In 2017, there were 6.9 million adults living with human immunodeficiency virus (HIV) in South Africa, of whom 61% were receiving treatment.¹ This equates to 4.2 million adults receiving antiretroviral therapy (ART), forming the largest treatment programme in the world.¹ South Africa’s ART programme has demonstrated successful initiation and management of clients on treatment for over a decade, despite rapid programmatic scale-up over this time,² and has achieved near-normal life expectancy for HIV-infected adults with timely ART initiation.³ The ART programme is thus an essential component of the third Sustainable Development Goal which aims to ensure healthy lives and end the AIDS epidemic by 2030.⁴

South Africa’s ART programme has evolved over time to accommodate increasing numbers of HIV-infected clients and to provide improved care and treatment services (Table 1).^5,6,7,8,9,10 This has encompassed expansion of the ART eligibility criteria from clients with CD4 counts < 200 cells/mm³ in 2004⁵ to universal test and treat (UTT) in 2016,⁸ as well as a shift from doctor- to nurse-managed care and decentralisation of ART services from hospitals to primary healthcare sites.¹¹ Across all years, cotrimoxazole preventive therapy (CPT) has been recommended for clients with baseline CD4 counts below 200 cells/mm³ as a means of protecting against infections and thereby reducing morbidity and mortality.^5,6,9 Triple-therapy antiretroviral drug regimens have also been updated over time, from combination single formulation regimens including stavudine in 2004⁵ and tenofovir in 2010⁶ to a fixed-dose combination (FDC), a single tablet containing three antiretroviral drugs (tenofovir, emtricitabine and efavirenz), in 2013.⁷ As the ART programme has evolved and expanded, it has become increasingly important to have an effective monitoring system, and in 2010 the Department of Health adopted an electronic monitoring and evaluation tool known as TIER.Net, which was developed by the University of Cape Town’s Centre for Infectious Disease Epidemiology and Research.¹² TIER.Net is used operationally to monitor baseline clinical care and client outcomes over time, providing a rich source of cross-sectional and longitudinal routine ART data.

Despite the widespread availability of ART, there is still a considerable burden of HIV-related morbidity and mortality. In South Africa, HIV accounted for almost two-thirds of medical admissions at one hospital in 2012 and 2013,¹³ with no improvement in the number of deaths due to AIDS nationally from 2013 to 2017,¹⁴ and in West Africa, AIDS-defining conditions remained the primary cause of hospitalisation among HIV-infected adults years after the scale-up of ART services.¹⁵ High morbidity and mortality is specifically associated with late presentation for HIV care, as indicated by low baseline CD4 counts or advanced clinical stage.^3,16,17,18 These poor outcomes potentially undermine the population-level impact of the ART programme, as ART coverage in individuals is known to impact HIV transmission, incidence and community viral load.^19,20 Interventions to improve ART initiation, specifically among clients at risk of presenting late for HIV care, are therefore essential to improve both individual- and programme-level outcomes. This study used operational programme data to describe ART initiation and outcome trends over time, with a focus on clients presenting late for care, so as to identify programmatic gaps that can guide interventions to reduce HIV-associated morbidity and mortality. Two districts were investigated as examples of the urban and rural HIV epidemics in South Africa, namely, Johannesburg and Mopani districts. Specifically, this study aims to use routine TIER.Net data from adult clients in an urban and rural district of South Africa to (1) describe ART programme growth and baseline CD4 count over time, (2) analyse 5-year mortality in the context of baseline CD4 count and (3) describe the population initiating ART at low CD4 counts (< 200 cells/mm³) in 2017 in order to identify priority groups at high risk of mortality for intervention.

The study was approved by the University of the Witwatersrand’s Medical Ethics Committee (M140461). Individual patient consent was not required as this study analysed anonymised TIER.Net data that were routinely collected at healthcare facilities for monitoring purposes. No data collection was performed for the purposes of this study. In addition, no patient files or electronic medical records were retrieved or accessed at any stage.

