It is estimated that by the end of 2006, nearly 90% of the 2.3 million children living with HIV worldwide were from sub-Saharan Africa. In the absence of antiretroviral therapy (ART), the median survival age for vertically infected African children is only 2 years compared to 8–10 years in US or Europe [1,2]. Malnutrition, respiratory and diarrhoeal diseases are among the leading causes of death, with considerable overlap in the spectrum of diseases with uninfected children [3–5].
Combination ART has radically improved prognosis for HIV-infected adults and children in industrialized countries. With access to ART for children increasing globally, the issue of when it should be started in an individual child remains complex. Recently updated World Health Organization paediatric treatment guidelines for resource-limited settings recommends that treatment initiation be based on clinical stage or monitoring by CD4 percentage and count, or total lymphocyte count (where CD4 technology is unavailable) . The marker thresholds chosen were, however, largely influenced by findings from the HIV Paediatric Prognostic Collaborative Study (HPPMCS), an individual patient meta-analysis of longitudinal data on approximately 4000 untreated children in Europe and US [7–9]. Reliable data of a similar nature required for validating these guidelines has been lacking from resource-limited settings . Furthermore, the prognostic value of simpler, low-cost clinical or laboratory alternatives to CD4 cell count or percentage, such as haemoglobin and weight-for-age, have been suggested in some small studies; these may be more appropriate for monitoring in resource-limited countries and require further validation [10–12].
In this Cross Continents Collaboration for Kids (3Cs4kids) study, individual child longitudinal data have been pooled for approximately 2500 untreated vertically HIV-infected children from mainly African studies. This is used to evaluate the prognostic value of selected laboratory and growth markers on the short-term risk of mortality.
All studies participating in 3Cs4kids are observational cohorts, apart from one randomized trial of cotrimozaxole prophylaxis (CHAP trial) [11,13–16]. Collaborators were requested to provide information on: date of birth, gender, date of enrolment or first clinic visit, date of starting any antiretroviral drugs, date of starting cotrimoxazole prophylaxis, date of death or last known to be alive, and available measurements for CD4%, CD4 cell count, total lymphocyte count (TLC), haemoglobin, weight and height.
Only measurements taken before start of ART and after 12 months of age were analysed. Data for infants were excluded because mortality rate in the first year of life was much lower than expected in exploratory analyses, with site investigators confirming that since HIV diagnosis and follow-up was particularly difficult in this age-group, early deaths were likely to be greatly under-ascertained. Children with less than 1 month of follow-up before starting ART were also excluded.
For each child, follow-up began at the first available measurement and ended at the earliest of: date of death, date last known to be alive, and 3 months after starting any antiretroviral drugs (censoring immediately at start of ART could have resulted in bias if children with advanced disease were more likely to be treated). For one study (from Brazil, the only non-African study) with extensive follow-up data before 1996, censoring was delayed until end 1995 for the 12% of children who started on ART earlier, mainly with zidovudine monotherapy. At a given time point, the value for a marker was set to the most recent measurement in the preceding 12 months or otherwise assigned missing if this was unavailable. Consequently, follow-up for a child was further censored at 12 months after the last measurement.
Crude mortality rates were calculated according to age and the most recent value for each marker, with markers categorized to approximately correspond to treatment guidelines where relevant . Univariate Poisson models were fitted with each marker included as a continuous, time-dependent covariate, adjusting for age. This assumes a constant mortality rate between successive measurements up to a maximum period of 12 months, which was verified in a Weibull survival model where a shape parameter of 1.04 [95% confidence interval (CI) 0.76 to 1.43] was obtained. Variation between studies was accounted for by a random-effect term, while there was no evidence of a calendar time effect (P-value = 0.6). Since the use of cotrimoxazole prophylaxis (estimated to reduce mortality risk by 43% ) varied between and within studies and was additionally influenced by the child's prognosis, we adjusted for its effect using a time-dependent covariate corresponding to before and after start of treatment, with its hazard ratio fixed at 0.57.
