The epidemiology of adolescents living with perinatally acquired HIV: A cross-region global cohort analysis

Background Globally, the population of adolescents living with perinatally acquired HIV (APHs) continues to expand. In this study, we pooled data from observational pediatric HIV cohorts and cohort networks, allowing comparisons of adolescents with perinatally acquired HIV in “real-life” settings across multiple regions. We describe the geographic and temporal characteristics and mortality outcomes of APHs across multiple regions, including South America and the Caribbean, North America, Europe, sub-Saharan Africa, and South and Southeast Asia. Methods and findings Through the Collaborative Initiative for Paediatric HIV Education and Research (CIPHER), individual retrospective longitudinal data from 12 cohort networks were pooled. All children infected with HIV who entered care before age 10 years, were not known to have horizontally acquired HIV, and were followed up beyond age 10 years were included in this analysis conducted from May 2016 to January 2017. Our primary analysis describes patient and treatment characteristics of APHs at key time points, including first HIV-associated clinic visit, antiretroviral therapy (ART) start, age 10 years, and last visit, and compares these characteristics by geographic region, country income group (CIG), and birth period. Our secondary analysis describes mortality, transfer out, and lost to follow-up (LTFU) as outcomes at age 15 years, using competing risk analysis. Among the 38,187 APHs included, 51% were female, 79% were from sub-Saharan Africa and 65% lived in low-income countries. APHs from 51 countries were included (Europe: 14 countries and 3,054 APHs; North America: 1 country and 1,032 APHs; South America and the Caribbean: 4 countries and 903 APHs; South and Southeast Asia: 7 countries and 2,902 APHs; sub-Saharan Africa, 25 countries and 30,296 APHs). Observation started as early as 1982 in Europe and 1996 in sub-Saharan Africa, and continued until at least 2014 in all regions. The median (interquartile range [IQR]) duration of adolescent follow-up was 3.1 (1.5–5.2) years for the total cohort and 6.4 (3.6–8.0) years in Europe, 3.7 (2.0–5.4) years in North America, 2.5 (1.2–4.4) years in South and Southeast Asia, 5.0 (2.7–7.5) years in South America and the Caribbean, and 2.1 (0.9–3.8) years in sub-Saharan Africa. Median (IQR) age at first visit differed substantially by region, ranging from 0.7 (0.3–2.1) years in North America to 7.1 (5.3–8.6) years in sub-Saharan Africa. The median age at ART start varied from 0.9 (0.4–2.6) years in North America to 7.9 (6.0–9.3) years in sub-Saharan Africa. The cumulative incidence estimates (95% confidence interval [CI]) at age 15 years for mortality, transfers out, and LTFU for all APHs were 2.6% (2.4%–2.8%), 15.6% (15.1%–16.0%), and 11.3% (10.9%–11.8%), respectively. Mortality was lowest in Europe (0.8% [0.5%–1.1%]) and highest in South America and the Caribbean (4.4% [3.1%–6.1%]). However, LTFU was lowest in South America and the Caribbean (4.8% [3.4%–6.7%]) and highest in sub-Saharan Africa (13.2% [12.6%–13.7%]). Study limitations include the high LTFU rate in sub-Saharan Africa, which could have affected the comparison of mortality across regions; inclusion of data only for APHs receiving ART from some countries; and unavailability of data from high-burden countries such as Nigeria. Conclusion To our knowledge, our study represents the largest multiregional epidemiological analysis of APHs. Despite probable under-ascertained mortality, mortality in APHs remains substantially higher in sub-Saharan Africa, South and Southeast Asia, and South America and the Caribbean than in Europe. Collaborations such as CIPHER enable us to monitor current global temporal trends in outcomes over time to inform appropriate policy responses.