This study of routine TIER.Net data demonstrates an increase in mean and median baseline CD4 counts over time, in line with South African guidelines that have repeatedly raised the CD4 cut-off for ART initiation^5,8,9,10 and in agreement with previous findings in South Africa and other countries.^{2,16,24,25,26} However, the improvement in these population measures is masking a high-risk group of clients with low CD4 counts who only initiate treatment once they have advanced HIV disease. Although the proportion of clients initiating ART with CD4 counts < 100 and < 200 cell/mm³ declined in the initial years of the ART programme, as has been previously demonstrated from 2001 to 2012,¹² there has been little change in recent years, with the proportion of clients with low baseline CD4 counts remaining concerningly high. This is unexpected in the context of UTT, but it confirms findings from a recent South African study of laboratory data which also noted little change in the proportion of adult clients initiating ART with CD4 counts < 200 cells/mm³ between 2012 and 2016, ranging from 32.9% to 34.8% nationally.²⁷ This is also a challenge in other low- and middle-income countries, with over 30% of clients initiating treatment with CD4 counts < 200 cells/mm³ despite generally widespread availability of ART.^24,25 These clients, who may either be ART-naïve or may have previously defaulted from the ART programme, represent a considerable burden of increased medical costs, risk of transmission and HIV morbidity.²⁷ Interventions to address HIV incidence as well as barriers to entering care are essential.

Mortality was found to be significantly higher in clients with baseline CD4 counts < 200 cells/mm³ compared to those with higher CD4 counts, as has been well established in South Africa and other settings.^3,16,17,28 Interestingly, it has been shown that standardised mortality rates which take background mortality into account are similar for all CD4 strata by 4 years after ART initiation, suggesting that baseline immunological status impacts mortality primarily at shorter durations of ART.² It is specifically at these early time periods after ART initiation that there is a notably high mortality rate, as has been demonstrated in this study, specifically among clients with baseline CD4 counts <200 cells/mm³. Other studies in South Africa and sub-Saharan Africa have similarly demonstrated high mortality rates in the first year after starting ART, with up to 26% of patients dying in the first year of treatment.^3,29 There is a strong association between early mortality and the degree of immunodeficiency at ART start; therefore, strategies to reduce mortality must address ‘up-stream’ entry points into the treatment cascade, including earlier diagnosis of HIV and improved longitudinal care pre-ART.²⁹

Although clients with low baseline CD4 counts are at increased risk of death, mortality rates have decreased in recent years of the ART programme in this and other studies.^2,16,17,26 Overall mortality in South Africa has declined in recent years³⁰ and the gains in life expectancy have been largely attributed to a decline in HIV-related mortality rates.³¹ This decline can be ascribed to improved accessibility to HIV services with the implementation of nurse-managed care,¹¹ as well as improved quality of care because of enhanced drug regimens. This study demonstrated significantly lower mortality among clients with CD4 counts < 200 cells/mm³ who initiated ART in more recent years when an FDC was in use compared to those initiating treatment in previous years, with the highest mortality in years when guidelines recommended the use of stavudine which has a relatively unfavourable toxicity profile.³² Antiretroviral therapy using an FDC has been shown in a European study to improve both adherence and quality of life of HIV-infected patients, while maintaining virological and immunologic efficacy.³³ In general, the simplification of ART with an FDC is associated with improved adherence because of decreased pill burden, reduced risk of resistance owing to the inability of patients to take partial regimens, reduced risk of treatment or prescription errors and lower risk of hospitalisation,^34,35 supporting the assertion that FDCs may contribute to reduced mortality rates. However, in the context of overall ART programme performance, it is important to note that LTFU appears to be increasing in more recent years, possibly because of the expansion of treatment programmes in the absence of sufficient resources.^26,36 It is important that overall programme performance is monitored in order to minimise LTFU and the knock-on effect that this could have on mortality.

To improve outcomes of the HIV programme, interventions should focus on clients most likely to present late for HIV care, including, as expected, clients with advanced clinical stage of HIV disease and those in hospitals. Interventions such as community-based strategies or home-based HIV testing should also target men,^37,38 in agreement with previous studies in South Africa and other low- and middle-income countries which demonstrate late presentation of men for both HIV diagnosis and treatment.^{39,40,41,42,43} This delayed health seeking may contribute to the growing gap between male and female life expectancy, which has doubled from 2003 to 2011.⁴⁴ Reasons for the disparity are likely to include reluctance of men to seek HIV care because of attitudinal barriers, HIV stigma, perceptions of masculinity and cultural practices,^45,46 as well as the scale-up of prevention of mother-to-child transmission (PMTCT) services,⁴⁷ which have increasingly engaged women in care. This study found that a higher proportion of pregnant women aged 15–50 years had baseline CD4 counts ≥ 200 cells/mm³ compared to women who were not pregnant during ART initiation, suggesting that PMTCT services play a role in linking women to timely HIV treatment. This is supported by findings of an association between PMTCT services and higher cohort median CD4 count, as well as reduced risk of ART initiation with advanced disease.^41,42,48 In addition, late presentation for ART initiation was found to be linked with older age in this study. Mixed findings regarding the association between age and late presentation for care have been reported in the literature, with one study finding no significant association,³⁹ others finding a decreased risk of late presentation for HIV care in older clients^41,42 and yet others demonstrating an increased risk of initiating ART with low CD4 in older clients.^25,40 The finding of this study that older age is linked to late presentation for care is supported by a recent study of South Africans aged over 50 years who reported seeking HIV testing only once they were symptomatic and referred by a provider or if a partner was diagnosed as infected.⁴⁹ Furthermore, older men in sub-Saharan Africa have been found to be less aware of and knowledgeable about HIV prevention compared to younger men, with increased risk behaviours in some settings.⁵⁰ Educational initiatives and HIV testing strategies that focus on men and the older population are thus required.