Based on the hazards derived from the models, mortality risk within the next 12 months was estimated from 100 * (1 − ehazard) . Estimates obtained across marker values at selected ages were standardized to a hypothetical child receiving cotrimoxazole. Results for the CD4%, CD4 cell count and TLC were compared against those previously derived in HPPMCS [7–9].
Multivariable Poisson models were then used to assess whether, after adjusting for either CD4% or CD4 cell count but not both, the effect of TLC, haemoglobin, weight, height and body mass index (BMI) remained. These analyses were restricted to follow-up periods with relevant markers available.
Weight, height and BMI, adjusted for age, were expressed as z-scores standardized to the UK reference for uninfected children  (as recently produced WHO growth charts are only available up to age 5 years and also require validation) . Age was log transformed and both TLC and CD4 cell count were square root-transformed to improve model fit. Effects of markers were examined for nonlinearity using a cubic spline term , and for interaction with age.
Sensitivity-analyses assessed the effect of: (1) changing the impact of cotrimoxazole on mortality from a 43% reduction in risk to either 20 or 70%; (2) excluding the Brazilian cohort from analysis; and (3) allowing a given measurement for a marker to remain valid for a maximum period of 6 months instead of 12.
A total of 2510 children aged at least 12 months from ten studies (nine African, one Brazilian) were included, nearly half (48%) of whom were girls (Table 1). Median age at first measurement was 4.0 [interquartile range (IQR), 2.2 to 7.0] years, with 11% over 10 years. Median CD4% was 15% (IQR, 9 to 22%) and weight-for-age was −1.9 (IQR, −3.2 to −0.8). As expected, the distribution of CD4%, CD4 cell count and TLC declined with age (results not shown). Younger children were more likely to be anaemic, with 14.8% of those aged less than 3 years had an initial haemoglobin < 8 g/dl in comparison with 9.4% in older ages. Forty percent of children had both weight-for-age and height-for-age < −2 at first measurement; 17% had BMI-for-age < −2.
Median follow-up per child was 12.7 months (IQR, 6.4–24.6) (Table 1). Cotrimoxazole prophylaxis was prescribed for 88% of children (62% within a month of presentation), with 81% of follow-up occurring after starting cotrimoxazole. Ninety percent of follow-up was from 2000 onwards, when seven of the 10 studies began collecting data.
Overall, 357 deaths were recorded during 3769 child-years-at-risk (Table 2). The median number of growth measurements per child was 5 (IQR, 2–9) in comparison with 2 (IQR, 1–3) for CD4%, with corresponding median interval between successive tests of 1.8 and 6.4 months, respectively. Between 12 and 38% of children contributed only one measurement to different analyses, although these accounted for only 6–16% of overall follow-up. As follow-up was censored at 12 months from the last measurement, the proportion of children who died was lower in analyses for laboratory versus growth markers.
Mortality rates by most recent value for each marker showed clear trends across all age groups, with the risk increasing as marker values decrease (Table 3). The exceptions were CD4 cell count, which was less prognostic among children aged 1–2 years, and TLC which only predicted higher rates at values below 2000 cells/μl.
After adjusting for age, cotrimoxazole use and study effects, all markers individually predicted mortality. Comparison of the log-likelihood for univariate models restricted to clinic visits with all markers available showed that CD4 cell count (log-likelihood = −701.8) and CD4% (−709.4) gave the best statistical fit to the data and so were the strongest predictors, followed by weight-for-age (−727.7) then height-for-age (−755.7), haemoglobin (−761.1), BMI-for-age (−762.5) and TLC (−763.3).
Comparing results in Cross Continents Collaboration for Kids versus HIV Paediatric Prognostic Collaborative Study
When compared against corresponding results in the HPPMCS study, the 12-month risk of death estimated by CD4%, CD4 cell count and TLC was generally higher in 3Cs4kids at any given marker value and age (Fig. 1). Furthermore, the increase in risk with decreasing marker values was more gradual and occurred at higher thresholds for children from resource-limited settings. For example, in 3Cs4kids, the annual risk at age 2 years increased from 5 to 10% as CD4% fell from 22 to 14% whereas the same increase in risk was observed at CD4% change from 16 to 11% in HPPMCS. Corresponding changes in CD4% at age 5 years were 13 to 8% in 3C4kids and 10 to 7% in HPPMCS, respectively. Of note, TLC was only weakly prognostic in 3Cs4kids, despite being a good predictor of disease progression in HPPMCS .
