Predicting patterns of long-term CD4 reconstitution in HIV-infected children starting antiretroviral therapy in sub-Saharan Africa: a cohort-based modelling study

Abstract

Background: Long-term immune reconstitution on antiretroviral therapy (ART) has important implications for HIV-infected children, who increasingly survive into adulthood. Children's response to ART differs from adults', and better descriptive and predictive models of reconstitution are needed to guide policy and direct research. We present statistical models characterising, qualitatively and quantitatively, patterns of long-term CD4 recovery.

Methods and findings: CD4 counts every 12 wk over a median (interquartile range) of 4.0 (3.7, 4.4) y in 1,206 HIV-infected children, aged 0.4-17.6 y, starting ART in the Antiretroviral Research for Watoto trial (ISRCTN 24791884) were analysed in an exploratory analysis supplementary to the trial's pre-specified outcomes. Most (n = 914; 76%) children's CD4 counts rose quickly on ART to a constant age-corrected level. Using nonlinear mixed-effects models, higher long-term CD4 counts were predicted for children starting ART younger, and with higher CD4 counts (p<0.001). These results suggest that current World Health Organization-recommended CD4 thresholds for starting ART in children ≥5 y will result in lower CD4 counts in older children when they become adults, such that vertically infected children who remain ART-naïve beyond 10 y of age are unlikely ever to normalise CD4 count, regardless of CD4 count at ART initiation. CD4 profiles with four qualitatively distinct reconstitution patterns were seen in the remaining 292 (24%) children. Study limitations included incomplete viral load data, and that the uncertainty in allocating children to distinct reconstitution groups was not modelled.

Conclusions: Although younger ART-naïve children are at high risk of disease progression, they have good potential for achieving high CD4 counts on ART in later life provided ART is initiated following current World Health Organization (WHO), Paediatric European Network for Treatment of AIDS, or US Centers for Disease Control and Prevention guidelines. In contrast, to maximise CD4 reconstitution in treatment-naïve children >10 y, ART should ideally be considered even if there is a low risk of immediate disease progression. Further exploration of the immunological mechanisms for these CD4 recovery profiles should help guide management of paediatric HIV infection and optimise children's immunological development. Please see later in the article for the Editors' Summary.

Figure 1. An illustrative representation of the mathematical model used to describe CD4 reconstitution.

Figure 2. Modelling CD4-for-age in children showing asymptotic reconstitution.

(A) Mean CD4-for-age over time since ART initiation for 914 children with asymptotic CD4 reconstitution. Errors bars give standard error of the mean. Points are shown for times where >20% of children in this group remained in the trial. Only children participating in a substudy had measurements available at week 4, accounting for the wider error bar. The numbers of children with measurements available at 0, 4, 12, 24, 36, 48, 60, 72, 84, 96, 108, 120, 132, 144, 156, 168, 180, 192, 204, 216, and 228 wk were 914, 153, 897, 893, 900, 893, 894, 898, 989, 900, 902, 898, 899, 897, 896, 892, 861, 790, 516, 354, and 232, respectively. (B) Population-average (fixed-effects) ln(CD4-for-age) trajectories predicted by our nonlinear mixed-effects model (Table 2) for “average” HIV-infected children (female, weight-for-age z-score at ART initiation −2.2) aged 2 (dashed line), 6 (solid line), and 10 (dot-dashed line) y. (C) Predicted CD4-for-age trajectories as shown in (B), but plotted on the linear scale (i.e., CD4/expected CD4). (D) Correlations between child-level random effects (difference from population average) in the three model parameters. Each panel refers to correlations between the estimated random effects (differences between child-specific parameters and the population average) in two of the three parameters int (CD4-for-age at ART initiation), asy (long-term CD4-for-age), and c (proportional recovery rate). The pair of parameters being considered is identified by the row and column labels in the panels on the diagonal. The upper-right panels plot the values of the random effects against one another. The lower-left panels give Pearson correlations between the estimated random effects, with 95% confidence intervals and p-values.

Figure 3. Predicted long-term CD4 counts in children starting ART at different ages and CD4 levels.

(A) CD4 trajectories predicted for children starting ART having reached WHO CD4 count thresholds at age 2 (dashed line), 6 (dotted line), or 10 (dot-dashed line) y. The thin dashed line indicates WHO thresholds for ART initiation, and the thin solid line the trajectory in CD4 count with age expected in a healthy child. (B) Expected CD4 count on immunological maturity (estimated at age 20 y) for different ages and CD4 counts at ART initiation. Values at the ends of the grey contour lines indicate expected adult CD4 count in children starting ART at the ages and CD4 counts given on the horizontal and vertical axes. The black line indicates the current WHO CD4 thresholds for ART initiation. Grey point markers show the age and fitted CD4 count at ART initiation of the 914 children on whom the model is based. They indicate at which ages/CD4 counts the model has substantial evidence, and where it represents an extrapolation from the available data.

Figure 4. CD4-for-age in children not showing asymptotic reconstitution.

Mean CD4 cell count for age with time since ART initiation (error bars: standard error of the mean) for: (A) 23 children with significant (p<0.05) decreasing CD4-for-age with time; (B) 79 children with significant (p<0.05) increasing CD4-for-age with time but not fitting the asymptotic model; (C) 153 children with no significant change (p>0.05) in CD4-for-age with time; and (D) 37 children with only baseline CD4 count available. For comparison to the asymptotic group, grey lines indicate average (fixed-effects, for a 6-y-old girl with weight-for-age z-score −2.2) CD4-for-age in the 914 children with asymptotic CD4 reconstitution. In (A–C), data points are shown where >20% of children in each group remained in the trial.