In this multicentered cohort study conducted in West Africa, we described a large sample of HIV-infected adolescents who started ART at a median age of 10.4 years. Despite this information was not documented with reliability in this cohort, it is estimated that more than 90% of the children/adolescents acquired HIV through mother-to-child transmission and can be considered as slow progressors 
. We observed that about two thirds of HIV-infected adolescents on ART beyond the age of ten were not aware of their own HIV status. When the age at disclosure was known, it was delayed around a median of 15 years. We demonstrated a beneficial effect of HIV disclosure on retention in care among adolescents on ART.
Although low, the frequency of disclosed HIV status in our study was higher than previously reported figures in studies conducted in Nigeria, Ethiopia, Thailand and India where the children were younger (mean age
8–9 years) 
or Ghana, with a median age of ten 
. It was similar to other studies conducted in Uganda and Zambia in children aged 12 in median 
or Thailand (mean age 9.2 years) 
but a bit lower than an Ivorian study involving adolescents older than 13 years 
. Indeed, disclosure was often associated with an age>10 or even older 
, supporting our observations. In a qualitative study conducted in South Africa, the age of ten has been reported by health workers as the most appropriate to start having discussion regarding HIV infection 
In a US perinatally HIV-infected cohort, HIV disclosure was observed to occur at younger age over time, which may suggest a decline in the perception of HIV stigma 
. In this industrialized country, the social and medical network may have been strengthened, leading to this condition. In resource-limited settings, despite local interventions, such as peer support groups, or community organizations for social support, stigma and fear of negative reactions or psychosocial outcomes may remain strong barriers to an earlier HIV disclosure to children 
Conditions of disclosure varied according to settings. In Abidjan, relatives were encouraged to conduct the disclosure while the medico-social staff was more involved in the other countries. In Nigeria, most of the disclosure was conducted by family, preferably the mother, at home 
. In Ghana, it was also mostly conducted by caregivers 
, but among those who had not disclosed, one-third wanted to defer to the health workers. Similarly, in Ethiopia, 60% of 193 caregivers interviewed thought that the doctor was responsible for disclosure 
and in Thailand, 50% of the interviewed caregivers reported the need of assistance from health workers 
. These findings were supported by a study on health workers' perceptions conducted in a South African setting 
. In this present study, the sites implemented different strategies to conduct the disclosure process (described in the Methods
), which could explain, in part, the differences observed between across the cities in terms of retention. Also, the investigators of the participating centers were aware of the recommendations from the French-speaking association called Grandir 
. However, they faced difficulties not addressed by these recommendations. For instance, they had problems to manage the caregivers' reluctance to disclose or to make understand to the adolescent what HIV is and its impact on life. They required sharing field experiences to harmonize process and guide local policy on HIV disclosure to be conducted in optimal conditions involving both adolescents and their caregivers. Recent guidelines proposed by the World Health Organization underlined the absence of evidence as to who best can disclose to the child his/her HIV status, caregiver or health care worker with or without specific training 
In this large cohort with three-year follow-up, we observed a relatively low death rate but a quite important loss-to-follow up rate, similar to a previous report on clinical outcomes in West African children 
. In a study conducted in Uganda, including 575 patients starting ART during adolescence, the cumulative survival at 36 months was 90% (95% CI 87.9–93.1) 
. This figure was similar to the survival rate found in our study where 55% has initiated ART after the age of ten. We found that HIV disclosure was associated with better retention in care as already reported in a Romanian setting among adolescents with a mean age of 13 years 
Several limitations could be discussed in our study. First, the main limit of this study is its partially retrospective design and the fact that it relied heavily on the patients' charts, leading to missing data and possible information bias. In particular, the HIV disclosure status was not systematically reported in the patients' charts. However, the forms were filled in by clinic staff, who knew the patients and their family environment quite well and could remember and cross-check who was informed of his/her HIV status and who was not. Some charts could not be found or used and were classified as missing data and the corresponding adolescents were not included in the analyses. Sensitivity analyses, coding these records either as disclosed HIV status or not disclosed did not lead to a large variation in our findings.
Second, the HIV disclosure status was reported as a dichotomic variable, while disclosure is a process evolving over time. Thus some adolescents may have been categorized as not informed of their HIV status while they may have been told that they harbour a virus, but not specifying HIV. Some studies, preferably those with a qualitative design, specify that some children are partially disclosed. It refers to the following broad category: “child not fully aware of his/her HIV disease but is suspicious, asks questions to the caregiver about the disease and the drug, and, in many cases, assumes that the drug is a cure” 
. This status could not have been taken into account in our study with no qualitative data recorded. In addition, the age at disclosure was frequently missing and it was not possible to study this variable as a time-dependant information in the survival analyses.
Third, some of the adolescent lost-to-follow-up might have been unreported deaths. We have addressed this possible misclassification by specifically retrieving and reviewing every patient's chart meeting the definition of loss-to-follow-up in case the charts had been completed after the closure of the database for this analysis.
Finally, due to the retrospective design, we could not address the psychosocial effect of disclosure. A cross-sectional study conducted in New York City demonstrated that Youths (n
196, mean age 12.7 years) with disclosed HIV status were significantly less anxious than those who had not been told but there were no other differences in psychological functioning 
. This should be verified in African settings as the caregivers' fear for negative psychosocial outcomes is a common barrier to disclose as discussed previously.
In conclusion, most of HIV-infected adolescents on ART in these West African settings were not aware of their HIV status. However, our study showed a strong beneficial effect of HIV disclosure on retention in care after ART initiation beyond the age of ten. This sample of clinics had initiated different strategies to carry out the HIV disclosure process in the absence of guidelines or specific training. Such initiatives need to be promoted and developed as they may provide individual benefits. For this purpose, they need to be described, standardized, evaluated and shared. Also, further studies should look at the effect on other outcomes such as immunological failure, treatment adherence, virological progression, viral resistance, but also anxiety, depression, school performance, family and social relations and sexual risk behaviours. This would provide deeper understanding of HIV disclosure process in adolescents in resource limited settings in order to tailor age-adequate interventions.