The strong impact of age on mortality risk by CD4 cell count, CD4% or TLC was seen in both studies, with younger children having worse prognosis across marker values. In addition, an interaction between markers and age was evident, with older children experiencing steeper rise in risk as marker values decrease.
Weight-for-age and haemoglobin remained strongly predictive of mortality after controlling for CD4% or CD4 count, as well as each other, as shown in Fig. 2 (both with P-values < 0.001). For example, in a 5-year old child with CD4% of 20%, the 12-month risk of death increased by 3.6 times when weight-for-age decreased from −1 to −3 and by 1.9 times when haemoglobin fell from 11 to 8 g/dl.
Conversely, there was still a steady gradient in risk with CD4% or CD4 count at any specific weight-for-age and haemoglobin values, though both markers decreased in predictive value with lower weight-for-age (P-value for interaction < 0.001 and 0.001, respectively). Mortality remained high even at high CD4% or count values for young children with severe malnutrition (weight-for-age ≤ ????%) and anaemia (haemoglobin ≤ 8 g/dl), particularly those aged 1–2 years. In contrast, high CD4% or CD4 cell count values predicted low risk at age 5 years onwards and also in younger children with neither severe malnutrition nor anaemia.
After adjustment for CD4%, TLC retained only a small independent effect, with a decrease from 4000 to 2000 cells/μl leading to an approximate 1.4-fold increase in 12-month risk of death (P-value < 0.001). Of note, TLC had no prognostic value when CD4 cell count was known (P = 0.6), suggesting that any additional value of TLC is captured in CD4 cell count. In an adjusted analysis including weight-for-age, the estimated hazard ratio for BMI-for-age was 0.87 (95% CI, 0.80 to 0.95; P = 0.002) per unit increase in z-score, corresponding to an approximate 1.3-fold increase in annual risk when BMI-for-age decreases from 0 to -2. By contrast, the association between height-for-age and mortality reversed direction giving a hazard ratio of 1.21 per unit increase in z-score (95% CI, 1.06–1.38, P = 0.004), probably reflecting a greater degree of malnutrition among taller children of the same weight.
Relationship between markers
As expected, weight for age and haemoglobin were strongly associated with CD4%. The proportion of follow-up time spent with CD4% below 15% was 68% at weight-for-age < −3, compared to 43% at weight-for-age −3 to < −1 and 25% at ≥ −1. Corresponding proportions at haemoglobin < 8g/dl, 8 to < 11 and ≥ 11 were 65, 49 and 30%, respectively.
Since most follow-up occurred after receipt of cotrimoxazole prophylaxis, varying the effect of cotrimoxazole on mortality had a relatively small effect on results. For example, the risk of death for a 5-year-old child with CD4% of 15%, weight-for-age of −3 and haemoglobin of 8 g/dl changed from 7.6% (corresponding to a 43% efficacy) to 8.0 and to 6.6% when the efficacy was set to 20 and 70%, respectively. The results were also not appreciably affected when the Brazilian cohort was excluded or when measurements of markers were assumed to remain valid for up to 6 months instead of 12, with the estimated risks in Fig. 2 changing by a maximum absolute value of 1.2 and 2.0%, respectively.
Based on longitudinal data of approximately 2500 untreated HIV-infected children aged over 12 months from mainly African studies, we found that CD4 percentage and CD4 cell count were the strongest predictors of short-term risk of death among the laboratory and growth markers examined. Similar to children from Europe and US in HPPMCS, prognosis was poorer at younger ages for a given CD4% or CD4 count value while the predictive value of both markers improved with age across childhood. As expected, however, mortality level was generally much higher in 3Cs4kids compared with HPPMCS when controlling for either of the CD4 parameters and age. Furthermore, the steep rise and threshold effect in mortality risk with falling marker values observed in HPPMCS, which influenced the choice of marker thresholds for ART initiation in the WHO guidelines, was less pronounced in 3Cs4kids. Consequently, both CD4% and count were less effective in discriminating between low and high mortality levels for children in resource-limited settings. This is comparable with results for untreated adults; mortality rate in a South African cohort was similar to that from European cohorts at a CD4 cell count below 200 cells/μl, but was 8–10 times greater at higher values [21,22].
