Bulletin of the World Health Organization

Scaling up antiretroviral therapy for HIV-infected children in Côte d’Ivoire: determinants of survival and loss to programme

M-F Anaky a, J Duvignac b, L Wemin a, A Kouakoussui a, S Karcher b, S Touré a, C Seyler c, P Fassinou d, F Dabis b, T N’Dri-Yoman a, X Anglaret b & V Leroy b

a. Aconda-VS-CI, Abidjan, Côte d’Ivoire.
b. Centre de Recherche INSERM U897 (ex-U593), Université Victor Segalen Bordeaux 2, 146 rue Léo Saignat, 33076, Bordeaux Cedex, France.
c. Institut de Santé Publique, Epidémiologie et Développement, Université Victor Segalen, Bordeaux, France.
d. Elizabeth Glaser Pediatric AIDS Foundation, Abidjan, Côte d’Ivoire.

Correspondence to Valériane Leroy (e-mail: valeriane.leroy@isped.u-bordeaux2.fr).

(Submitted: 30 May 2009 – Revised version received: 19 August 2009 – Accepted: 02 November 2009 – Published online: 21 December 2009.)

Bulletin of the World Health Organization 2010;88:490-499. doi: 10.2471/BLT.09.068015


In 2007, an estimated 2 million children and 13 million women of childbearing age worldwide were infected with HIV.1 The HIV epidemic in children continues to grow, partly because only about 33% of HIV-infected pregnant women receive antiretroviral therapy (ART) for the prevention of mother-to-child transmission (PMTCT).1

The problems of caring for children with HIV differ from those in adults. First, over 50% of untreated HIV-infected children die before their second birthday.2 Second, routine diagnosis of paediatric HIV infection is more difficult: definitive diagnosis in children aged < 18 months requires virological testing, which is often unavailable in resource-limited settings.3 Moreover, clinicians must also take a dynamic view of HIV exposure among children because of the likelihood of infection during breastfeeding.4 This need to take uninfected children into account is difficult to explain to parents. Third, providing a continuum of care covering postnatal diagnosis in infants and HIV treatment after a PMTCT intervention in mothers still remains a challenge in 2009.1,5 Fourth, chronic diseases requiring lifelong daily treatment are more difficult to manage in children, especially when paediatric formulations and dosages are not available.6,7 As a result of these difficulties, in 2008 only an estimated 38% of all children aged < 15 years worldwide thought to need ART actually received it.8 Finally, additional issues, such as the need for informed consent, have led to a relative scarcity of studies on paediatric care in low-income countries and have put children at a disadvantage in AIDS research.914 Increasing the availability of HIV care in developing countries provides an unprecedented challenge.8 In particular, the public health impact of initiatives on paediatric HIV care in resource-limited settings must be assessed using conventional indicators of effectiveness.15 The aims of this study were to investigate the effectiveness of a paediatric HIV treatment and care programme in Côte d’Ivoire, West Africa, over a 3-year period and to identify factors that influenced mortality and the number of children lost to the programme.


Setting and patient characteristics

In June 2004, Aconda, a nongovernmental organization created by researchers who had studied cohorts of HIV-infected adults and children in Abidjan, Côte d’Ivoire, between 1996 and 2003,9,16 formed a partnership with the Institute of Public Health, Epidemiology and Development in Bordeaux, France, to study access to HIV care and treatment. The study was funded by the United States President’s Emergency Plan for AIDS Relief through the Elizabeth Glaser Pediatric AIDS Foundation in Washington, DC, United States of America (USA).

Details of the Aconda programme have been described elsewhere.17 Briefly, the Aconda team trained health workers in HIV care and implemented a standardized computer data management system which was controlled by a designated person in each study centre. Aconda staff visited study centres regularly and offered database management support online.

This analysis used data from children with HIV-1 infections, with or without concurrent HIV-2 infections, aged ≤ 15 years at 17 urban and 2 semi-urban health-care centres who had had at least one contact with the Aconda programme between 1 June 2004 and 30 November 2007. In 2008, the Aconda data management system evolved into a national HIV data management system for all Côte d’Ivoire under the auspices of the Ministry of Public Health.

