To our knowledge this is the first analysis to examine the determinants of CD4 cell count recovery among patients receiving ART in resource limited settings. These data indicate that those with baseline CD4 cell counts <50 cells/μl had similar rates of phase 1 CD4 recovery (0–16 weeks) and greater rates of phase 2 recovery (16–48 weeks) compared to rates among those with higher baseline CD4 cell counts. Moreover, those with the lowest baseline counts were least likely to be immunological non-responders to ART. Despite these observations, those with baseline CD4 cell counts <50 cells/μl were nevertheless least likely to attain a CD4 cell count 200 cells/μl at 48 weeks. Taken together, these results suggest that although patients with very low baseline CD4 cell counts retain capacity for similar or slightly greater rates of CD4 cell count recovery compared to those with higher counts, they are nevertheless likely to require a longer period of time to attain a CD4 cell count threshold of 200 cells/μl. Thus, a prolonged period below a 'safe' CD4 cell count threshold rather than a diminished rate of immunological recovery is likely to underlie the high rates of morbidity and mortality observed among those with advanced disease during the early months of ART [4, 5, 8].
The findings of this analysis are strengthened by the relatively homogeneous study population receiving treatment at a single facility using standardised clinical protocols. Patients were all ART-naïve and received a standard triple-drug regimen with uniform follow-up time points. Quality-assured laboratory assays were all performed in a single nationally accredited laboratory. In contrast, previous studies of the determinants of CD4 cell count recovery have examined heterogeneous study populations in multiple centres and used diverse treatment regimens. Moreover, these studies included many patients with prior ART exposure [20–22] and some included only those who maintained suppression of plasma HIV load [9, 23, 24]. Our patient population was treated under the government ART roll-out programme and data are therefore likely to be generalisable to other ART programmes in sub-Saharan Africa. Our study is limited to analysis of CD4 cell recovery during the first year of ART and long-term outcomes and their sustainability remain to be determined. Moreover, in the present analysis we have examined recovery of CD4 cell counts but not CD4 cell functional responses.
The immunological response to ART among those with low CD4 cell counts was excellent with the proportion of patients with a CD4 cell count <100 cells/μl decreasing from 51% at baseline to just 4% at 48 weeks. However, our most important finding was that in multivariate analysis baseline CD4 cell counts <50 cells/μl were independently associated with similar rates of phase 1 CD4 cell recovery and greater rates of phase 2 CD4 cell recovery compared to individuals with higher baseline CD4 cell counts. This has not been clearly highlighted in previous publications from Europe and North America although comparison of our data with previous studies is difficult in view of differing cohort compositions. However, this overall observation is consistent with the findings of Bennett et al., and Le Moing et al. showed a similar but non-significant trend when comparing patients with baseline CD4 cell counts <200 cells/μl with those with higher counts . Kaufmann et al. found that CD4 cell increases in the first year of ART were similar comparing those with baseline counts <100 cells/μl with those with counts 100–199 cells/μl .
Survival bias could potentially have affected our findings. We have previously shown in this cohort that patients with the lowest baseline CD4 cell counts had a higher risk of death  and immunological non-responders or poor responders may have a greater mortality risk, leading to a survivor effect. However, we were careful to ensure that rates of CD4 cell recovery in the CD4 cell strata did not differ when comparing analyses of the whole cohort with those for whom data points were available at every time-point. Secondly, WHO clinical stage of disease is the strongest predictor of death in this cohort  and yet this variable was not associated with CD4 cell responses. Thirdly, the majority of deaths occurred in the first few weeks of ART among those whose disease was simply too far advanced; such deaths probably do not reflect a lack of immunological response to ART. Finally, we have previously reported that over 20% of early deaths in this cohort are due to immune reconstitution disease [5, 25]; such deaths typically occur among those with low baseline CD4 cell counts and good CD4 cell recovery. These deaths would tend to cause exactly the opposite bias. Thus, although an important consideration, we do not think that survival bias had an overall dominant effect.
We have previously shown that low baseline CD4 count at entry to an ART programme was associated with increased risks of tuberculosis and of mortality during the first year of ART [5, 8]. Results from the present study suggest that these increased risks are likely to reflect an increased period of time required for such patients to achieve a 'safe' level of immune function rather than reflecting impaired rates of immune recovery and this is consistent with the findings of a study from Spain with longer term follow-up . More recently we have found that risk of mortality and risk of incident tuberculosis is strongly associated with the current CD4 cell count during ART rather than the baseline count (unpublished data). Thus, if patients with profound baseline CD4 lymphocytopenia survive the initial few months of treatment and achieve full viral load suppression, then high rates of immune recovery are likely to result in a better prognosis that might have been anticipated.
Rates of phase 1 and phase 2 CD4 cell increase were similar in magnitude to those previously reported from high-income countries [9, 20]. The rapid phase 1 CD4 cell recovery was strongly associated with baseline viral load as described previously [20, 21]; patients with viral loads >105 log10 copies/ml had 11-fold greater CD4 count increases than those with lower viral loads. Immune dysregulation and immune cell activation promote sequestration of CD4+CD45Ro+ memory T cells in lymphoid tissue; suppression of viral replication then triggers rapid redistribution of this cell pool and a reduction in apoptotic cell death during the initial weeks of ART [26, 27]. A positive correlation between the plasma viral load and the degree of cell sequestration may provide a possible mechanism underlying the observation that those with the highest viral loads have the greatest initial CD4 cell increment. The fact that patients with profound CD4 lymphocytopenia have good immunological recovery during ART is likely to be important in the pathogenesis of immune reconstitution disease associated with Mycobacterium tuberculosis, Cryptococcus neoformans that is frequently observed in this patient population.
Greater phase 2 CD4 cell recovery was strongly associated with age as reported elsewhere [9, 24]. Sustained suppression of viral replication is also critical to second phase recovery [20, 21] and we confirmed that viral load at 16 weeks was a strong independent predictor of subsequent immunological recovery. Ten per cent of patients were immunological non-responders (CD4 cell increment <50 cells/μl at 48 weeks) and 7% of patients had discordant responses, having immunological non-response despite a fully suppressed viral load. These rates are much lower than those reported in previous series from industrialised countries [14, 29], possibly because our study only included antiretroviral-naïve patients, rates of HIV primary drug resistance in this setting are likely to be low, and because rates of treatment adherence in our cohort are very high. However, those with CD4 cell counts <50 cells/μl were least likely to be immunological non-responders. Moreover, 19% of 'super-responders' had baseline CD4 cell counts <50 cells/μl, indicating that a very low baseline CD4 cell count does not preclude an excellent CD4 cell count response to ART.