Antiretroviral therapy (ART) reduces HIV-1-related morbidity and mortality in children but does not prevent the establishment of a persistent replication-competent HIV-1 reservoir [1,2]. The time to rebound viremia following ART discontinuation is influenced by the size of the latent HIV reservoir . Multiple studies have shown that the size of the HIV-1 reservoir in children is dependent on the timing of effective ART initiation (<3 months of age) and the rapidity with which HIV-1 replication is controlled [4–8], with shorter time to virologic control [5,6] associated with reduced levels of circulating cell-associated HIV-1 DNA. Decay of circulating HIV-1 DNA levels over time has also been reported in early-treated children and is thought to be because of either loss or dilution of HIV-1-infected CD4+ T cells over time . Some of the lowest reported levels of circulating cell-associated HIV-1 DNA have been reported in children who initiate ART within hours of birth  or in children with over 10 years of viral suppression (viral suppression) after early ART [6,11]. Although blood levels of cell-associated HIV-1 DNA overestimate the size of the replication-competent HIV-1 reservoir that serves as the barrier to cure, they remain the most feasibly measured approximation of the HIV-1 reservoir size in children.
Although early ART initiation preserves immune function [12,13], many early-treated children lack persistent HIV-1 antibody responses measured by standard HIV-1 ELISA assays during routine clinical testing [13,14], often leading to questions in later years about the initial diagnosis. As the generation and persistence of antibodies is dependent on the magnitude and duration of antigenic exposure, we hypothesized that HIV-1 antibody levels might be useful predictors of HIV-1 DNA levels in ART-suppressed children. Indeed, we  and others  have previously described lower levels of cell-associated HIV-1 DNA in children with negative or indeterminate western blots. In a recent study of children who initiated ART prior to age 2 years and were studied through age 4 years, antibody levels to p31 (integrase) and p17 (gag) were significantly associated with HIV-1 RNA levels whereas antibody levels to HIV-1 envelope proteins (gp160 and gp41) were significantly associated with cell-associated HIV-1 DNA levels . Altogether, these data suggest that HIV-1 antibody levels might be useful to predict the size of the HIV-1 DNA reservoir in children who have achieved sustained viral suppression [16,17].
In this study, we examined HIV-1 antibody decay trajectories after sustained viral suppression on ART, and further evaluated the utility of using HIV-quantitative antibodies as a screening test for low HIV-1 DNA reservoir size in children and adolescents with perinatal HIV-1 infection (PHIV), and in whom longitudinal analyses of proviral reservoir decay was previously characterized . We used two prediction-based approaches to identify the most significant antibodies for predicting low HIV-1 DNA levels and to quantify the discriminative ability of a model for low HIV-1 DNA levels including all antibody levels as predictors.
This study utilized 514 plasma specimens from 61 perinatally infected study participants living with HIV and enrolled in the Pediatric HIV/AIDS Cohort Study Adolescent Master Protocol (PHACS AMP) longitudinal follow-up cohort study . Study participants achieved viral suppression (HIV-1 plasma levels <400 copies/ml) at or before 5 years of age on ART and maintained virologic control (allowing for isolated viral loads ≥400 copies/mL [i.e. blips]) as of their last specimen before 2 July 2012. PHACS AMP was designed to evaluate the impact of HIV-1-infection and ART on youth with PHIV. Four hundred and fifty-one youth with PHIV from 15 study sites in the United States and Puerto Rico were enrolled if they were 7 to 16 years of age, and had complete medical history of ART use, plasma HIV-1 viral load concentrations, and lymphocyte subset measurements since birth. PHACS AMP was approved by the institutional review board at the Harvard T.H. Chan School of Public Health and at each participating site. Written informed consent was obtained from each participant's parent or legal guardian. Assent was obtained from child participants according to local institutional review board guidelines. Among the PHACS AMP participants with PHIV, there were 242 specimens where data was available for both PBMC-associated HIV-1 DNA (measured as previously described ) and plasma to measure HIV-1 antibody levels. A total of 78 specimens from 20 seroreverters who were perinatally HIV-1-exposed but uninfected (PHEU) from a previously described cohort were also utilized as controls for analyses of antibody levels from birth through age 2 years . Study of PHEU participants beyond 2 years of age was not possible as they were discharged from routine follow-up for HIV-1 infection if found to be uninfected by 2 years of age.