Entry into the programme

The Aconda programme adopted a comprehensive family-based approach. Health-care workers were trained to offer HIV testing to every pregnant woman attending antenatal clinics and to encourage HIV-infected mothers to bring their children and partners with them for testing and counselling. Each pregnant woman with an HIV infection was immediately referred for an adult consultation. She then received PMTCT therapy, either a short regimen or ART depending on her clinical and immunological status, and underwent assessment at both antenatal and adult clinics.18 Children aged ≤ 15 years entered the Aconda programme in one of two ways: (i) after referral for HIV testing at the age of ≥ 6 weeks because their mother had been diagnosed with HIV infection and had received PMTCT therapy, or (ii) after HIV testing at a paediatric clinic following presentation with AIDS-related symptoms, even if they had not been previously classified as exposed to or infected by HIV and even if their parents had not participated in the Aconda programme.

Standardized paediatric follow-up

The Aconda paediatric HIV care package included systematic paediatric HIV testing which varied according to the child’s age. In those aged ≥ 18 months, the standard serum testing algorithm comprised a series of two rapid HIV assays: the Determine® HIV-1/2 assay (Inverness Medical, Bedford, United Kingdom of Great Britain and Northern Ireland) followed by the Genie II® HIV-1/HIV-2 assay (Bio-Rad laboratories, Marne-La-Coquette, France).19 Children aged < 18 months were diagnosed virologically using a TaqMan HIV-1, ribonucleic acid (RNA), real-time polymerase chain reaction test (Hoffmann–La Roche, Basel, Switzerland) with a threshold of 300 copies/ml.20 Children were regarded as having HIV-1 infections if, at any age, they tested positive for HIV-1 RNA in plasma at least once or if, at age ≥ 18 months, they were positive for HIV-1 on serum testing. Children who tested negative but who were still breastfeeding were defined as HIV-undetermined and were retested 2 months after the cessation of breastfeeding or at 18 months. A negative diagnosis was regarded as definite if it was made at least 2 months after the cessation of breastfeeding and children with this diagnosis were excluded from the programme.

All children with a confirmed HIV infection were seen monthly and had unrestricted free access to antiretroviral drugs and comprehensive care.21 In all children, whether on ART or not, the CD4+ T lymphocyte (CD4 cell) count and CD4 cell percentage were measured every 6 months. Plasma viral load testing was not performed routinely after the diagnosis of HIV infection, even in children on ART. Pulmonary radiographs were available for children whose history and symptoms suggested tuberculosis.

Children initiated ART if they were either at World Health Organization (WHO) HIV/AIDS clinical stage 3 or 4 or at clinical stage 1 or 2 with impaired immunity (i.e. a CD4 cell percentage ≥ 25 at age < 12 months, ≥ 20 at 12–35 months or ≥ 15 at ≥ 36 months).3 Until June 2007, first-line ART consisted of zidovudine or stavudine plus lamivudine plus nevirapine, efavirenz or nelfinavir. The drug combination used depended on the child’s age, weight and blood haemoglobin level. Co-trimoxazole prophylaxis was given to all HIV-exposed children from the age of 6 weeks and to all HIV-infected children regardless of age, as recommended by national Ivorian guidelines.21 A community-based women’s organization provided psychological, social and nutritional support, advice on how to disclose test results, and adolescent and orphan care.

The Côte d’Ivoire National Programme paid for ART, co-trimoxazole prophylaxis, plasma HIV-1 RNA testing, HIV serum testing and CD4 cell counts for children aged < 15 years.21 From April 2006, patients paid US$ 0.90 per prescription for each additional drug used against opportunistic infections. Patients received antiretroviral drugs and co-trimoxazole every month. A community-based team of experienced social workers and members of associations of people living with HIV/AIDS made telephone calls or home visits within 2 weeks to children who did not keep their scheduled appointments.17

Data management system

The national ethics committee of Côte d’Ivoire approved the Aconda data management system.17 Standardized forms were used to record the following variables: (i) at the initial visit, date, sex, date of birth or age, height, weight, PMTCT history, breastfeeding status, haematology results and CD4 cell count and percentage; (ii) at each subsequent visit, date and weight; (iii) at ART initiation, date, WHO HIV/AIDS clinical stage and weight; (iv) date, name and quantity of drugs delivered for each prescription, whether or not for ART; (v) at each CD4 cell and blood cell count measurement, date, CD4 cell count, CD4 cell percentage, blood haemoglobin level, and blood platelet, granulocyte and leukocyte counts; and (vi) date of death or, for patients whose death had not been reported, the last date at which the patient was known to be alive.

Statistical analysis

Kaplan–Meier analysis was used to determine the probability that a child aged < 18 months at first contact with the programme would be diagnosed with an HIV infection. We examined three outcomes for children with HIV-1 infection, with or without concurrent HIV-2 infection: death, loss-to-follow-up (i.e. last contact ≥ 3 months before 30 November 2007 if death had not been reported) and loss-to-programme (i.e. death or loss-to-follow-up).

