Long-Term Clinical, Virological and Immunological Outcomes Following Planned Treatment Interruption in HIV-Infected Children

R Freguja; A Bamford; M Zanchetta; P Del Bianco; C Giaquinto; L Harper; A Dalzini; TR Cressey; A Compagnucci; Y Saidi; Y Riault; D Ford; D Gibb; N Klein; A De Rossi

Disclosures

HIV Medicine. 2021;22(3):172-184. 

In This Article

Methods

Our methods were as previously described.[8–10] HIV-infected children were eligible if they had been on three-drug ART for at least 24 weeks, had HIV-1 RNA < 50 copies/mL and two recent CD4% of at least 30% (age 2–6 years), or, alternatively, a CD4% of at least 25% and a CD4 cell count of at least 500 cells/μL (age 7–15 years). Randomization was stratified by age of ART initiation, age and lowest recorded pre-ART CD4. Patients were followed until the last randomized child had completed 72 weeks' follow-up.

The main trial ended in May 2008 and ART re-initiation was recommended in the PTI arm. Routine annual data (ART, AIDS events, important clinical events, weight/height, CD4 and CD8 count/percentage, HIV-1 RNA viral load) were collected for 5 years from end of the main trial. The last data from long-term follow-up were collected in April 2014. Full ethics committee approval was obtained at each participating centre (summarized in previous reports[8–10]). Informed consent was obtained at the main study entry and additional consent was obtained at the beginning of long-term follow-up.

In centres able to separate and store cells, 10 mL whole blood was collected in EDTA annually. Peripheral blood mononuclear cells (PBMCs) were isolated by density gradient centrifugation and frozen for storage. Children with at least two PBMC samples available during 5 years' follow-up were included in the sub-study aimed at providing in-depth cell phenotyping and virological data for exploratory analysis.

Sub-study Laboratory Methods

HIV-1 DNA Quantification. Cell-associated HIV-1 DNA was measured in PBMC by real-time polymerase chain reaction (PCR), as previously described.[11] HIV-1 copy number was normalized against the housekeeping gene telomerase reverse transcriptase (TERT) and results were expressed as HIV-1 DNA copies/106 PBMCs.[12]

Intracellular HIV-1 RNA Quantification. RNA was extracted from 3 × 106 PBMCs using Trizol Reagent (Invitrogen, Carlsbas, CA, USA). Trizol (500 μL) and quantitation standard (7 μL; internal control; Roche Diagnostic Systems, Branchburg, NJ, USA) were added to PBMCs. Samples were incubated with 200 μL chloroform for 15 min on ice. After centrifugation, RNA was stored at −20°C overnight with cold isopranolol. Samples were centrifuged and supernatant removed. Each RNA pellet was resuspended in 75 μL of elution buffer heated to 70°C. HIV-1 RNA levels were determined by real-time PCR with the Amplicor HIV-1 Monitor Test using Cobas TaqMan48 (Roche Diagnostic Systems).[13,14]

Immunophenotyping. Aliquots of frozen samples were thawed and cell suspensions incubated for 15 min in the dark with labelled monoclonal antibodies (Beckton Dickinson Bioscience Pharmingen, San Diego, CA, USA): anti-CD3 fluoroscein isothiocyanate, anti-CD4 peridinin chlorophyll protein (PerCP), anti-CD8 PerCP, anti-CD38 phycoerythrin (PE), anti-CD45RO allophyocyanin (APC) and anti-CD45RA APC. Appropriate isotype controls (mouse IgG1-PE and mouse IgG2b-APC) were used to evaluate non-specific staining. Cells were washed with Automacs Buffer (Miltenyi Biotec Inc., Auburn, CA, USA) and resuspended in phosphate-buffered saline supplemented with 1% paraformaldehyde. Samples were analysed by four-colour flow cytometry using a Calibur fluorescence-activated cell sorter (Beckton Dickinson) equipped with a 488-nm argon-ion laser and a 635-nm red diode laser. In all, 50 000 events were collected in the lymphocyte gate according to forward and side scatter.[15] Data were processed using CellQuest Pro Software (Becton Dickinson) and analysed using Kaluza software v.1.2 (Beckman Coulter).