The strong effect of CD4% and CD4 cell count over and above other markers underlines the importance of access to low-cost laboratory monitoring, given that standard flow cytometry CD4 technology remains impractical and costly in many resource-limited settings. Many cheaper and simpler assays exist which can measure absolute CD4 cell count but not CD4 percentage . Use of the CD4 cell count for monitoring prior to ART is supported by the suggestion in the study that the two markers were similarly prognostic, as previously observed in HPPMCS . The CD4 count is difficult to interpret in young children, however, due to its dramatic decline in early life in HIV-uninfected children . The financing of currently available CD4% technology and the development of more affordable and feasible technology for resource-limited settings is therefore a priority. Comparison of monitoring by CD4% versus CD4 cell count is being further investigated using data from both 3Cs4kids and HPPMCS.
A key finding from this analysis is that TLC was a poor predictor of mortality for children in resource-limited settings, despite being highly prognostic among children in industrialized countries . This could be due to differences in population distribution of TLC, possibly because of the greater exposure to endemic infections in resource-limited settings. Higher levels of TLC had been reported among uninfected African children compared to European black children over 2 years of age . A cohort in Nairobi also found that perinatally HIV-infected children had overlapping TLC levels with uninfected children across ages . Of interest, TLC was more weakly associated with CD4 cell count in 3Cs4kids compared to HPPMCS, with age-specific correlation coefficients ranging from 0.43–0.47 and 0.69–0.72, respectively.
Evidence from our study emphasizes the importance of nutritional support and prevention and treatment of anaemia in HIV-infected children [10–12]. In accordance with Walker et al., we observed weight-for age was more prognostic than either height-for-age or BMI-for-age. We found mortality was high among young children who were both severely malnourished and anaemic, regardless of their immunological status. At older ages, children with low mortality risk, in whom ART could be deferred, were more effectively identified when CD4% (or CD4 cell count) was used in conjunction with weight-for-age and haemoglobin than alone. In spite of this, it is not straight-forward to define the appropriate thresholds of these markers for ART initiation due to the complex inter-relationships between HIV infection, nutritional status and anaemia. In sub-Saharan Africa, infected children are more likely to be malnourished, anaemic and have chronic diarrhoea in comparison with to uninfected children while conversely, many HIV and non-HIV-related illnesses are attributable to malnutrition and anaemia [3,5,12,26,27]. These factors are likely to contribute to our findings that both CD4% and CD4 cell count were less prognostic at lower weight-for-age values, suggesting that the increased mortality associated with malnutrition is not due to advanced HIV disease alone. This is supported by a previous observation that children who were successively treated for prior malnutrition did not have higher subsequent mortality . Current WHO treatment guidelines recommend ART initiation for children with ‘unexplained severe malnutrition’ who do not respond to nutritional therapy, although data on how to assess such response is lacking .
Our study has several important limitations. Children under 1 year of age were excluded because mortality rate was substantially under-estimated in this age-group among participating cohorts. Of note, early results from the ongoing randomized South African ‘Children with HIV Early antiRetroviral therapy’ (CHER) trial demonstrated a significant 75% reduction in mortality after median follow-up of only 32 weeks among asymptomatic babies initiating ART before 12 weeks of age compared to deferring treatment according to guidelines . The wide variation in mortality rates between studies in 3Cs4kids, even after adjustment for age and prognostic markers (data not shown), suggest that under-ascertainment of deaths from loss to follow-up could also have occurred among older children, as observed in adult cohorts with passive follow-up in resource-limited settings . Given also that longitudinal follow-up in 3Cs4kids was considerably less than in HPPMCS (median follow-up per child of 13 months versus 24 months, respectively), we caution against over-interpretation of our quantitative estimates of mortality risk which, although higher than HPPMCS, could be under-estimated. Another factor which complicates the comparison of results between the two studies is their different statistical methods, especially with respect to analysis of cotrimoxazole prophylaxis. In 3Cs4kids, cotrimoxazole use was adjusted for by assuming the efficacy reported in the CHAP trial (found to be independent of age and CD4% values) , with risks presented standardized to a hypothetical child receiving continuous treatment. By comparison, in HPPMCS where some follow-up data occurred before the widespread use of cotrimoxazole prophylaxis in Europe and US, cotrimoxazole use could not be adjusted for since relevant information was unavailable. This would, however, potentially only further under-estimate the difference in mortality levels between the studies.
