We compared the CD4 cell count declines in untreated patients in a European and African setting, analyzing two cohorts from Cape Town, South Africa, and Switzerland. Both in South Africa and Switzerland Non-whites had slower CD4 declines than Whites, and older patients had faster declines. Furthermore, the CD4 decline was more rapid in patients with higher initial CD4 counts than in patients with lower counts.
We applied the same analytical approach to the data from the CTAC and SHCS cohorts and results are therefore directly comparable. We have taken account of the demographic characteristics in the cohorts by excluding injection drug users, adjusting models for age and sex. We focused on estimating short-term CD4 trajectories within four years of the first measurement to reduce bias due to slow progressors having more CD4 counts than fast progressors. Finally, we examined the effect of deaths on estimated CD4 decline and found that CD4 declines in the higher CD4 strata are over-estimated by the standard mixed-effects model. The joint model takes into account survival time and deaths and adjusts for the steeper decline of CD4 counts in very ill patients. A previous analysis of the CTAC data by Holmes and colleagues [6
] estimated the CD4 decline to be 47.1 cells/μL per annum for patients with initial CD4 counts above 500 cells/μL, 30.6 for those with CD4 counts between 351 and 500 cells/μL and 20.5 cells/μL in patients with counts between 201 and 350 cells/μL [6
]. Our estimates are somewhat higher. The different methodological approach may at least partially account for the difference between our estimates and those of Holmes and colleagues [6
Our study has a number of limitations. There were few white patients in CTAC and few non-white patients in SHCS. The study thus had limited power to examine whether the effect of ethnicity differed between the two settings. Also, follow-up was short for many patients in CTAC and therefore there are few CD4 measurements for most patients. A further problem with analyzing CD4 trajectories in seroprevalent cohorts is the lack of the time of infection, which would be the natural point in time to line up the trajectories. This is to some extent overcome by stratifying on initial CD4 count and reflects what the treating physicians see in practice and are interested in, namely estimates of short-term CD4 decline from the current value. The validity of using first CD4 count measurement as a surrogate for time from infection has been questioned in survival analyses from the Concerted Action on Seroconversion to AIDS and Death in Europe (CASCADE) study, with time to death as the outcome [16
], which have shown variation in CD4 set point associated with rate of subsequent decline in CD4. Our analyses, which examined CD4 decline according to initial CD4 strata, may thus be grouping together patients who have different lengths of time since infection. Furthermore, the mean time since infection for each initial CD4 strata could vary by cohort or ethnicity.
The data that are available from patients with well-documented seroconversion are limited, and particularly limited in resource-limited settings: a recent collaborative analysis of time to ART treatment eligibility was based on just over 2000 seroconverters from five cohorts from sub-Saharan Africa and Thailand [4
]. In contrast, the CASCADE collaboration of cohorts in Europe, Canada and Australia is based on over 17000 patients with documented seroconversion [16
]. There are also disadvantages to seroconverter cohorts, which are generally not representative of the HIV-infected population but include many patients infected through intravenous drug use or the transfusion of blood products. Also, independently of the route of transmission, patients whose seroconversion was documented because they experienced symptomatic illness are likely to have steeper CD4 declines and more rapid clinical progression compared with those who were asymptomatic [17
Other seroprevalent cohorts comparing white and black patients also found slower declines in the black group [19
]. A recent analysis of the SHCS showed similar differences in CD4 cell count declines between patients of African and European descent, but declines in general were less pronounced, probably because the analysis was restricted to patients in CDC clinical stage A with at least five CD4 cell counts and did not take into account informative censoring due to death [21
]. Interestingly, in CASCADE, non-white, compared with white ethnicity, was associated with higher odds of spontaneously achieving undetectable viremia and, in those who did control viral load, with a longer period of undetectable viremia [22
]. Analyses of seroprevalent cohorts such as the CTAC and SHCS are helpful to complement and extend the evidence that is available from seroconverter cohorts. In particular, our results are relevant when modeling time to ART eligibility and the need for ART at the population level, and estimating and projecting the future course of the AIDS epidemic [23
Whilst the decline in CD4 cell count is a determinant of the time from HIV-1 infection to AIDS or death, other factors include the level of CD4 before infection in the population and the rapid drop in CD4 in the weeks immediately after seroconversion. A study of CD4 count declines in seroincident and seroprevalent individuals in Tanzania showed that the median initial CD4 count of the seroincident individuals who seroconverted within 1 year of the first CD4 measurement was around 500 cells/μL whereas in non-infected individuals it was around 800 cells/μL, indicating that about 300 cells/μL were lost soon after seroconversion [24
]. Reference ranges for CD4 counts in people not infected by HIV vary across geographical regions, gender and age and racial groups [25
], as well as by lifestyle and biological factors such as smoking or contraceptive use [27
]. A study of Swiss blood donors found that the reference range was higher in women than men, and lower at older ages [28
]. African studies showed heterogeneity across populations, with, for example, markedly lower CD4 counts in Ethiopians [29
], and also found higher counts in women than men [25
The association between ethnicity and CD4 count declines may have several explanations. Ethnicity reflects social and economic position of participants within their respective cohorts. In CTAC, non-white participants tended to be of lower socioeconomic position than white participants due to historical limitations on socioeconomic opportunity. In the SHCS, the non-white participants are migrants from sub-Saharan Africa who are also generally of lower socioeconomic position than white participants. In untreated patients socioeconomic position may affect CD4 declines through increased exposure to opportunistic infections and resultant immune activation and by influencing access and adherence to prophylactic therapies and other relevant health behaviors. In the Swiss cohort, but not in South Africa, the slower decline of CD4 counts appears to have translated into better survival in Non-whites compared to Whites. Of note, once enrolled in the SHCS access to ART and prognosis of non-white participants is equivalent to white participants [14
]. Socioeconomic conditions, health-seeking behaviors, access to health care, and exposure to pathogens are more important determinants of mortality in South Africa [31
It is also possible that our results relate to host genetic differences: Non-whites may have adapted to frequent infectious diseases by selection over many generations for the ability to survive despite chronic immune activation [20
]. Of note, the low immune activation phenotype is also found in HIV-1 infected patients with slow disease progression in the Western world [32
], and in asymptomatic nonhuman African primates infected with Simian immunodeficiency viruses (SIVs) [33
]. A recent study suggested that genetically determined divergent Toll-like receptor signaling and interferon production distinguishes pathogenic (“immune activated”) from non-pathogenic infection in the animal model [34
]. In Caucasians genetic polymorphisms explained about 15% of the variation in viral load set points during the asymptomatic period of infection [35
]. Of note, a recent analysis of the SHCS found that the CD4 cell count decline was less steep in patients of African descent compared to patients of European descent independently of whether patients were infected with HIV-1 subtype B or subtype C [21
]. The slower CD4 cell count decline in patients of African descent is therefore unlikely to be due to infection with less virulent subtypes.
In conclusion, further studies on reference ranges of CD4 counts and on rates of declines in CD4 counts, in different countries and populations are needed to inform the development of guidelines for when to start ART, and to improve projections of the epidemic, particularly in resource-limited settings. The methodology used in the present study, which addressed a number of issues not generally considered in previous studies, might serve as a model for future studies.