Our study showed that the majority of the patients attending TASO Mulago clinic, Kampala, had good adherence to HAART with 78.2% of the patients achieving a mean adherence of greater than 95%. Comparison of adherence to HAART between studies is fraught with many obstacles including differences in study population and design, definition and measurement of adherence, as well as differences in sample size and type of HAART. Nevertheless, the adherence rate in our study was similar to that reported in previous studies in sub-Saharan Africa [14
]. The high adherence rate to HAART observed in TASO is remarkable considering that its HAART programme is in a non research or government setting. The high adherence to HAART can be attributed to the continuity of care and close support of the patients that is the hallmark of TASO and which has also been described elsewhere [19
]. TASO originated as a patient-centred care and support organization and this philosophy has persisted even after the introduction of HAART.
The overall mortality was 12 per 100 person-years but the estimate may have been affected by four major factors. Compared to settings where HAART is initiated at CD4 counts of <350 cells/mm3
the patients in our study were initiated late on HAART (<200 cells/mm3
), though it is the policy of Ministry of Health. This late initiation on HAART may have increased the mortality in our study. We also excluded participants who had been on HAART for less than one year in an effort to capture mortality that occurred only after HAART was deemed to be effective. By so doing however we may have inadvertently excluded early mortality and yet several studies have reported that mortality among HAART patients is greatest in the first three months of treatment [3
]. Thus exclusion of early mortality may have led to an underestimate of overall mortality in our study.
In our study, there were difficulties in ascertaining the cause of death. The cause of death was determined by the TASO medical team and was mostly in the form of verbal autopsy which classifies deaths in broad categories but does not permit assignment of a specific cause of death [21
]. Thus there were limitations in determining the underlying cause of death and we therefore performed all-cause mortality analysis. The use of all-cause mortality in our study may have led to an under or overestimation of the overall mortality among HAART patients. The use of the multiple imputation method to impute missing data may also have distorted our estimates if the missing data were not random[22
]. Finally, under or overestimation of mortality could have occurred because a significant proportion (16.4%) of the participants on HAART were lost to follow up. Nevertheless, the overall mortality in our study was similar to that reported in previous studies [10
Non-adherent participants had a mortality of 42.5 deaths per 100 person-years and, after adjusting for age, sex and education level, were two times as likely to die as adherent participants. These findings are consistent with other studies that have reported that non-adherence is associated with increased mortality [12
]. Non adherence to HAART leads to virologic, immunologic and clinical failure that is mediated mainly by two potentially re-enforcing mechanisms. Non-adherence to HAART leads to failure to suppress viral replication, thus increasing the likelihood of developing HIV mutations that could lead to the development of drug-resistant viral strains. Secondly, non-adherence to HAART fails to prevent further viral destruction of the cellular immune system with consequent reduction in the level of CD4+ cells and development of opportunistic infections [15
]. It is worth noting that mortality among adherent participants in our study (6.1 per 100 person-years) was comparable to mortality rates reported in developed world settings.
In our study we found overwhelming evidence of a statistical interaction between baseline CD4 count and non-adherence i.e. baseline CD4 count modified the effect of non-adherence. In patients with a CD4 count <50 cells/mm3
non-adherence increased the risk of mortality four fold, whereas in patients with a CD4 count of 50 cells/mm3
or above at baseline, non-adherence doubled the risk of mortality. Previous studies have reported that CD4 counts at the initiation of HAART independently modify survival with decrease in CD4 count being associated with increased mortality [3
]. Although Nachega et al reported that there was no interaction between adherence and baseline CD4 count, they found that CD4 count strongly modified the effect of adherence on survival, with the greatest mortality (HR 4.54, 95% CI: 2.83–7.29) occurring in the most immunosuppressed patients (≤ 50 cell/μL.
The findings of this analysis have important policy implications particularly for HAART programmes in developing countries. Our study has shown that good adherence to HAART and improved HIV survival are feasible in a community HIV/AIDS treatment and care programme. Uganda, like many developing countries, has limited capacity to provide HAART through the government health care system. Accordingly, there is increasing need to provide HAART through non governmental programmes including home or community care programmes. As of June 2007, of the more than 100,000 people in Uganda taking HAART, 14% received their drugs through home or community care programmes, mostly through TASO.
The success of the TASO community HAART programme can be attributed to a number of factors including the personalized patient-centred approach; the continuity of care and follow up through home visits; total patient and family care and support; and free access to services. These factors have contributed to the high adherence rates in TASO that are critical for improved survival of HIV infected people. The major challenges of the TASO community HAART programme include the provision of high quality clinical and laboratory services (e.g. CD4 cell counts) to support HAART; funding to provide sustainable provision of antiretroviral drugs and other supplies; and the need to balance the expansion of the TASO community HAART programme to meet increasing demand without losing the personal touch that has over the years ensured the quality of services and success of TASO.
Our study highlighted the poorer outcomes in HIV infected people with low CD4 counts and echoes the need to identify and treat eligible HIV infected people early and for intensive adherence interventions to target HIV-1 infected patients with the lowest CD4 counts [16
]. The home-based VCT programme that is also offered in TASO is one way in which HIV infected people can be identified early so that they benefit maximally from HAART.
The principal limitation of our study was the study design. The study was a retrospective cohort, and therefore we did not control for all the potential confounders that could confound the association between non-adherence to antiretroviral therapy and patients' survival, such as viral load and opportunistic infections at initiation of HAART. Viral loads were not recorded by TASO, while there was no precise information on opportunistic infections at the time of initiation of antiretroviral therapy. This situation may have resulted in an under or over estimate of the association between non adherence and survival. However, random error was unlikely to have biased our findings because the study needed to observe 121 deaths to show a significant effect in survival but we observed 164 deaths.
The measurement and definition of adherence to antiretroviral therapy may have introduced bias. There is no gold standard for measuring adherence to antiretroviral therapy [14
]. The approaches employed in this study included patient self-report and pill counts. The self-report obtains a patient's subjective evaluation of his or her own level of treatment compliance behaviour. The advantages of the self-report method include its simplicity, speed, and viability of use. The disadvantages include reliance on recall and social desirability bias, with a tendency to overestimate adherence. Nevertheless, several studies highlight the usefulness of the self-report as an adherence measurement tool and show it to correlate well to virologic outcome [30
]. Adherence measurement depended on a clinician's subjective judgment. Also using mean adherence over the period a patient was on treatment could have introduced measurement bias into our study.
In our study, the adherence assessments for non-adherent and adherent patients were unequal. One interpretation of this finding is that unequal assessment of adherence between the adherent and the non adherent patients may have introduced bias in our estimates. However, an alternative and more plausible explanation is that differences in adherence levels between patients especially in the clinic-based programme may instead have lead to differences in adherence assessments between the adherent and non adherent patients since the adherence assessments were based on patient visits to the TASO clinic. The type of regimen patients were taking could have had an impact on their survival and hence distorted our estimates. As already mentioned, we could not control for the type of regimen patients were taking. While our findings may have been affected by such factors, we believe that our overall conclusions are robust.
In conclusion, our study showed that good adherence to HAART and improved survival are feasible in community HIV/AIDS treatment and care programmes such as that of TASO, Uganda. However, there is need to support community HIV/AIDS HAART programmes to overcome the challenges of funding to provide sustainable supplies particularly of antiretroviral drugs; provision of high quality clinical and laboratory support; and achieving a balance between expansion and maintenance of quality of services. Measures for early identification and treatment of HIV infected people including home-based VCT and HAART should be strengthened.