Worldwide, around 330,000 infants were infected perinatally in 2011(1). These HIV-infected children now live longer because of greater access to cART and they are reaching adolescence. Perinatally HIV-infected adolescents are an emerging and growing population with specific features, observed initially in high-income countries (Europe and USA), then progressively in middle-income countries (Asia and South America) and, more recently, in low-income countries (sub-Saharan Africa) (2)(3)(4)(5)(6)(7)(8)(9)(10). So far, we have limited understanding of the patient-and regionallevel factors associated with long-term care of these HIV-infected adolescents in these different settings. A description of their socio-medical characteristics and their ability to remain in paediatric care, as well as research on the temporal trends of these characteristics and their consequences on transitioning to adult care, are also needed.
Within CIPHER, a meeting of paediatric cohorts and cohort collaborations was held in Venice in May 2013, during which the need for a global analysis of perinatally HIV-infected adolescents was identified. This forum and the associated collaboration provides a unique opportunity to bring together multiple, large data sources to better describe the global epidemiology of adolescent HIV and compare programme and clinical outcomes in both regional-and individual-level contexts.

Key aims/research questions Objectives
The primary objective of this study is to describe the global epidemiology, geographical and temporal trends of characteristics of perinatally HIV-infected adolescents according to baseline variables at entry into HIV care, at cART initiation, and at entry to adolescence (i.e., from the age of 10 years). These variables will capture demographic data, clinical status, antiretroviral therapy outcomes and laboratory monitoring, by region, site and study.
The secondary objective of this study is to evaluate the HIV disease status (CDC/WHO stage), and rates and predictors of death, loss to follow up (LTFU), loss to programme (death or LTFU) and transition to adult care outcomes from the age of 10 (baseline time for entry into the adolescent cohort) by regional-, site-, study-and individual-level characteristics.
Study design: This will take the form of an international pooled analysis of individual-level retrospective data from eligible paediatric cohort collaborations and individual cohorts, including: EPPICC, MSF, Optimal Models/ICAP, IMPAACT, PHACS, Baylor and IeDEA (East Africa, West Africa, Central Africa, Southern Africa, Asia-Pacific and CCASAnet).

Justification for use of CIPHER collaboration
Within CIPHER, a meeting of paediatric cohorts and cohort collaborations was held in Venice in May 2013, during which the need for a global analysis of perinatally infected adolescents was identified. This forum provides a unique opportunity to describe the global epidemiology of adolescents infected perinatally with HIV and to compare their retention in care and transition to adulthood care within both regional-and individual-level contexts.

Study population, including estimated sample size if possible
Cohort eligibility criteria: Cohorts must have data available in an electronic database in 2013 in order to participate.

Patient eligibility criteria:
Participants will be HIV-infected children entering HIV care at any age before 10 years of age (proxy for perinatal infection) and followed beyond the age of 10 (follow-up data would be included up to age 19 years, encompassing the period of adolescence as defined by WHO).

Exclusions:
• Children with known horizontally acquired infection.

Data required Key variables and definitions:
Not all variables will be available from all cohorts. Viral load data may be missing for several cohorts.
Potential overlaps would have to be identified at the regional level before data transfer to the data management center.  Baseline characteristics at the age of 10 (including history of HIV care) will be reported as median (IQR) for continuous variables and counts (percentage) for categorical variables.
All the above descriptions will be done globally, then stratified by calendar at first visit period (before 2004; 2004-2008; 2009-2013) and by region (geographical and according to the income status of countries). Comparisons will be carried out for some of the characteristics of interest using Chi-square, ANOVA methods or non-parametric equivalence.
• Time to event analysis using Kaplan-Meier survival analysis with delayed entry given survival to the age of 10 years (adapted for left truncation), global, and stratified by region and by time periods: o To ART initiation (for those not starting before 10 years of age) o To death o To transfer out to adult care o To loss to follow-up • A Cox proportional hazard model analysis from the age of 10 will be conducted to analyse the predictors of the study outcomes (death/LTFU/transfer to adult care). This analysis will be done using a competing risk analysis between death and LTFU.
• • Multivariate analyses will be based on the Cox proportional hazards model to summarize predictors associated with survival and retention in care and using a stepwise descendant approach, including variables that were selected by univariate analyses at the 25% level. When explanatory variables are available, the Poisson model and the Cox model with delayed entry may be used for estimating relative risks (11).
Mixed models will be used to account for clustering at multiple levels (country, site/city and, possibly, time period). Sensitivity analyses will be conducted using different LTFU definitions and ways of dealing with missing data: exclusion, inclusion as a missing category, or inclusion using missing data imputation. Ethical issues This study will use site-level data and retrospective data from individual cohorts with previous IRB approval for data collection and analysis. Only anonymized data will be transferred to the data management centre.