This analysis, including nearly 50, 000 patients, over eight years and covering a network of 30 sites in three countries, suggests that in addition to rapid expansion, the roll out of ART services for HIV-infected patients in east Africa is improving in effectiveness. First, we documented that the median CD4 count at the time of ART initiation in the east Africa region has increased substantially from 87 cells/mm3 in 2002-03 to 185 cells/mm3 in 2008-09. Second, we found improvements in the pharmaco-epidemiology of the roll out over time with reductions in more toxic regimens: the use of D4T in the first regimen fell from a peak of 88% to 58% in 2008-09, and the fraction of patients starting NVP-based regimens decreased as more regimens made use if EFV instead, even after taking into account changing trends in sex of patients starting ART over time.
Third, the fraction of patients who had longer travelling times to a clinic declined by 50% in the programme from which we had these data available, and this likely reflects the ongoing process of decentralization of ART services, a key element of improving access to ART services. Lastly, our analysis underlines that, after 2005 when the Global Fund and PEPFAR began supporting large-scale ART programmes, self-pay for ART, which has been shown to be associated with poor outcomes in Africa [19
], essentially disappeared.
This analysis, which covers a large span of both time (indeed, capturing data before the rapid Global Fund- and PEPFAR-funded rise in ART availability in 2004-05) and geographic regions, also presents the opportunity to identify "gaps" in access to ART in east Africa. Existing literature has suggested that more patients with advanced disease at ART centres are men [20
]. One explanation is that many asymptomatic women with high CD4 counts are detected when screened during pregnancy and referred to ART centres.
Although we also found male sex to be associated with advanced disease at presentation (defined as either WHO Stage 4 or a CD4 count of ≤ 50 cells/mm3), we also found that a history of PMTCT did not explain all of this association. Men were 50% more likely to access ART with advanced disease compared with women without a history of PMTCT. But women without a history of PMTCT were nearly twice as likely to access ART with advanced disease compared with women with a history of PMTCT. Because most women tested during pregnancy receive some form of PMTCT, it appears that socio-behavioural characteristics of men in east Africa confer additional risk of having advanced disease that is not completely explained by public health services targeting pregnant women. Further evaluation of the causal reasons for the association between late disease presentation and male sex among patients accessing ART is required.
Although the magnitude of the differences were not large, patients in Uganda and Tanzania in this analysis were more likely to start ART with advanced disease as compared with Kenya after adjustment for socio-demographic factors and calendar time. This is likely explained by the fact that the Tanzanian scale up of ART services occurred slightly later and coverage has only recently begun to catch up with - a history that is consistent with a programme seeing more advanced disease at ART initiation.
In Uganda, the risk of presentation with more advanced disease is driven by a larger fraction of patients with a WHO Stage 4 condition despite approximately equal CD4 cell count levels. This association may be due to differential clinical assignment of WHO stage when definitive diagnoses cannot always be made. Further study, in cohorts where definitive diagnoses are available, are needed to exclude the possibility that Ugandan patients have a higher disease burden even after accounting for CD4 count levels. Overall, when compared with recent reports from southern Africa, the demographics in our patients in east Africa are similar (i.e., proportion of women and age). The median CD4 count at ART initiation, however, is slightly higher at 122/mm3
in our analysis as compared with 103/mm3
in South Africa during a similar period. This likely reflects the higher burden of disease in the epidemic population in South Africa [21
]. Furthermore, the CD4 count at ART initiation in our study during 2008-09 period was 154 cells/mm3
, which is not far from contemporaneous figures of 187 cells/mm3
in the United States, 159 cells/mm3
in Brazil and 157 cells/mm3
in China [22
Given the increasing consensus of the high D4T toxicity, crystallized in the WHO 2010 recommendations [14
], monitoring declines in D4T use is an important objective of contemporary pharmaco-epidemiology. A "gap" we observed is the marked differences in the use of D4T and NVP over time and across countries with similar economic resources. Although the fraction of patients starting D4T-based regimens declined in all countries, programmes in Uganda moved away from D4T earlier and more extensively than those in Kenya and Tanzania, and by 2009, included nearly 8% of regimens, which contained tenofovir.
The reasons for these differences, including the differences in healthcare systems and national policies, across countries with similar economic resources available for health require further explanation and research. In particular, interdisciplinary research focused on the economic, programmatic and policy issues that inform national implementation programmes may yield further explanations. Furthermore, the cost effectiveness of these differences should be quantified: although countries using less D4T may have higher per patient costs in the short run, the long-term quantification and prevention of morbidities associated with prolonged D4T use must also be assessed.
Our analysis of factors associated with specific ART drugs reflected, in general, rational drug choices and are reassuring from a public health perspective. We found several notable associations with D4T use. Male sex, older age, higher pre-therapy CD4 counts and TB were associated with a reduced risk of D4T use. Lower D4T use in men and at higher CD4 count levels may be explained by the lower prevalence of anaemia in men and in healthier patients, and hence the absence of a contraindication to zidovudine (AZT) use. Reduced D4T use in patients with TB is potentially explained by the desire to avoid the neurotoxic combination of isoniazid and D4T [23
]. The observed preponderance of efavirenz (EFV) use in men likely reflects the desire to avoid EFV in women who desire children, and the elevated use of EFV in those with TB reflects recognition of the interaction between NVP and rifampin [24
A limitation of this manuscript is the cross-sectional nature of the analyses. In cross-sectional studies, inferences must be made with care because time ordering is not possible and selection bias is difficult to control. For example, the finding that more men have advanced disease among patients accessing ART leads to the nominal inference that the socio-behaviour characteristics of HIV-infected men in east Africa prevents them from seeking care. However, an alternative may be considered: if the distribution of advanced disease is similar in men and women in the community (a plausible assumption especially early in the roll out when few patients were accessing ART), then the observation that male sex is associated with at advanced disease among patients accessing ART could imply that female sex is associated with advanced disease in patients unable to access ART in the community.
A second limitation is that, although IeDEA spans three countries, the patients from each country may not be representative of the country as a whole in all aspects. We believe, however, that sites are in general prototypical ART scale-up clinics staffed and stocked by national ministries of health and implementing partners of the Global Fund and PEPFAR and should be fairly representative of clinics in these countries. Third, the data we analyzed were collected in the course of routine clinical care, and the accuracy of certain measurements, such as WHO staging, may be limited by few diagnostic options. This may explain why a later calendar year was associated with a slightly high proportion of patients classified as WHO Stage 4 in Uganda even though the CD4 count levels rose during that interval.