Introduction

Children with HIV (CHIV) require appropriate dosing of pediatric antiretroviral therapy (ART) to prevent HIV-related morbidity and mortality. HIV programs at national and international levels procure annual quantities of pediatric ART based on forecasts of the number of children likely to require each drug formulation and dose; dosing is weight-based in children < 15 years of age. Accurate forecasting of antiretroviral needs is particularly important, because the pediatric antiretroviral market is small. Procurement volume has a large impact on market prices and failure to procure sufficient quantities of antiretroviral medications (ARVs) leads to gaps in treatment access in low-income and middle-income countries [1]. While validated estimates exist of the numbers of children at each age likely to need ART [2], additional information about the weight distribution for CHIV by age are needed to permit accurate forecasts of the medication quantities that will be required. To date, ART forecasts rely on WHO and CDC growth curves for all children in the general population [3,4,5]. However, children and adolescents living with HIV are often malnourished at the time of initiating ART and their catch-up growth can be delayed even after initiating ART [6,7,8,9]. Because of these key differences for CHIV, growth curves derived from the general population likely overestimate weight-for-age among CHIV and therefore may lead to inaccurate estimates of the number of formulations and doses of pediatric ART that will be required globally [10]. Accurate data about volume trends in weight-for-age evolution among ART-treated CHIV are therefore needed to appropriately inform ART forecasting efforts [11].

This study reports a secondary analysis of data from a parent study conducted within the global International epidemiology Databases to Evaluate AIDS (IeDEA) pediatric research consortium (https://www.iedea.org/). The parent study sought to analyze age- and CD4-stratified risks of opportunistic infections and mortality for CHIV before and after ART initiation in six global IeDEA regions [12]. Using the database assembled for the parent study, we conducted a separate analysis, reported here, to evaluate the weight-for-age distributions in CHIV on ART in the IeDEA consortium in order to support accurate forecasting and procurement of pediatric ART formulations.

Main text

Methods

We analyzed individual patient data from the six regional pediatric cohorts within IeDEA: Asia–Pacific, West Africa, East Africa, Central Africa, Southern Africa, and the Caribbean, Central, and South America network (CCASAnet) [13]. Using eligibility criteria from the parent study [12], all patients were included if they enrolled into care at age < 24 years, were followed-up at any of the participating IeDEA sites between 2004 and 2016, had a confirmed HIV diagnosis, were ART-naïve at enrolment, and had at least one CD4 count or percent measurement during follow-up. In the current analysis, we limited the dataset to children aged < 15 years, because older youth primarily use adult ARV formulations and doses. Although clinic protocols vary across the IeDEA consortium, CHIV are usually seen at least every 3 months while on ART. The data were generated during routine care encounters and included region, country, site, patient demographics (sex, date of birth, date of HIV diagnosis if available, and date of enrolment in care), laboratory values (CD4 count, CD4 percent), date of ART initiation, initial ART regimen, date of death, date of last clinical contact, and date of transfer out. Each participating IeDEA region obtained local institutional review board approval to participate. Written informed consent requirements were deferred to the local institutional review boards. The analysis only used de-identified data that had been collected as part of routine clinical care.

For the current study, we used data on weight at each visit as recorded in the medical record. For each one-year age stratum from 0 to 15 years, we calculated the number of children in each of the weight bands designated by the World Health Organization (WHO) to be relevant to pediatric ART formulations: 0 to  < 3 kg, 3 to  < 6 kg, 6 to  < 10 kg, 10 to  < 14 kg, 14 to  < 20 kg, 20 to  < 25 kg, 25 to  < 30 kg, 30 to  < 35 kg, 35 to  < 40 kg, 40 to  < 45 kg, 45 to  < 50 kg, 50 to  < 55 kg, 55 to  < 60 kg, and ≥ 60 kg. We derived these numbers, and resulting proportions, for the entire cohort of CHIV, total and stratified by sex; we additionally derived numbers and proportions stratified by sex and IeDEA region, calendar year of enrolment (before versus beyond January 1, 2013), and time on ART (< 12 months or ≥ 12 months). Children for whom sex was unknown were excluded from the analysis. Age at each visit was calculated based on the date of birth recorded in the database. A single child contributed multiple weight measurements over time and we deleted outlying measurements based on the following criteria: (1) weight-for-age z-scores (WAZ) > 3 for infants aged < 1 year, WAZ < − 8 for children aged < 5 years, WAZ < − 6 for children aged 5–10 years, and WAZ > 5 for all ages; (2) two values measured within the same month with a difference greater than 5 kg if age < 5 years, 8 kg if age 5 to  < 10 years, and 10 kg if age 10–15 years.