TREC Quantification. Thymic output in PBMCs was assessed using T-cell receptor excision circle (TREC) levels measured by real-time PCR,[11,16] expressed as TREC copies/105 PBMCs.[16,17]

Statistical Methods

Main Study. Laboratory measurements at end of main trial were defined as those nearest to but no more than 3 months prior to 29 May 2008 and annually (up to end of the fifth year) thereafter as those taken nearest to but within ±3 months of these respective time points.

Comparison between the two randomized arms was according to intention-to-treat, adjusting for stratification factors in regression analyses. Rates of clinical events from end of main trial to last clinic visit were compared using a random-effects Poisson model allowing more than one event per child. Virological suppression < 50 copies/mL and CD4/CD8 ratio < 1 were analysed using Poisson regression with robust error variance. CD4 and CD8 percentages and counts and CD4/CD8 ratio were analysed using linear regression, adjusting for values at randomization. Mixed-effects models were fitted over time, including all measurements in the CT arm and measurements after restarting treatment after the last PTI in the PTI arm. Cubic splines were used to investigate the shape of the curves over time by arm; curves were then approximated by a linear model in the CT arm and a piecewise linear model in the PTI arm, allowing for a different slope for the first 3 months after restarting continuous treatment. For continuous outcomes, heteroscedastic random effects were modelled by arm with random effects for intercept and slope in the CT arm (unstructured covariance matrix), random effects for intercept and slope in the PTI arm (unstructured covariance matrix) and an additional random effect for the slope in the first 3 months after restarting treatment in the PTI arm. For virological suppression < 50 copies/mL, two random effects were included, one for each arm; there was too little variability by slope to include additional random effects. Time 0 was defined as the end of the main trial (prior measurements were given negative time). Models were extrapolated as necessary to estimate where curves would cross (after those in the PTI arm had been back on continuous treatment for at least 3 months), indicating that laboratory values in the PTI arm had reached those in the CT arm. Two-sided 95%-based confidence intervals of the difference between arms at different time points from the end of main trial and for the time point where the PTI and CT curves would cross were obtained using 1000 bootstrap samples.

Based on all measurements of CD4/CD8 ratio in the PTI arm after each child restarted ART following the most recent PTI, effects of the following factors on CD4/CD8 ratio recovery were assessed: baseline characteristics, CD4/CD8 ratio at ART re-initiation, RNA plasma viraemia at ART re-initiation, nadir CD4% and number of PTIs. Mixed-effects models were fitted as described earlier, although time 0 was defined as the time of restarting continuous ART; effects of predictors were similar when time 0 was defined as the end of the main trial, as previously (data not shown). All models were adjusted for time since restarting continuous ART and baseline (trial entry) CD4/CD8 ratio. The final multivariable model included all factors significant at P < 0.05 in univariable analyses. Current RNA was then added to the final model as a surrogate for recent adherence to ART. Data were analysed using STATA v.13.

Sub-study. This was an exploratory analysis using samples available from children followed up in centres able to store PBMCs. Results were included if a child had at least two samples available in the 5-year period following the end of the main trial. Samples were assigned to their nearest annual time point to maximize the number of samples included.

Mixed-effects linear models were used to examine change over time in binary and continuous outcomes within and between CT and PTI groups, using mixed-effects Poisson and linear regression models, respectively, as for the main trial. The models included as explanatory variables the number of years since the end of the main trial, as a categorical variable, the treatment group and their interaction, in addition to a compound symmetry covariance structure for the random effects for intercept and slope. Data were analysed using SAS v.9.2 and STATA v.13.0.

Comments

3090D553-9492-4563-8681-AD288FA52ACE

processing....