherence to HAART in Rwanda. A 2003 study in the Ensemble de Solidarité Therapeutique Hospitaliere En Reseau (ESTHER) antiretroviral treatment program at the Centre Hospitalier Universitaire de Kigali in Kigali, Rwanda, found that only 5% of patients reported missing a dose of their HAART medication in the 3 days prior to their adherence visit.[18] This level of adherence was confirmed using therapeutic drug monitoring in a group of 41 patients taking Triomune or Triviro (single-pill fixed dose combinations of stavudine/lamivudine/nevirapine). Only 5% of patients in this group had undetectable drug levels and 93% had levels within the therapeutic range.[18]

Our own skepticism about the high adherence in this cohort prompted us to informally compare our findings with other Rwandan health facilities. Three of the authors [SM, EM, FM] looked at HAART adherence rates by examining clinical pharmacy statistics at King Faisal Hospital Kigali (KFH, K), WE-ACTx and Gicumbi Hospital respectively. Pharmacy data from KFH, K indicated that 94% (367/400) of patients on HAART returned to refill their medications within three days of their scheduled appointment during September to December 2010, and between January to February 2011, 100% (400/400) of patients returned to fill their medications. Similarly, in the first quarter of 2010, 94% (1071/1139) of patients on HAART at the WE-ACTx clinic refilled their medications at the appropriate time. For the last quarter of 2010, 93% (2081/2237) of patients on HAART at Gicumbi Hospital appropriately refilled their medications.

Another indication that the level of adherence found in this study is accurate can be assessed by analyzing the observed changes in CD4 cell count and MCV from zero to six months prior to HAART initiation to approximately 12 months post HAART initiation. On a population level median CD4 cell counts rose about 80 cells/mm³ from 185 to 265 while median MCV increased about 16 fL from 88 to 104. The observed 80 cells/mm³ change in CD4 cell count is consistent with data from clinical trials which demonstrate that patients on HAART with good virologic control show an average increase of approximately 50–100 cells/mm³ per year until a steady state level is achieved.[19] The observed rise in median MCV of 16 fL from 88 to 104 may also indicate adherence to HAART for those patients taking zidovudine or stavudine. Prior studies have correlated the development of macrocytosis with zidovudine and stavudine use and suggested that simple observations of patients' MCV could be an effective method of monitoring antiretroviral adherence.[20] While we considered looking at individual changes in MCV and CD4 to see if these could identify specific non-adherers who were misreporting adherence, the repeat visit random change and measurement error of laboratory parameters such as this is so high as to negate this use of between visit changes.[21]

There are several possible reasons why adherence in this cohort of Rwandan women could be very high. Unlike many other settings, in Rwanda, when a patient becomes eligible to start antiretroviral treatment, national guidelines require that person to attend 2- to 3-day educational sessions with a “buddy” who can support his or her adherence to HAART.[22] There are mandatory social criteria for starting HAART, including acceptance to be visited by a health care worker, having a trusted person to help with compliance usually called a treatment buddy, having a fixed residence within Rwanda in a known catchment area of a health facility and disclosure of HIV status to a trusted family member. To further promote adherence, there are also follow up visits at the village level guided by a community health worker and or a member of associations of people living with HIV and AIDS.

Certain cultural aspects of Rwandan society may also contribute to high levels of HAART adherence. Following the 1994 war and genocide there has been a strong societal focus on community coherence and the sharing of meager resources. Within this communal environment the loss of a family member is a loss to the whole village or umudugudu and to a wider Rwandan community; thus, there is a strong focus on promoting health and reducing morbidity and mortality. Family members within a community are obliged by custom to support an HIV-infected person through the journey of compliance to antiretroviral therapy.

Other national characteristics specific to Rwanda may have contributed to the high adherence to HAART seen in this study. In Rwanda there is almost universal health coverage with 96% of Rwandans having health insurance, 91% of which is provided through the government-supported Mutuelle de Santé program.[22] Furthermore, in Rwanda there is a strong emphasis on personal and public health. For example, immunization coverage for children under one year of age for the six killer diseases (tuberculosis, poliomyelitis, tetanus, diphtheria, pertussis, measles) is the highest globally with 100% coverage as reported by both national statistics and UNICEF.[23] Lastly, in Rwanda, all HAART medications are provided free of charge to patients once they meet eligibility criteria for treatment initiation. The general promotion of health, almost universal insurance coverage and availability of medical care and free medications all likely contribute to better overall HAART adherence.

Some limitations of this analysis should be noted. Our study was in a cohort consisting entirely of Rwandan women and may not be generalizable to other sub-Saharan African populations. Prior studies have indicated that women may have higher HAART adherence compared to men.[17], [24], [25] This gender disparity could have been a contributing factor to the high adherence found in this study. All study participants were being followed as part of a larger research study. The majority of these participants received their HIV care at a non-governmental organization clinic where there were greater outreach efforts such as patient home visits and extra resources invested in care and treatment of women. Although there are remarkable initiatives to improve adherence even in government health care clinics, the extra resources and close follow-up for majority of patients in RWISA cohort study, could have resulted in better overall adherence compared to the general population. While only 15 of the 490 women who began HAART dropped out of the study we did not have adherence data for the 6 woman who died prior to 12±3 months post HAART initiation. There were an additional 80 women who remained in RWISA but did not have a study visit within the 12±3 months time points designated for this study. The median time from HAART initiation to follow-up for these 80 women was 639 days. Similar to the 12±3 month cohort 4 women had stopped HAART without being told to do so by their physician and the remaining 76 women reported 100% adherence. This adherence measurement occurred outside of the 12±3 months time frame set for this study so this data was not included in our analysis.

We measured HAART adherence by self-report using structured interview questions that can be subject to overestimation as patients tend to overstate their adherence to treatment. Other measures of adherence, such as pill counts, pharmacy records, electronic devices or therapeutic drug monitoring were not available to us. Even so, measuring adherence using patients' self-report can be easily replicated in most resource-limited settings including Rwanda making it a good measure for comparison. We examined the median CD4 cell count and MCV at the pre and post adherence visit to corroborate the self-reported adherence of the study participants. Given the significant variability in these between-visit hematologic parameters it is possible that individual-based changes could be due in part to underlying variability in the observed data.[21], [26], [27] However, due to the central limit theorem when comparing median CD4 cell counts (or MCV levels) in a sample size this large (354) such variability tends to cancel out, making population-based comparisons a more reliable indicator of what is happening in the population as a whole.[21] Lastly, we chose to define adherence as taking all prescribed HAART medications as directed in the three days prior to the 12±3 months adherence visit. This definition of adherence has been used in prior studies which observed substantially lower rates of adherence than we observed.[28], [29] We did not ask patients if they had missed medication doses in the month prior to their clinic visit or if they “more generally take most or nearly all of their prescribed medications”. It can be argued that this single pre-clinic visit measurement of adherence may not accurately reflect more general adherence.