Risk factors for poor virological outcome at 12 months in a workplace-based antiretroviral therapy programme in South Africa: A cohort study

An unprecedented effort by global organisations, governments and health care workers has achieved access to antiretroviral treatment (ART) for 2.0 million HIV-infected individuals in low- and middle-income countries by December 2006. Despite these efforts, 72% of the estimated 7.1 million who need ART globally are not on treatment, and hence strategies to improve access to care for HIV-infected persons have been the subject of much discussion [1, 2]. As more patients start ART, treatment programmes face the emerging challenge of achieving and maintaining virological suppression in increasing numbers of patients for sustained periods. Treatment success requires excellent adherence to ART, which is difficult to maintain [3]. A suboptimal virological response to ART may lead to increased transmission of ART-resistant virus strains and ultimately to a dramatic reduction of the therapeutic lifespan of available ART regimens [4–6].

ART programmes in developing countries have reported varied treatment outcomes. Pilot projects established by non-governmental organisations in South Africa and Haiti have reported good virological results [7–9] although others have been less encouraging [10–12]. Adherence to ART is a major determinant of virological outcome [3, 13, 14] and in industrialized countries other baseline factors influencing virological response have included disease stage [15], viral load [15, 16], and previous exposure to ART[15], although associations with baseline CD4 count and viral load are not consistent [17, 18]. Determinants of virological outcomes, and the effect of differences in the health delivery system on outcomes, have not been studied in detail in resource limited settings. Understanding these issues is important to guide the development of effective ART programmes.

Another obstacle to effective HIV care in resource-limited countries is limited access to laboratory facilities to monitor the virological and immunological markers that are considered the standard of care in wealthy nations. For middle-income settings, where there is limited laboratory support, data are needed to inform guidelines on how best to use scarce resources for laboratory monitoring; for example to decide at what point after starting ART it is most useful to measure viral load to ensure the best outcomes for patients.

We report risk factors for suboptimal virological outcome at 12 months among HIV-infected adults receiving ART in a workplace-based HIV care programme in South Africa.

A total of 1760 individuals had started ART before 1st July 2004, satisfied the other inclusion and exclusion criteria and were therefore eligible for the analysis. The median age was 41 years (range 21–66 years) and 97% (1701/1760) were male, reflecting the gender distribution of the workforce. Of the 1760 individuals, 175 (9.9%) and 213 (12.1%) had died or left the workforce before the 12 month follow-up, respectively, and 200 (11.4%) stopped ART before 12 months (Figure 1). Of the remaining 1172 individuals, 81.3% (953/1172) either had a viral load measurement at 12 months or started second line therapy prior to the 12 month visit, and data are subsequently reported for these individuals. Of those in the programme at 12 months, baseline CD4 counts, viral load and 6 week viral loads were similar in the 219 individuals who did not have viral load measured at 12 months compared with the 953 who did have viral load measured at 12 months. The median time from initiation of ART to the 6 week HIV-1 RNA measurement was 7.1 weeks (inter-quartile range [IQR]: 6.3–8.9). For the 6 and 12 month HIV-1 RNA measurements the median times from initiation of ART were 5.9 (IQR: 5.6–6.3) and 12.0 (IQR: 11.4–12.6) months, respectively.

The multivariable analysis is reported in Table 2. Virological response and self-reported adherence at six weeks were included in the same model as the data suggests these variables may capture different information (Table 3). Amongst n = 719 patients with both viral load and adherence measured at 6 weeks 18% had differing results; 10% (72/719) reported 100% adherence but had <1 log decrease in viral load, whereas 8% (60/719) reported <100% adherence but had satisfactory virological response. Site of ART delivery, WHO stage, viral load at baseline, unsatisfactory decrease in viral load at six weeks and less than 100% self-reported adherence at six weeks were all independent risk factors for suboptimal virological outcome.

The earliest reports from ART programmes in low-income countries were from programmes managed by non-governmental organizations. In their early phases when ART availability was limited, some programmes had stringent entry criteria, requiring good adherence to visits and to preventive therapy before ART was started [7]. These entry criteria are likely to have resulted in a clinic population pre-selected for good adherence. An unintended additional consequence may have been very high mortality among individuals waiting to start treatment [23]. In our programme the criteria for offering ART were medical need, with no prerequisite to demonstrate adherence, making it more similar to routine programmes in industrialized countries. Virological outcomes in those who had a 12 month blood sample taken were similar to those reported from clinic cohorts in the United States [24], but losses from our programme are higher than reported from such cohorts, and this is a limitation of our study.

We can distinguish two main reasons for losses from our programme; individuals who have left the workforce, which may not relate to virological outcome, and those remaining in the workforce who have stopped ART, who may be assumed to have treatment failure. The reasons for default have been briefly reported elsewhere [25] and are currently being investigated in detail. From patient interview the reasons cited for leaving the programme included use of traditional medicines, side effects and not being convinced of the benefits of treatment. A systematic review of patient retention in ART programmes in sub-Saharan Africa found that the percentage of patients retained at 12 months from initiation of ART (where retention was defined as patients known to be alive and receiving ART) varied between 49%–90.3% (median 76%) [26]. Based on this definition patient retention in our study was comparable at 67%.

As ART roll out progresses, retention within programmes will become a key issue determining long-term success. Among patients in our programme who had a viral load measured, 71.5% had suppressed viral load at 12 months following initiation of ART. Using the same analysis approach other South African studies have reported better outcomes [7, 9] whilst a small study in Senegal reported 61.9% with viral load < 500 copies/ml at 12 months [10].

We found that high baseline viral load and more advanced WHO stage were risk factors for a suboptimal virological outcome at 12 months. A fall in viral load of less than one log at 6 weeks predicted a suboptimal virological outcome at 12 months, consistent with reports from industrialized countries [15, 16, 27, 28], as did reported less than perfect adherence at six weeks. In our setting there was very little ART use prior to the start of this cohort and hence transmitted resistance among our participants is unlikely; poor virological response is most likely to be due to poor adherence. Suboptimal adherence should be identified as early as possible, so that additional support can be provided, aiming to minimize the risk of ART resistance developing. In resource-limited settings, if it is possible to measure baseline viral load, an early measure of viral load on treatment (for example at four to six weeks) would be arguably more valuable than at the six month visit as suggested, for example, in South African guidelines [29]. Self-reported adherence was also a strong predictor of outcome at 12 months which may be useful if viral load measurements cannot be done.

Our programme provides an unusual opportunity to investigate the influence of health facility level, as well as individual level, factors on ART outcomes. Site was an important determinant of virological outcome which persisted after adjustment for individual factors. Given that all sites used the same clinical guidelines, criteria for starting ART, counseling protocols and drug regimens, it is interesting that site had such a large effect on the probability of achieving virological suppression. With relatively few large sites to compare we cannot formally investigate at the health facility level which may influence virological outcomes. However our experience suggests some factors which may be important include long waiting times for clients (both in clinics and in pharmacy); using staff who do not have specialized training in HIV care; failure to contact individuals who do not attend clinic and lack of communication between pharmacy and clinic staff if ART is not collected. We plan to explore these issues in further work, including comparison of costs at different clinic sites.

In conclusion, in this cohort of adults starting ART in South Africa, the strongest independent predictors of virological outcome at 12 months were the decrease in viral load and self-reported adherence at six weeks. In settings with limited access to viral load measurements, the virological response would be more usefully determined at this point rather than later, so that appropriate interventions can be implemented. Self-reported adherence at six weeks was also an independent predictor of outcome. Health system factors have an important influence on virological outcomes; this needs further investigation to guide implementation of effective ART programmes.