Unfortunately, we were not able to analyse progression to AIDS, as in HPPMCS, since most studies did not record such data. A lack of association between CD4% and Centers for Disease Control clinical staging was reported in a retrospective paediatric South African study . Conversely, growth variables and haemoglobin were not collected in HPPMCS, so cannot be compared in the two studies. Information on causes of death was not available in either study for comparison. Cause of death is frequently difficult to ascertain for HIV-infected and uninfected children in Africa although bacterial infections, tuberculosis, malaria, malnutrition and diarrhoeal disease are important in both [3,4].
In conclusion, the issue of when to start antiretroviral therapy in children in resource-limited settings remains complex, particularly with evidence from randomized clinical trials remaining lacking. The PREDICT trial coordinated by the National Institute of Allergy and Infectious Diseases is currently recruiting children aged 1 to 12 years with CD4% values of 15–24% in Thailand and Cambodia to compare immediate versus delayed treatment on AIDS-free survival. Our findings indicate that growth markers and haemoglobin should also be considered if similar trials are to be planned for sub-Saharan Africa. In addition, the role of malnutrition on ART response warrants further research as well as, conversely, the impact of ART on subsequent growth patterns among malnourished children . Finally, for effective care of HIV-infected children in resource-limited settings, the prevention and treatment of malnutrition and anaemia need to be integrated within routine clinical management, while the optimal timing of starting ART in relation to these interventions requires clarification.
Charlotte Duff coordinated the data extraction with collaborating studies and assembled the database. We thank all the patients and staff from all the centres contributing data to 3Cs4kids.
Trin Duong, Diana M Gibb, David Dunn, Chifumbe Chintu, Veronica Mulenga, Mark Cotton, Brian Eley, Heather Zar, Jane Ellis, Stephen Graham, Carlo Giaquinto, Maria Nanyonga, Philippe Msellati, Tammy Meyers, Harry Moultrie, Chris Hani, Jorge Pinto, Paul Roux, Ralf Weigel.
Collaborators and centres in Cross Continents Collaboration for Kids
South Africa: Professor Heather Zar, Professor Brian Eley (School of Child and Adolescent Health, Division of Paediatric Pulmonology, Red Cross Children's Hospital, Cape Town); Professor Paul Roux (Paediatric HIV/AIDS Service, Groote Schuur Hospital, Cape Town); Professor Mark Cotton (Children's Infectious Disease Clinical Research Unit (KID-CRU), Faculty of Health Sciences, Stellenbosch University, Tygerberg Children's Hospital, Cape Town); Dr Tammy Meyers, Dr Harry Moultrie (Harriet Shezi Children's Clinic, Chris Hani Baragwanath Hospital, Soweto, Johannesburg).
Zambia: Veronica Mulenga, Chifumbe Chintu, Chipepo Kankasa (Department of Paediatrics, University Teaching Hospital and School of Medicine, Lusaka).
Cote d'Ivoire: Philippc Msellati, Patricia Fassinou, Dr N Elenga (Programme Enfant Yopougon, Abidjan).
Malawi: Stephen Graham, Jane Ellis (Department of Paediatrics, College of Medicine, Blantyre); Ralf Weigel (Lighthouse Clinic, Kamuzu Central Hospital, Lilongwe).
