We compared pooled data collected over two successive west-African PMTCT cohorts conducted in a similar environment to assess the field effectiveness of a specific paediatric set of interventions combining early HIV-diagnosis in all HIV-exposed newborns from age 6-week and cotrimoxazole prophylaxis targeted to those HIV-infected. Early virological HIV-diagnosis was routinely feasible in Ditrame-Plus, and despite the systematic daily cotrimoxozole prophylaxis, the occurrence of severe events remained extremely high reaching 66% at 18 months of age. These residual severe events remained also mainly attributable to infectious causes. After adjustment on known determinants of child morbidity-mortality, the daily cotrimoxazole provided to HIV-infected children was well tolerated and reduced by 45%, although not significantly, the occurrence of severe morbidity-mortality when compared to the control group. The effectiveness of cotrimoxazole prophylaxis for optimizing early case management of HIV-infected infants and reducing severe morbidity-mortality within their first year of life was not statistically significant within our study. In fact, the statistical power estimated to show an existing difference between the two cohorts was estimated to be only 11%.
As the access to care was different between the two cohorts with no day-care hospital in Ditrame while this existed in the Ditrame Plus cohort,.some of the Ditrame plus children were first managed at the day-care hospital before going back to home on the same day or to be hospitalised at the pediatric unit if overnight care was needed. To get the most comparable definition of hospitalisation between the two cohorts, we defined hospitalisation as a hospitalisation > one day, at the pediatric unit of the Yopougon Hospital for both cohorts. Consequently, we feel that we have minimized the potential information bias in documenting comparably the risk of severe event.
The 6-week MTCT rates were about 15% in the Ditrame trial  while it was 5% in the Ditrame-Plus cohort, explained by a greater efficacy of PMTCT interventions in Ditrame-Plus  but also leading to the likely selection of HIV-infected newborns at higher risk of rapid disease progression. At inclusion in Ditrame-Plus, mothers from HIV-infected children were more advanced in the HIV/AIDS disease, more immune-suppressed, and their infected children were more likely to have been infected earlier (in utero). Therefore, we observed a trend of a protective effect of cotrimoxazole in Ditrame-Plus HIV-infected children having a poorer prognosis than children from Ditrame.
In the Ditrame-plus cohort, 20% of the children started an antiretroviral therapy as they were symptomatics, with a likely higher risk of experiencing a severe event: if these children would have not been right censored at the time of starting antiretroviral therapy in our analysis, this would have led to a probably lower risk of severe event in the ditrame plus cohort compared to the ditrame cohort but attributable to the ART effect rather than the cotrimoxazole effect, so with an overestimation of the cotrimoxazole effect.
We conclude that we have underestimated the true cotrimoxazole protective effect: we hypothesize the cotrimoxazole protective effect would have in fact probably exist with a greater public health impact if it was administrated to HIV-infected infants with a less severe HIV disease than in Ditrame-Plus and issued from HIV-infected mothers who would have received a less effective PMTCT intervention than the Ditrame-Plus antiretroviral combination. So, this would even have a greater impact in population with a lower standard of care access than in our well structured research context.
The ideal study design to assess the efficacy of cotrimoxazole prophylaxis given early in newborn would have been a randomized placebo controlled trial. Despite the lack of evidence of cotrimoxazole's effectiveness specifically in African infants less than 12 months in the CHAP trial , this would have been unethical because of the lack of equipoise between the two randomised cotrimoxazole and placebo arms as cotrimoxazole was already demondtrated to be efficacious in children older than 12 months  and in adults . Based on these latter finding, the good clinical practice was to provide the recommended standard of care to all study participants: national guidelines recommend cotrimoxazole prophylaxis in children since 1999 in Côte d'Ivoire . We choose this historical comparison to assess the cotrimoxazole effectiveness adjusting for the other known prognosis factors of infant mortality. We acknowledge that although important confounding variables such as maternal clinical stage of disease, 6-week paediatric viral load, birth-weight, breastfeeding were systematically controlled for in the analysis, this potential confusion was only partly taken into account notably because infant CD4 cell count was not available in the Ditrame study. Finally, the low rates of lost-to-follow-up of these HIV-infected children were acceptable in this difficult context and strengthened our conclusion.
The infant mortality figures among HIV-infected children reported in the Ditrame and Ditrame-Plus projects (59% and 50% respectively) were higher than those reported in a pooled analysis in which 35% of African children with early HIV-infection had died by age one year . The difference could be due to the large proportion of South African children in this pooled analysis, where infant mortality rate was reported to be lower than in West Africa according to 1998 demographic health surveys .: i.e. 59 per 1,000 livebirths in South Africa  versus 118 per 1,000 in Côte d'Ivoire . Our results suggest a reduced incidence of pneumonia as cause of death but a significant higher rate of diarrhoea as cause of death was observed in Ditrame-Plus compared to Ditrame. About 40% of the Ditrame-Plus children were formula-fed while their HIV-status was not yet ascertained, exposing formula-fed HIV-infected children to an increased mortality risk. Formula-feeding was not retained as a significant explaining factor for morbidity-mortality in our adjusted analysis, but HIV-infected children who are still breastfed at the time of HIV-diagnosis should better continue to be breastfed as recently reported elsewhere  and recommended .
As in others studies, our results confirm maternal death as the major determinant of infant mortality suggesting the crucial role of caregivers including mothers in the access to care for children in Africa [5, 34]. These findings thus emphasise the need for an early HIV-diagnosis, and an adequate access to antiretroviral care and support for HIV-infected children and all members of affected families, including mothers .
Cotrimoxazole demonstrated its efficacy in 2004 in a Zambian trial conducted among children mainly older than 12 months . In two South African studies, the history of cotrimoxazole use was significantly lower in children with severe pneumonia attributable to Pneumocystis jiroveci than those without [35, 36]. Lastly, HIV-infected infants with access to cotrimoxazole prophylaxis had a significantly lower incidence of pneumonia than those without access to prophylaxis . Our results are consistent with the Zambian and South African results observed in older children and reinforce the message suggesting that all HIV-infected children should receive cotrimoxazole prophylaxis in low-income settings, irrespective of their age. In 2006, WHO recommends that all infants exposed to HIV should receive cotrimoxazole prophylaxis from six weeks until HIV infection is excluded (rarely before 18 months in routine) .
While the number of antiretroviral-treated adults in low-income countries is increasing in Africa with the World Health Organization (WHO) 3 × 5 initiative since 2003, it is estimated that only 15% of people in need of antiretroviral treatment in 2005 are covered . There is particularly a lack of access to antiretroviral therapy in African children . The lack of routine available access to an early diagnosis of HIV-infection in children may explain this delayed access to antiretroviral therapy . In our setting, the real-time PCR test used was low cost (10 euros/sample) and highly accurate . The cost of providing systematically a daily cotrimoxazole prophylaxis to any HIV-exposed child (3 euros/month) during their first year of life is estimated to be (100*10 months*3) = 3000 euros for 100 HIV-exposed children. When using a targeted strategy guided by the early paediatric HIV-diagnosis (10 euros per unit), given the hypothesis of a 5% mother-to-child transmission rate, the cost for diagnosing 100 HIV-exposed children and providing to HIV-infected children a daily cotrimoxazole prophylaxis during their first year of life is estimated to be [(100*10)+(5*10 months*3)] = 1150 euros for 100 HIV-exposed children. Thus, not only is the cost of the targeted strategy more affordable, but also will it allow detecting early HIV-infected children to give them an opportunity to access to antiretroviral therapy.