An additional intervention to improve outcomes of the HIV programme is delivery of CPT to high-risk clients presenting for HIV care with low baseline CD4 counts, as recommended by the WHO.⁵¹ It has been well established that CPT significantly reduces mortality in clients with low baseline CD4 counts,^52,53 and it is therefore concerning that delivery of this intervention appears to be declining over time, possibly reflecting competing demands on healthcare providers as the ART programme expands. Interestingly, a consistently higher proportion of clients with WHO stage III or IV disease received CPT compared to clients presenting with WHO stage I or II disease, indicating that healthcare providers are more likely to give CPT to clients who appear ill. This is concerning, as clients with immunosuppression who require CPT may not present with advanced disease, as seen in this and other studies,⁵⁴ and delivery of CPT may therefore be missed in these clients. This is likely exacerbated in the era of UTT, as baseline CD4 counts which would alert healthcare providers to the need for CPT are not required at the time of ART initiation; delivery of CPT therefore relies on the healthcare provider later checking the CD4 results and subsequently initiating CPT if required, increasing the chance that CPT may be overlooked. Interventions to increase awareness of the importance of baseline CD4 testing and delivery of CPT to clients with low baseline CD4 counts and/or advanced clinical disease are warranted.

The urban and rural districts analysed in this study had comparable findings in terms of overall trends in population CD4 counts, the proportion of clients initiating ART with CD4 < 200 cells/mm³ in 2017 and characteristics of clients with low baseline CD4 counts. However, mortality was markedly higher in rural Mopani than in Johannesburg. This highlights the inferior quality of care in the rural district, likely because of poor access to doctors as has been previously described in rural settings, as well as the deficit in specialist medical care in rural provinces of South Africa.^55,56,57 In addition, characteristics of rural settings, such as lower levels of formal education, are also known to be associated with higher mortality.⁵⁸ While urban settings need to be aware of the rate of programme expansion owing to the increased risk of attrition associated with rapid expansion of ART services,³⁶ efforts in rural settings need to focus on quality of care, perhaps by employing mobile units that are staffed by doctors or by providing training for primary healthcare staff.

This study demonstrated the rich source of data available in TIER.Net. The analysis was performed using data spanning over a decade during which there were substantial guideline changes in the ART programme, producing a robust and realistic analysis. On the other hand, this study has a number of limitations. Firstly, TIER.Net was introduced in South Africa in 2010, necessitating the back-capture of data from the ART programme prior to this time. There may have been challenges with the quality of the back-captured data, which may have led to an underestimation of the absolute numbers of clients on ART, including those who died, although we believe that the proportions are representative. Secondly, not all clients in TIER.Net could be included in the analysis because of missing baseline CD4 counts. Thirdly, the study was not designed to follow up clients who were LTFU from the ART programme and it is therefore unclear how many of these clients died. The Kaplan–Meier analyses may therefore underestimate mortality. Finally, although these data are representative of urban and rural areas in South Africa, they should be generalised to other settings with caution, particularly because only two districts were included in the analysis.

In conclusion, this study of routine programme data demonstrates rapid expansion of South Africa’s adult ART programme since 2004. Although median CD4 count has increased over time, this population measure is masking a high-risk group of clients who present late for HIV care with low baseline CD4 counts. Despite improvements in mortality over time, these clients are still at risk of increased mortality, particularly in rural settings. Therefore, targeted interventions are urgently warranted to improve early engagement in care and clinical management, with a specific focus on men, older clients and those presenting at hospitals. Ongoing monitoring of the routine programme coupled with targeted interventions will ensure continued improvements in the outcomes of these high-risk individuals.