In this observational study, we found evidence not only of earlier presentation for ART but also a substantial reduction in early mortality on ART subsequent to the change in CD4+
cell count eligibility criteria that took effect in South Africa in August 2011. Median CD4+
cell count at ART initiation remained higher and mortality remained lower for women during all time periods, consistent with the data from this region.22,23
However, it is encouraging that the increase in CD4+
cell count and reduction in early mortality for those who initiated ART in the last time period (2011–2012) were consistent for both the nonpregnant females and males. Pregnant women only represented about 7% of all adults who initiated ART and there were very few deaths in this group; so there was limited statistical power to observe any temporal trends. These data provide further encouragement at a time when substantial reductions in population-level mortality have been documented in this community and more broadly in South Africa.24–27
We have previously shown that during the first phase of rapid programme scale-up, there was no significant change in CD4+
cell count at ART initiation.4
It seems that after the first policy change in 2010 to recommend ART at <350 cells per microliter for TB patients and pregnant females, there was minimal change in the overall median CD4+
cell count at ART initiation, yet after the expansion to <350 cells per microliter for all, there has been a definite increase. This suggests that, although antenatal care and TB services remain important entry points to HIV treatment, the contribution from these groups alone was too small to see a broader impact of the expanded eligibility criteria. The increase in overall median CD4+
cell count between 2010–2011 and 2011–2012 was 54 cells per microliter, which corresponds approximately to 6 months earlier in terms of disease progression according to our previous data.17
The relationship between mortality risk and time period was maintained after adjustment for baseline CD4+
cell count, consistent with the fact that CD4+
cell count does not fully predict mortality,28
and that other factors associated with earlier presentation, such as the absence of opportunistic infection, may have contributed to the mortality decline. It is noteworthy that in the last time period, there was a decline in the absolute number and the proportion of people who initiated ART with CD4+
cell count <50 cells per microliter, as this remains the factor most strongly associated with early mortality, with a 4-fold increased risk of early mortality compared with CD4+
cell count of 201–350 cells per microliter.
These observations provide robust evidence that the reduction in early mortality has occurred since the policy change to recommend earlier initiation of ART, although a direct causal association cannot be inferred from these observational data. In common with data from other large programmes in South Africa, early mortality did not change in the first few years of programme scale-up.29–31
Only in the last time period (2011–2012), there was a real decline in early mortality. There have been no specific interventions or programmatic changes that could easily explain the decline in early mortality. There is no formal package of care for individuals before eligibility for ART and retention in this group is suboptimal.32
We have previously shown that TB is the most common cause of early mortality on ART and that prevalent TB at time of ART initiation is associated with almost double the risk of mortality on ART, in line with other studies in the region.32–34
In this analysis, there was no independent association between TB and early mortality, although the statistical power to observe an association might have been affected by the lower risk of TB disease at higher CD4+
The implementation of isoniazid preventive therapy was scaled up from early 2010, yet provision remains patchy in this programme. It could be that improved integration of ART with TB treatment has contributed to the reduction in mortality, in line with findings from clinical trials.36,37
Alternatively, although there was no specific programmatic change, it could be related to clinic systems becoming more efficient over time, increased use of nurse-initiated ART, or by the shorter time needed for pre-ART work-up for people with higher CD4+
cell counts. Finally, given the change in the first-line antiretroviral regimen in 2010 from a stavudine-based to tenofovir-based regimen, we cannot discount that this may have accounted for some of the mortality change, although against this would be the fact that the majority of fatal adverse events related to stavudine (eg, lactic acidosis) occur after the first 3 months.38
Earlier presentation for treatment may be not only indirectly related to the policy changes but also to improved HIV testing coverage in the community, facilitated by the nationwide HIV counseling and testing (HCT) campaign and local initiatives to scale up home- and community-based HIV counseling and testing.39
This is supported by the fact that the CD4+
cell count at enrollment in the programme also shifted substantially in the period since August 2011, to the extent that almost half of the adults who enrolled into care between August 2011 and July 2012 had a CD4+
cell count >350 cells per microliter. Policy changes alone have no impact unless the health system is able to provide the services and individuals are able and willing to access those services. Our data are, therefore, important to show not only that people are accessing treatment earlier but that, so far, the health system has been able to cope with the increasing patient load. Not all countries are following WHO recommendations with regard to ART eligibility criteria, and these data should contribute to the evidence base around the impact of earlier access to ART.40
As early LTF increased over time, mortality may have been underestimated, especially in the last time period because of the shorter time for deaths to be captured. The correction of mortality for LTF is complicated in our programme as, before analysis, some correction occurs routinely through linkage to the hospital information system and demographic surveillance system. Under a worst-case but unrealistic scenario, where the proportion of LTF in the last time period attributable to mortality was 40%, the estimated mortality decline was much less marked, but under a more realistic assumption of 20%, the decline remained significant.
The strengths of this analysis relate to the simplicity and quality of the data from a well-characterized treatment programme. Only 6% of adults who initiated ART had a missing baseline CD4+
cell count (considerably lower than the 25% reported from a multicohort collaboration in South Africa41
), thus minimizing potential bias. Nevertheless, interpretation of the findings should be subject to some caution. We were unable within this analysis to explore outcomes for those people who, while eligible, did not initiate ART. We cannot, therefore, exclude the possibility that unmeasured variations over time in mortality for those who did not access ART could offset the mortality decline observed for those who access ART, although the consistent decline in population-level mortality would argue against this.27
Data regarding in-hospital deaths from the Hospital Information System were only available from January 2011 and in-hospital deaths during earlier time periods might have been missed if they had not been reported through the routine systems of the HIV treatment and care programme. However, this would be expected to bias the findings away from the null hypothesis and therefore would strengthen our findings. Although the findings should be broadly generalizable to public-sector programmes in South Africa, it is possible that not only extra financial support to the programme but also the research and community engagement activities in the area might influence access to and uptake of treatment and replication of this analysis should be performed with other programmes.
In conclusion, we have demonstrated that in this rural area, since the policy shift to recommend ART for all adults with CD4+ cell count <350 cells per microliter, there has been a significant increase in the median baseline CD4+ cell count and a significant reduction in mortality within the first 3 months of ART. These findings provide support for national and international ART policies and highlight the importance of earlier ART initiation for achieving reductions in HIV-related mortality.