Uganda: Carlo Giaquinto, Maria Nanyonga, Erika Morelli, Betty Atai (St Francis Nsambya Hospital, Kampala).
Brazil: Jorge Pinto, Claudete Araújo, Andrea Carvalho, Inácio Carvalho, Ana Diniz, Flávia Ferreira, Vanessa Lobato, Talitah Sanchez (Federal University of Minas Gerais).
United Kingdom: Trinh Duong (Medical Research Council, Clinical Trials Unit, London School of Hygiene and Tropical Medicine, London); David Dunn, Diana M Gibb, Charlotte Duff (Medical Research Council Clinical Trials Unit, London).
Disclaimer: We declare that we have no conflict of interest.
Sponsorship: 3Cs4kids received a research grant from World Health Organisation.
Funding towards meetings of collaborators was provided by the Paediatric European Network for Treatment of AIDS (PENTA), GlaxoSmithKline and the Department for International Development, UK.
1. Newell M-L, Coovadia H, Cortina-Borja M, Rollins N, Gaillard P, Dabis F, et al
. Mortality of infected and uninfected infants born to HIV-infected mothers in Africa: a pooled analysis. Lancet 2004; 364:1236–1243.
2. European Collaborative Study. Fluctuations in symptoms in HIV-infected children: the first 10 years of life
3. Berkley JA, Lowe BS, Mwangi I, Williams T, Bauni E, Mwarumba S, et al
. Bacteremia among children admitted to a rural hospital in Kenya. N Engl J Med 2005; 352:39–47.
4. Mulenga V, Ford D, Walker AW, Mwenya D, Mwansa J, Sinyinza F, et al
. Effect of cotrimoxazole on causes of death, hospital admissions and antibiotic use in HIV-infected children. AIDS 2007; 21:77–84.
5. Taha TE, Graham SM, Kumwenda NI, Broadhead RL, Hoover DR, Markakis D, et al
. Morbidity among human immunodeficiency virus-1-infected and -uninfected African children. Pediatrics 2000; 106:e77.
6. World Health Organization. Antiretroviral therapy of HIV infection in infants and children in resource-limited settings: towards universal access, 2006 revision. Geneva: World Health Organization. Available at: http://www.who.int/hiv/pub/guidelines/art/en/index.html
. Accessed May 2007.
7. HIV Paediatric
Prognostic Markers Collaborative Study. Short-term risk of disease progression in HIV-1-infected children receiving no antiretroviral therapy or zidovudine monotherapy
8. HIV Paediatric
Prognostic Markers Collaborative Study. Predictive value of absolute CD4 cell count for disease progression in untreated HIV-1-infected children
9. HIV Paediatric
Prognostic Markers Collaborative Study. Total lymphocyte count for informing when to initiate antiretroviral therapy in HIV-infected children: a meta-analysis of longitudinal data
10. Walker AS, Mulenga V, Sinyinza F, Lishimpi K, Nunn A, Chintu C, et al
. Determinants of survival without antiretroviral therapy and infancy in HIV-1 infected Zambian children in the CHAP trial. J Acquir Immune Defic Syndr 2006; 42:637–645.
11. van Kooten Niekerk NK, Knies MM, Howard J, Rabie H, Zeier M, van Rensburg A, et al
. The first 5 years of the family clinic for HIV at Tygerberg Hospital: family demographics, survival of children and early impact of antiretroviral therapy. J Trop Pediatr 2006; 52:3–11.
12. Clark TD, Mmiro F, Ndugwa C, Peryy RT, Jackson JB. Risk factors and cumulative incidence of anaemia among human immunodeficiency virus-infected children in Uganda. Ann Trop Paediatr 2002; 22:11–17.
13. Chintu C, Bhat GJ, Walker AS, Mulenga V, Sinyinza F, Lishimpi K, et al
. Co-trimoxazole as a prophylaxis against opportunistic infections in HIV-infected Zambian children (CHAP): a double-blind randomised placebo controlled trial. Lancet 2004; 364:1865–1871.