Mortality and loss-to-programme rates, expressed in events per 100 child-years of follow-up, were calculated for the period between programme inclusion and ART initiation (i.e. the off-ART period), for the first 12 months on ART as a whole, and for months 0–3 and 4–12 of ART.

We estimated the probability that a death or loss-to-programme would occur during ART using the Kaplan–Meier method. We used crude and adjusted Cox proportional hazards regression models to identify baseline characteristics associated with death or a loss-to-programme during 12 months of ART. Characteristics considered included sex, haemoglobin level, CD4 cell percentage, weight-for-age z-score, initial WHO HIV/AIDS clinical stage and initial ART regimen. Data were censored on 30 November 2007 for patients known to be alive at that date and, for those whose death had not been reported by that time, on the date of last contact with the care centre. Cox models took missing data into account. Analyses were carried out using SAS software, version 9.1 (SAS Institute Inc., Cary, NC, USA).


In total, 4182 children attended a programme centre at least once. Of these, 290 (6.9%) were excluded from the analysis because they had started ART at first presentation and 16 (0.4%) were excluded because their age was unknown.

Fig. 1 provides details of the children who entered the programme, were tested for HIV, and received ART. Of the 3876 children included in the analysis, 1766 (46%) were diagnosed with an HIV-1 infection, with or without HIV-2, 19 (0.5%) were diagnosed with an HIV-2 infection only, 1870 (48%) tested negative for HIV and were excluded from the analysis, and 221 (5.7%) were classified as having an HIV-undetermined status at the end of the study period.

Fig. 1. Flow diagram of the cohort enrolled in the Aconda HIV treatment programme for children, Côte d’Ivoire, 2004–2007
Fig. 1. Flow diagram of the cohort enrolled in the Aconda HIV treatment programme for children, Côte d’Ivoire, 2004–2007
ART, antiretroviral therapy; RNA, ribonucleic acid.a Children aged < 18 months who were exposed to maternal HIV underwent virological testing (i.e. HIV-1 RNA measurement).b HIV-negative children were excluded after complete cessation of breastfeeding.c Children whose HIV status was undetermined were still being exposed to maternal HIV during breastfeeding at their last visit.d Children were lost to follow-up if their last contact was more than 3 months before the end of the study.

Among the 1215 children exposed to HIV and aged < 18 months at first presentation, the Kaplan–Meier probability of undergoing HIV RNA testing was 59% (95% confidence interval, CI: 56–63) at 1 month after the first contact with the programme, 76% (95% CI: 72–80) at 3 months and 80% (95% CI: 76–84) at 6 months. Of these 1215 children, 307 (25%) were confirmed as having an HIV infection at a median age of 9 months (interquartile range, IQR: 4–14), as shown in Fig. 1. The majority of children who did not undergo HIV RNA testing attended programme centres at the rural periphery of Abidjan. Only 413 (34%) children aged < 18 months had mothers who were involved in a PMTCT programme.

Of the 1766 children diagnosed with an HIV infection, 124 (7.0%) died without starting ART, 52 (2.9%) were transferred out of their programme centres, 354 (20%) were lost to follow-up before starting ART, 224 (13%) did not meet ART initiation criteria, 35 (2%) met ART initiation criteria but remained in care without starting ART, and 977 (55%) started ART before the end of the study.

Table 1 (available at: http://www.who.int/bulletin/volumes/88/7/09-068015) summarizes the demographic and clinical characteristics of children at diagnosis of HIV infection and at ART initiation. The median age of the 1766 children at diagnosis of HIV infection was 54 months. Of these children, 307 (17%) were aged < 18 months, 1378 (78%) underwent at least one CD4 cell measurement and 1142 (65%) were immediately eligible for ART: 374 (21%) on the basis of clinical criteria and 768 (44%) on immunological criteria. Overall, 977 children initiated ART a median of 1 month (IQR: 0–4) after inclusion in the Aconda programme.

Table 2 shows mortality and loss-to-programme rates for children who were and were not on ART. The mortality rate in all children who were off ART was 13.0 per 100 child-years of follow-up; in those aged < 18 months, 18–59 months and ≥ 60 months at first presentation, the rate was 35.5, 12.2 and 9.2 per 100 child-years, respectively (Table 2). In total, 478 (27%) children were lost to the programme before ART initiation, giving a loss-to-programme rate of 50.3 per 100 child-years of follow-up. The loss-to-programme rates before ART were 3 to 4 times higher than mortality rates for all age groups.

