Reply to Okatch et al

With the recent publication of landmark studies supporting immediate antiretroviral therapy (ART) initiation for persons infected with human immunodeficiency virus (HIV), the World Health Organization has recommended ART for all [1]. Although these recommendations are intended to improve health outcomes and decrease transmission, their short-term impact could be challenged by the ability of HIV programs to identify and refer patients to care at early stages of disease. We based this conclusion on a recent meta-analysis evaluating all studies reporting such data in the published literature [2]. We found no change in CD4 count at presentation to care or ART initiation from 2002 to 2013. In 2013, we estimated the CD4 count at presentation and ART initiation in sub-Saharan Africa were 309 cells/µL and 140 cells/µL, respectively, leading us to conclude that much work remains to improve identification and prompt linkage to care of HIV-infected individuals.

Okatch et al now provide novel data demonstrating that the mean CD4 count at ART initiation increased by approximately 26 cells/µL per year during 2009–2013 in Botswana [3]. Their results are cause for optimism that improvements in HIV care delivery are possible in countries that dedicate resources to the HIV epidemic. Botswana more than doubled its investments in HIV control between 2006 and 2011, and is now one of 2 countries (with South Africa) that relies on international support for <25% of total funding [4].

As discussed in responses to our article [3, 5], one potential reason for differences between our estimates and those by Okatch et al is that meta-analyses aggregate data from multiyear cohorts into single estimates, and can be subject to ecological fallacy, such that within-cohort trends become obfuscated by summary estimates across multiple years of data. To account for this potential bias, we conducted 2 sensitivity analyses. First, we restricted our sample to studies with durations of ≤3 years to minimize masking of within-cohort trends from studies of longer duration. Second, we conducted an analysis including disaggregated, annual CD4 data, when available. Neither of these sensitivity analyses yielded estimates that differed meaningfully from the primary analysis (Table 1). Moreover, assuming that these sensitivity analyses were insufficient to account for positive within-cohort trends, our analysis still would have identified positive trends over time between cohorts, had they existed. Our data therefore offer strong evidence that CD4 count trends at presentation and ART initiation have been stagnant over the past decade.

Table 1.

Trends in CD4 Count and Presentation to Care and Antiretroviral Therapy Initiation in Sub-Saharan Africa, 2002–2013

Sample	Presentation to Care					Initiation of Antiretroviral Therapy
	Studies, No.	Patients, No.	β Coefficient	95% CI	P Value	Studies, No.	Patients, No.	β Coefficient	95% CI	P Value
Total sample	56	295 455	5.84	−10.7 to 22.4	.48	71	549 702	−1.13	−8.4 to 6.2	.76
Excluding studies with >3 y of data	43	42 890	1.38	−18.3 to 21.1	.89	41	65 027	5.50	−1.6 to 12.6	.13
Disaggregated by year	56	295 455	6.52	−7.3 to 20.4	.35	71	549 702	3.91	−.7 to 8.5	.10

Abbreviation: CI, confidence interval.

The findings of Okatch et al support our contention that the HIV care community has a long way to go to fully realize the clinical and public health benefits of ART. Their analysis notes important improvements in linking patients to ART in Botswana. Yet, at the conclusion of 2013, the median CD4 count at ART initiation remained 238 cells/µL, well below international goals. While the global health community debates changing the treatment threshold from 500 cells/µL to abandoning immunological criteria entirely, the vast majority of patients are linking to care and initiating ART late in their disease course. Although changing guidelines to promote earlier treatment initiation is important, resources must be dedicated to earlier identification and linkage to care. Otherwise, the current guidelines will continue to offer too little, and much too late.