14. Kouakoussui A, Fassinou P, Anaky MF, Elenga N, Laguide R, Wemin ML, et al
. Respiratory manifestations in HIV-infected children pre and post-HAART in Abidjan, the Ivory Coast. Paediatr Respir Rev 2004; 5:311–315.
15. Zar HJ, Cotton MF, Strauss S, Janine K, Hussey G, Schaaf HS, et al
. Effect of isoniazid prophylaxis on mortality and incidence of tuberculosis in children with HIV: randomised controlled trial. BMJ 2007; 334:136.
16. Laufer MK, van Oosterhout JJG, Perez MA, Kanyanganlika J, Taylor TE, Plowe CV, Graham SM. Observational cohort of HIV-infected African Children. Pediatr Infect Dis J 2006; 25:623–627.
17. Collett D. Modelling survival data in medical research. 2nd edition London: Chapman & Hall; 2003.
18. Cole T. Babies, bottles, breasts: is the WHO growth standard relevant? Significance 2007; 4:6–10. Available at: http://www.blackwell-synergy.com/toc/sign/4/1
19. Cole TJ, Freeman JV, Preece MA. British 1990 growth reference centiles for weight, height, body mass index fitted by maximum penalized likelihood. Stat Med 1998; 17:407–429.
20. Durrleman S, Simon R. Flexible regression models with cubic splines. Stat Med 1989; 8:551–561.
21. Badri M, Lawn S, Wood R. Short-term risk of AIDS or death in people infected with HIV-1 before antiretroviral therapy in South Africa: a longitudinal study. Lancet 2006; 368:1254–1259.
22. HIV Paediatric
Prognostic Markers Collaborative Study and the CASCADE Collaboration. Current CD4 cell count and the short-term risk of AIDS and death in HIV-infected children and adults
. J Infect Dis
23. World Health Organization. CD4+ T-cell enumeration technologies. Technical information. Available at: http://www.who.int/diagnostics_laboratory/CD4_Technical_Advice_ENG.pdf
. Accessed May 2007.
24. Embree J, Bwayi J, Nagelkerke N, Njenge S, Nyange P, Ndinya-Achola J, et al
. Lymphocyte subsets in human immunodeficiency virus type 1-infected and uninfected children in Nairobi. Pediatr Infect Dis J 2001; 20:397–403.
25. Bunders M, Lugada E, Mermin J, Downing R, Were W, Thorne C, et al
. Within and between race differences in lymphocyte, CD4+, CD8+ and neutrophil levels in HIV-uninfected children with or without HIV exposure in Europe and Uganda. Ann Trop Paediatr 2006; 26:169–179.
26. Lepage P, Msellati P, Hitimana DG, Bazubagira A, Van Goethem C, Simonon A, et al
. Growth of human immunodeficiency type 1-infected and uninfected children: a prospective cohort study in Kigali, Rwanda, 1988 to 1993. Pediatr Infect Dis J 1996; 15:479–485.
27. Caulfield LE, de Onis M, Blossner M, Black RE. Undernutrition as an underlying cause of child deaths associated with diarrhea, pneumonia, malaria, and measles. Am J Clin Nutr 2004; 80:193–198.
28. Violari A, Cotton M, Gibb DM, Babiker A, Steyn J, Jean-Phillipe P, McIntyre J. on behalf of the CHER Study Team. Antiretroviral therapy initiated before 12 weeks of age reduces early mortality in young HIV-infected infants: evidence from the Children with HIV Early Antiretroviral Therapy (CHER) Study. Fourth International AIDS Society Conference on HIV Pathogenesis, Treatment and Prevention, 2007 [Abstract WESS103].
29. The Antiretroviral Therapy in Lower Income Countries (ART-LINC) Collaboration and ART Cohort Collaboration (ART-CC) groups. Mortality of HIV-1-infected patients in the first year of antiretroviral therapy: comparison between low-income and high-income countries
2006; 11; 367
30. Castleman T, Seumo-Fosso E, Cogill B. Food and nutritional implications of antiretroviral therapy in resource limited settings. Washington, DC: Food and Nutritional Technical Assistance Project, Academy for Educational Development, 2004.