CD4 Cell Count Distributions and HIV‐Associated Mortality in Africa

To the Editor—We read with great interest the article by Williams et al. [1] in which they reported the development of a dynamic model of CD4 cell count decline in HIV-infected people living in Africa. A CD4 cell count model in this context is greatly needed, because it could be used to estimate the incidence of tuberculosis and other opportunistic infections, predict life expectancy, and project the need for antiretroviral treatment (ART) and patterns of health care utilization. The inputs to this model were carefully drawn from existing data on CD4 cell count distributions in African populations, both among HIV-positive and HIV-negative adults. The best-fitting model was used to compare the survival of HIV-infected people in 2 African countries. This model predicted that HIV-infected people with CD4 cell counts of 200 cells/μL would have a substantially shorter survival time in South Africa than in Zambia. An important assumption of this model is that death occurs in HIV-infected individuals at a CD4 cell count of 50 cells/μL. This assumption conforms to the long-held view that HIV-associated deaths occur with very advanced immunosuppression and do not typically occur in patients with higher CD4 cell counts or before the development of AIDS [2].Williams et al. based their assumption on data from 2 cross-sectional surveys of hospital inpatients in Abidjan, Côte D'Ivoire, and on data from an HIV seroconversion cohort in rural Uganda [3–5]. In the latter study, the median CD4 cell count before death (n = 68) was 61 cells/μL. However, the CD4 cell counts were broadly distributed, with a 75th percentile of 199 cells/μL. Moreover, just 61% of deaths occurred in patients who had received a diagnosis of AIDS at their clinic visit within the preceding 3 months [5]. We have recently reported data from a South African natural history cohort of HIV-infected people (n = 1399) [6]. Although these data showed that risk of mortality was greatest for those with the lowest CD4 cell counts and those with World Health Organization stage 4 disease (AIDS), the 6-month risk of mortality was, nevertheless, appreciable even in patients with CD4 cell counts >200 cells/μL. Moreover, 52% of deaths occurred in patients who had not previously received an AIDS diagnosis [6]. This challenges the existing dogma that death occurs only with advanced immunosuppression. Recent data from the SMART trial, a large, international, multicenter study of 5472 patients, also challenge this notion [7]. HIV-infected patients receiving ART were randomized to receive either continuous treatment or drug-conserving treatment using structured interruptions. Interruptions were initiated when CD4 cell counts were >350 cells/μL, and treatments were restarted when counts decreased to <250 cells/μL. This trial was discontinued prematurely at the interim analysis of data, the vast majority of which was from patients living in high-income countries. The unexpected finding was that patients randomized to the treatment-interruption strategy had an excess mortality risk, despite CD4 cell counts being maintained within what might have been perceived as a “safe” range [7]. These data again illustrate the fact that HIV-associatedmortality does occur at high CD4 cell counts. In resource-limited settings, a high prevalence of other important HIV-associated but non—AIDS-defining infections (such as malaria [8]), plus limitations in health care, are likely to further increase the potential for mortality among those with well-preserved CD4 cell counts. The data from these 2 studies may have important implications for mathematical models of HIV infection in Africa, such as the CD4 cell count model described by Williams et al. [1]. First, predictions of survival must take into account the mortality that occurs at higher CD4 cell counts. Second, it should be recognized that substantial mortality occurs before the development of AIDS. Third, because the rate of CD4 cell count decline is extremely heterogeneous, non-AIDS mortality may be greatest for those with the slowest rates of CD4 cell count decline. Although it would considerably increase the complexity of the modeling, we suggest that future models should take these observations into account. We thank Dr. John Beslile and Dr. Karen Dobos, Colorado State University (National Institute of Allergy and Infectious Diseases, National Institutes of Health, Tuberculosis Research Materials contract N01-AI-40091) for kindly supplying Mycobacterium tuberculosis-related reagents.