We analyzed participants in the Reaching for Excellence in Adolescent Care and Health (REACH) study. The characteristics of the cohort, recruitment, follow-up and observational research objectives are described in detail elsewhere [8
]. Briefly, HIV-1 seropositive and high-risk seronegative adolescents (13 to 19 years old) were recruited from 15 clinical sites in 13 US cities. Participants were followed on a quarterly basis for basic science and clinical evaluations, including documentation of demographic and risk behaviors, collection of medical history and various biological samples, along with tests for HIV-1 infection, other sexually transmitted infections, CD4+
cell counts, and immunological outcomes. HIV-1 RNA concentration (viral load) and immunological outcomes were measured every three months. CD4+
lymphocytes were quantified by flow cytometry in National Institute of Allergy and Infectious Disease (NIAID)-certified laboratories at the clinical sites. Plasma viral load was measured in a centralized laboratory using either nucleic acid sequence-based amplification (NASBA, lower limit = 400 copies/mL) or NucliSens assays (Organon Teknika, Durham, NC, lower limit = 80 copies/mL). Viral loads under the lower limit of detection were assigned half of log10
(lower detection limit) values for each respective visits. All individuals included in the analysis had multiple measurements (≤ 4 visits) of HIV-1 RNA (transformed to log10
for normalization), CD4+
count, and CD8+
count at treatment-free visits.
Our analyses focused on a subset of African American adolescents [8
] who were free of clinical AIDS and not receiving any antiretroviral therapy (ART) at multiple 3-month follow-up intervals (total = 396 person-visits). All participants had acquired HIV-1 infection through sexual activity or injection drug use. Although these relatively young patients were HIV-1 seropositive at baseline, vertical transmission could be excluded and seroconversion was deemed recent as judged by i) self-reported sexual and other risk behaviors, ii) high CD4+
count and iii) relatively stable viral load that could be treated as a proxy for viral load set-point [8
]. We estimated the viral load set-point based on observations from 2–4 sequential visits, which is consistent with earlier reports that it is cost-effective to estimate the viral set-point based on one or two measurements obtained between 5 and 12 months after HIV-1 infection [10
High-resolution HLA class I (HLA-A, HLA-B and HLA-C) typing was performed by a combination of molecular techniques, including automated DNA hybridization with sequence-specific oligonucleotide (SSO) probes, PCR with sequence-specific primers (SSP), and automated sequencing-based typing (SBT). SNP genotyping was performed by the allelic discrimination TaqMan assay (Applied Biosystems). We tested for consistency with Hardy-Weinberg equilibrium expectation using one-way goodness-of-fit chi-square. To confirm the specificity of the rs9264942 genotyping (due to several SNPs in the region), we sequenced the regions (flanking 250 bp each side) in a selected subset of African American individuals from the entire REACH cohort (19 with B*5703, 4 with B*5701, and 45 with other HLA-B genotypes). We assessed patterns of linkage disequilibrium (LD) of rs9264942 along with other variants in the sequenced region with HLA-B alleles (specifically B*5701 and B*5703) and alleles of HLA-A and HLA-C.
Among those participants with multiple clinical visits in the study, we excluded the measurements from the visits where the subsequent viral loads did not appear to be in steady state, i.e. inter-visit viral load variation exceeded 3-fold (0.5 log10
) copies per ml of plasma. Individuals with no more than two steady state viral loads were excluded from the study. We used Proc Mixed in SAS version 9.2 (SAS Institute, Cary, NC) to fit the uneven repeated measurement ANOVA model and estimated mean viral load [11
] by genetic variant (allele or genotype) after accounting for confounding covariates and multiple comparisons using Tukey-Kramer adjustment. To account for the differences in the evaluation between viral load observations which are close to each other than to those which are further apart we used autoregressive covariance structure. This method is powerful in providing the precision of the mean and the standard error of the viral set-point with such uneven viral load data over uneven period of time. We also performed alternate analysis by first estimating the viral load set point as the average of multiple viral-load measurements and used the standard linear regression model. Additionally, we used the mixed model to assess how genotype influenced CD4+
over time, post viral load set-point, as surrogates for treatment free HIV progression. We also determined the allelic and genotypic heterogeneity among 19 “controllers” (VL<1000 copies/ml and CD4 >450 × 106
cell/l), 23 “non-controllers” (VL>16,000 copies/ml and CD4 <450 × 106
cell/l) and 79 “intermediates”, as previously described [12
]. Categorical outcomes were assessed by χ2
and Cochran-Armitage trend tests and the continuous outcomes (log10
viral load and absolute CD4+
) by t-tests and F-tests.