Results

Overall, 59,862 children and adolescents with HIV contributed to the analysis, of whom 50.7% were females. Demographic information is shown in Table 1. Age and weight data were available from 530,080 clinical encounters for girls and 537,894 clinical encounters for boys. Sex-stratified results for all CHIV in the study are shown in Table 2. Additional tables are included in Additional file 1:

  1. 1.

    Overall population (combined).

  2. 2.

    By sex (Table 2).

  3. 3.

    By IeDEA region, boys.

  4. 4.

    By IeDEA region, girls.

  5. 5.

    By ART duration and sex.

  6. 6.

    By year of visit and sex.

  7. 7.

    By IeDEA region, ART duration, boys.

  8. 8.

    By IeDEA region, ART duration, girls.

  9. 9.

    By IeDEA region, ART duration, < 2013, Boys.

  10. 10.

    By IeDEA region, ART duration, < 2013, Girls.

  11. 11.

    By IeDEA region, ART duration, ≥ 2013, Boys.

  12. 12.

    By IeDEA region, ART duration, ≥2013, Girls.

Table 1 Characteristics of the 59,862 children in IeDEA at baseline
Full size table
Table 2 Weight-for-age distribution of CHIV aged 0–15 in the IeDEA global consortium (N = 59,862 children from 2004 to 2016)
Full size table

Limitations

This study is subject to a number of limitations. Several characteristics of patients participating in the IeDEA cohort may lead to overestimation of the weight-for-age distribution of children receiving ART and thus may overestimate the amount of active pharmaceutical ingredient required per child. The first is survivorship bias, because CHIV who survived to initiate ART at IeDEA sites from 2004 to 2016 may have been healthier (and perhaps with higher weights for age) than those who did not. Children in this cohort do not represent a complete cohort of CHIV followed from birth, and there was likely a high risk of death in children before the age of 2 years who therefore never entered the study. Similarly, a high proportion of CHIV were lost to follow-up during the study period, possibly leaving a healthier cohort to remain in care and with evaluable weights and ages. Nevertheless, these data likely accurately reflect weight-for-age among children who remained on treatment and for whom ARV procurement estimates currently apply, even if they may overestimate weight for the full population of CHIV in need of ART. On the other hand, approximately two-thirds of IeDEA sites are located at university hospitals in capital cities, where the standard of care may be different than in rural areas; if more intensive HIV care is provided, children at IeDEA sites may be healthier than at more rural sites in the same regions [13]. We also note a potential selection bias induced by the parent study, which only included children with available CD4 counts during follow-up, thus excluding the sicker children who may have initiated ART immediately, without their caregivers waiting for CD4 counts. This may also be possibly overestimating weights for ART-treated CHIV.

A second set of limitations may counteract these trends and may underestimate the quantities of active pharmaceutical ingredients needed. For example, before ART was recommended for all CHIV in 2013, it was offered only to the sickest children, potentially introducing an indication bias. This selection bias may particularly apply to older children in the IeDEA cohort, as access to early ART has improved over time for children, so adolescents aged 10–15 years in the future may be in better health than those observed in the IeDEA cohort over the study period reported in this study [14].

Despite these limitations, this dataset is, to our knowledge, the only available source of data about weight-for-age distributions among CHIV treated with ART built on the strength of a large cohort size and broad geographic coverage. Consequently, the resulting estimates and those, more generally, derived from observational cohorts conducted in real-world settings, offer a valuable opportunity to derive data-driven weight-for-age distributions among CHIV [15, 16].

These data were originally pooled to support adequate forecasting and procurement of pediatric ART formulations. Nevertheless, beyond this purpose, data about weight in children may also be helpful to create a secondary-hand framework for observing changes overtime regarding growth on ART, during childhood and adolescence as this has been done in the past [7, 11]. Especially, these data will be useful in refining projections of the number of CHIV who may need specific ARV formulations and doses allowing HIV care and treatment programs to procure sufficient quantities of medications for CHIV, avoiding both medication stock outs and medication wastage, furthering the critical goal of improving access to ART for CHIV globally [1].