In this multisite analysis of four large sub-Saharan African HIV programs, long-term immune response up to 6 years after ART start was higher in women compared with men,, independently of characteristics at initiation.
The gender-based difference in immune reconstitution described here has inconsistently been found in previous studies in resource-rich countries, some studies showing a better female immune reconstitution 
whereas two literature reviews indicated that most of the studies did not find any viroimmunological difference between men and women 
In resource-limited settings, studies showing worse male reconstitution are more common. In Cameroon and in China, two studies recently showed male poorer immune reconstitution even after initial patients characteristics were accounted for in the analysis. 
. In another large multicentric study in sub-Saharan Africa, a gender difference in immune reconstitution was found, but the difference was explained by baseline characteristics 
. In Senegal, a cohort study found a borderline gender effect which disappeared when viral load and baseline CD4 levels were integrated in the model. 
At last, a recent cohort study in Tanzania found gender differences in mortality, rates of loss of follow-up or side effects in men but failed to find a significant difference in immune reconstitution after 3 months of ART 
Our study confirmed that gender differences in immune reconstitution remained even when baseline factors are included the model. We also showed that this difference increased with time on ART up to 6 years after initiation. Furthermore, we quantified this difference to show that after 6 years on ART, it reached 140 CD4 cells/µL. When we estimated the time on ART required to reach 500 CD4 cells/µL, we found that men remained much longer under this threshold, known to be associated with higher morbidity 
and mortality 
Behavioural, biological factors or a combination of both could explain this difference. First, it is known that men living with HIV usually initiate ART at more advanced stages of disease, resulting in higher mortality 
highlighting gender differences in health seeking behaviours. Therefore, one possible reason for this gender difference in immune reconstitution could also be behavioural and related to a poorer adherence of men to ART 
. Decentralization of HIV care in sub-Saharan Africa has been implemented through primary health facilities, and traditionally such care has been well attended by women and children but less so by men. 
Interventions specifically addressing men's needs might be effective to improve treatment outcomes in this patient population. However, we could not verify this hypothesis as adherence and viral load were not routinely monitored in our cohorts.
Nevertheless, another explanation needs to be explored, reflecting a possible biological explanation. It is known that women have higher physiological CD4 levels compared to men 
and some studies also underlined the effect of male hormones on the thymic function 
. Therefore, this gender difference in immune reconstitution could also be linked to a higher potential for women to regenerate their CD4 lymphocytes stock on ART.
We lack of information on women previously exposed to “prevention to mother to child transmission” treatment. Previous exposure to ART (like single dose nevirapine) might underestimate the difference between men and women. The underlining hypothesis would be that women who took these prophylaxis would be more likely to be resistant to ART and thus to have a non optimal immune response. However, recent evidences seem to minimise this influence. 
Gender differences in long-term survival outcomes need to be explored and linked to this observed difference in immune reconstitution.
CD4 gains up to 6 years after ART start were similar regardless of the initial CD4 level. We demonstrated for the first time in sub-Saharan African countries that patients who started ART with CD4 count over 250 cells/µL achieved similar CD4 gains up to 6 years after ART initiation than those who initiated at lower CD4 levels. If the influence of initial CD4 level on future reconstitution was well known 
, such levels of reconstitution for patients who initiated ART at high CD4 levels had never been observed previously in sub-Saharan settings. This finding supports the theory that patients may have the same reconstitution potential regardless of baseline CD4. We need to evaluate if complete reconstitution could be feasible if patients were initiated ART at even earlier stage.
This study suggests that long-term CD4 recovery is independent of the level of CD4 cells at ART start, and patients who initiated ART at earlier stages of disease spend shorter time below the threshold of 500 CD4 cells/µL, which is associated with high morbidity 
and mortality 
. An earlier start of ART would decrease this high-risk period and facilitate the provision of care in areas suffering from human resources shortages.
Consistent with previous studies, advanced age at initiation was associated with poorer immune reconstitution 
. This might relate to greater difficulties to reactivate the thymic function and/or to poorer sustained adherence achieved by older patients 
We decided to study patient immune response after the first 9 months of ART because during the early stages of treatment, the release of pre-trapped CD4 cells could influence the observed response 
. The large number of patients included in this analysis allowed careful investigation and quantification of the dose-response effect of initial CD4 counts on immune reconstitution. Indeed, this study used data from four large operational HIV programs running for almost a decade, giving strength to our findings.
Nevertheless, given the definition of inclusion criteria, exclusion of patients because of incomplete initial characteristics is likely to have overestimated the extent of the overall immune reconstitution. Indeed, patients with advanced HIV disease have often been prescribed ART without having initial CD4 count data. Since these patients were likely to have low initial CD4 count values, missing values on initial CD4 count did not follow a random process. Imputation of these missing values was therefore not recommended. Cohort attrition might also have biased estimates. Patients who were included in the analysis were those who survived the first months of ART when mortality is higher. We analyzed routinely collected data without double entry, which could not be implemented for cohorts of this size, though consistency checklists were routinely used to try to reduce data entry errors.
Concerning the generalization of these findings, MSF support on these programs needs to be accounted, as these programs might differ from regular Ministry of Health programs.
In this analysis, we found that women treated for HIV experienced higher immune reconstitution than men, and patients who survived after the first 9 months on ART. Reasons for the gender effect should primarily be investigated through differences in health-seeking and adherence behaviour and addressed to improve outcomes among men.
The paper has not been published or submitted for publication elsewhere. This study was presented at the Conference on Retrovirus and Opportunistic Infections CROI in Boston (David Maman, Mar Pujades-Rodriguez, Fabien Subtil, Loretxu Pinoges, Megan McGuire, René Ecochard, Jean-François Etard. Gender Differences in Long-term Immune Response to ART and Mortality: A Cohort Analysis in 4 Sub-Saharan HIV Programs, CROI Conference, Boston, USA, February 27th–March 2nd 2011.).