The 977 children who received ART had the following characteristics at the start of therapy: median age: 5 years; median weight-for-age z-score: −2.99; median CD4 cell percentage: 10.7; and proportions with a history of completed tuberculosis treatment: 2.7%; with active tuberculosis at baseline: 3.3%; in WHO HIV/AIDS clinical stage 3 or 4: 43%; on co-trimoxazole prophylaxis at ART initiation: 54%; and on co-trimoxazole prophylaxis within the first year following ART initiation: 89.5%. Sixty-five percent of patients, 66% of whom were aged ≥ 60 months, initiated ART with a combination of two nucleoside reverse transcriptase inhibitors, such as azidothymidine, stavudine, didanosine or lamivudine, and one non-nucleoside reverse transcriptase inhibitor, such as efavirenz or nevirapine. Overall, 33% of children initiated ART with a combination of two nucleoside reverse transcriptase inhibitors and one protease inhibitor, such as nelfinavir, though among those aged < 18 months, 55% started on a regimen containing a protease inhibitor (Table 1). The most frequent combinations prescribed were stavudine, lamivudine and efavirenz, in 28%, and azidothymidine, lamivudine and efavirenz, in 25%.

The overall mortality rate on ART was 14.4 per 100 child-years of follow-up and was significantly higher in the first 3 months of ART than during months 4–12, at 32.8 and 6.9 per 100 child-years of follow-up, respectively. Sixty-six percent of deaths during ART occurred within the first 3 months of therapy. During this period, the mortality rate was 43.7 per 100 child-years in children aged < 18 months and around 30 per 100 child-years in older children. Mortality rates decreased 4 to 6 times in all age groups after the third month on ART (Table 2). There was no significant difference in Kaplan–Meier probability of death during 12 months of ART between the age groups (logrank test P = 0.48) (Fig. 2 and Table 3).

Fig. 2. Kaplan–Meier probability of death during antiretroviral therapy (ART) in HIV-infected children in the Aconda HIV treatment programme, by age group, Côte d’Ivoire, 2004–2007
Fig. 2. Kaplan–Meier probability of death during antiretroviral therapy (ART) in HIV-infected children in the Aconda HIV treatment programme, by age group, Côte d’Ivoire, 2004–2007
ART, antiretroviral therapy.

The loss-to-programme rate during ART was roughly double the mortality rate, both overall and in the different age groups (Table 2). It peaked at 53 per 100 child-years in the first 3 months of ART. The Kaplan–Meier probability of a child being lost to the programme 12 months after ART initiation was 0.30 in children aged < 18 months at ART initiation, 0.19 in those aged 18–59 months and 0.18 in those aged ≥ 60 months. The probability of a loss-to-programme differed significantly between the age groups on univariate analysis (logrank test P = 0.03) (Fig. 3 and Table 4).

Fig. 3. Kaplan–Meier probability of loss-to-programmea during antiretroviral therapy (ART) in HIV-infected children in the Aconda HIV treatment programme, by age group, Côte d’Ivoire, 2004–2007
Fig. 3. Kaplan–Meier probability of loss-to-programme<sup>a</sup> during antiretroviral therapy (ART) in HIV-infected children in the Aconda HIV treatment programme, by age group, Côte d’Ivoire, 2004–2007
ART, antiretroviral therapy.a Children were lost to the programme if they died or were lost to follow-up (i.e. their last contact was > 3 months before the end of the study).

Data on demographic and clinical variables of interest were recorded for 746 (76%) of the 977 children who received ART. Associations between these variables and mortality and loss-to-programme during ART were identified using Cox proportional hazards regression models. It was found that both mortality and loss-to-programme were independently and significantly associated with a weight-for-age-z-score < −2, a CD4 cell percentage < 10 and WHO HIV/AIDS clinical stage 3 or 4 (Table 5). A blood haemoglobin level < 8 g/dl was significantly associated with mortality alone. There was neither significant association between any pre-ART variable and the adjusted risk of mortality or loss-to-programme, nor between age, sex, co-trimoxazole prophylaxis at baseline or initial ART regimen and outcomes. Stratifying the data by age group gave similar results.


The results of our study revealed information important for guiding the expansion of paediatric HIV care. First, 86% of children in the programme who were eligible for ART according to WHO 2006 criteria started therapy. Second, although over 75% of those aged < 18 months had an HIV-RNA test, many were tested ≥ 3 months after first contact with the programme, which meant that HIV diagnosis and ART were too late.22 Third, although the percentage of children who received co-trimoxazole in the programme (i.e. 50% of those off ART and 90% of those on ART) did not meet guideline recommendations,23 useful information has been collected for sub-Saharan Africa, where data are generally lacking.22,24 Fourth, mortality rates after the first 3 months of ART in all age groups were consistently lower than before therapy or during the first 3 months, despite children being relatively old at ART initiation and often being severely immunocompromised. Loss-to-programme rates were twice as high but followed similar trends in all age groups.

Other studies have shown that paediatric responses to ART are similar in low- and high-income countries but that, in low-income countries, therapy is often started too late, when children are already immunodeficient.9,10,12,13,2528 This also occurred in our cohort and most likely resulted in high rates of mortality and loss to follow-up, with the latter often including underdocumented mortality.

To improve the accessibility and effectiveness of HIV care in Côte d’Ivoire, several systematic weaknesses must be addressed. First, strengthening the links between PMTCT services and paediatric HIV care could improve prognosis in children who still acquire HIV infections. Second, survival on ART has been associated with a low baseline CD4 cell count in both resource-rich29,30 and low-income countries.9,10,12,13 Earlier access to treatment could prevent many deaths.28 The significant drop in mortality during ART we observed after 3 months has been reported elsewhere in Africa.13,31 The early peak in mortality may have been due to severe malnutrition, ART toxicity or immune reconstitution inflammatory syndrome (IRIS), which may be common in young children with a low CD4 cell count at ART initiation, especially if therapy is started too late. Indeed, IRIS has been associated with high morbidity in South Africa32 and Thailand.33 Our findings support the WHO 2008 recommendation that ART should be started immediately in all HIV-infected infants.34

A third systematic weakness in Africa is the limited availability of HIV-RNA tests for infants. Tests should be offered to HIV-exposed infants as early as possible after the age of 6 weeks and should be repeated if children are still exposed via breastfeeding. The routine use of dried blood spot testing could help increase early HIV diagnosis in infants aged over 6 weeks.35,36 Fourth, severe malnutrition remains the most important predictor of both mortality and loss-to-programme, even compared with a low CD4 cell count. However, impaired immune status is generally an even stronger predictor among HIV-infected children off ART.37 Finally, although our findings on mortality are encouraging, mortality rates would double if we assumed that losses to the programme may be due to undocumented deaths, as observed in adult HIV studies.38,39

Overall, our paediatric ART cohort encountered the same operational problems observed in adult ART programmes: an early peak in mortality on ART and high cohort attrition when programmes are scaled-up.4042 Socioeconomic factors are among the primary reasons for losses to follow-up.43 New strategies are urgently needed to retain children and their caretakers in HIV programmes.

In this study, we showed that clinical markers, such as weight-for-age, haemoglobin level and CD4 cell count, can help predict responses to ART in low-income countries. Data on these markers are easy to collect and could provide information for improving paediatric HIV care and assessing programmes.

Finally, the successful and rapid implementation of paediatric ART programmes depends on dealing urgently with several programmatic issues. An effort should be made to identify all HIV-infected infants early. The barriers to scaling-up treatment include the failure to identify maternal HIV infection, weak links between PMTCT and child care programmes, and limited access to HIV-RNA testing for early diagnosis.5,7,44 Education about ART should start antenatally and should be monitored to increase postnatal infant HIV screening from the age of 6 weeks. Political commitment is needed to mobilize health-care and community leaders. Lastly, ensuring an adequate number of trained staff is essential for the successful scaling-up ART in this context.45


We would like to warmly thank all children and their families who were involved in this study, the whole Aconda Study Group in Abidjan, Côte d’Ivoire, the Aconda care centres in Côte d’Ivoire and the Abidjan EGPAF Foundation in Abidjan. Special thanks to Catherine Wilfert (EGPAF) for her helpful support.

Co-authors J Duvignac, S Karcher, F Dabis, X Anglaret & V Leroy are also affiliated with the Institut de Santé Publique, Epidémiologie et Développement, Université Victor Segalen, Bordeaux, France.


This study was funded by the United States President’s Emergency Plan for AIDS Relief (PEPFAR) through the Elizabeth Glaser Paediatric AIDS Foundation (EGPAF, Washington, USA), with additional support from the French National Agency for Research on AIDS and Viral Hepatitis, the International Epidemiological Database to Evaluate AIDS in Africa (IeDEA) of the National Institutes of Health (Bethesda, USA) and the Ministry of Public Health of Côte d’Ivoire.

Competing interests:

None declared.