Quantitation of HIV-1-specific antibodies
Antibody levels to HIV-1 envelope (gp160, gp41), gag (capsid, p24; matrix, p17); RT (p66/51), and integrase (p31) were quantified by ELISA, as previously described . The limits of detection (bolded) and dynamic ranges are as follows: p24: 0.1 μg/ml, 0.15–125 000 μg/ml; gp41: 0.4 μg/ml, 0.6–50 000 μg/ml; gp160: 0.1 μg/ml, 0.2–50 000 μg/ml; and RT: 0.1 μg/ml, 0.2–3100 μg/ml; p31: 0.06 OD405 nm, 0.299–2.000 OD405 nm; p17: 0.06 OD405 nm, 0.152–2.000 OD405 nm. The lower limit of the dynamic range was used as the lowest possible value in all calculations.
Quantification of HIV-1 DNA
Cell-associated HIV-1 DNA levels were quantified as previously described, using droplet digital PCR and expressed as HIV-1 DNA copies per million peripheral blood mononuclear cells (PBMCs) .
Locally Estimated Scatterplot Smoothing (LOESS) plots were used to qualitatively visualize HIV-1 antibody trajectories, including any decay, after viral suppression among participants with PHIV by age at viral suppression (<1 vs. 1–5 years)  and PHEU controls. LOESS plots depict a moving average of a series of smaller polynomial regressions of data providing an understanding of the general trajectory the data has across values of the x-axis, thereby aiding in the development of more formally specified models. A smoothing parameter can be specified to result in a more or less jagged trajectory line. The resulting LOESS plot depicts antibody level changes over time in a generally qualitative manner.
Antibody distributions were then compared by two thresholds for low HIV-1 DNA (<100 and <10 copies per million PBMCs) using linear models with generalized estimating equations (GEE). Gp41, gp160, p24, and RT were log10 transformed for analyses.
Two approaches were used to identify the most predictive antibodies for low HIV-1 DNA levels and to quantify the discriminative ability of a model for low HIV-1 DNA levels that included all antibody levels as predictors. For both approaches, we utilized the two thresholds for low HIV-1 DNA (<100 and <10 copies per million PBMCs) and two analysis populations (the full study population with matched antibody and HIV-1 DNA data, and a population restricted to study participants with no viral load blips).
The first prediction-based approach utilized receiver operator curve (ROC) analyses to identify the cutoff value for each individual antibody that maximized its sensitivity and specificity for low HIV-1 DNA levels. Quantitative antibody results were then dichotomized at their respective cutoff value and included as independent variables in individual logit-linked binomial repeated measures (GEE) models for low HIV-1 DNA. We then developed scores for each antibody based on the integer value of the resulting odds ratio estimates. These scores were then applied to any participant observation where the antibody level was at or below the ROC-determined cutoff. Observations that were above this cutoff were scored as a zero. The discriminative ability of this scoring system, as measured by the C-statistic, was then determined by including each applied antibody score as independent variables in GEE models for low HIV-1 DNA. Ranking of the individual C-statistics from these models identified the most predictive antibodies for low HIV-1 DNA. To evaluate the combined discriminative ability of the HIV-1-specific antibodies, we utilized a stepwise approach, adding the score of an antibody one at a time based on the ranking of individual C-statistics. At each step, scores were summed at the individual level and combined C-statistics were produced until all six antibodies had been included.
Our second approach utilized the Random Forests method, which outputs a variable importance plot ranking the importance of inputted antibody levels for prediction of low HIV-1 DNA levels and estimates of area under the curve for low HIV-1 DNA levels including data on all antibody levels. This ensemble learning approach utilizes robust prediction algorithms and reduces the risk of overfitting training data .
Analyses were conducted with SAS Version 9.4 (Cary, North Carolina, USA) and R version 3.2.2 with the Random Forest package version 4.6-14.
Sixty-one youth living with PHIV, born between 1991 and 2002, were studied from a median age of 4.5 years (IQR: 1.9--6.8) to 12.8 years (IQR: 11.0--14.5) (Table 1). Among the 13 children who achieved viral suppression before 1 year of age, the median age at viral suppression was 8 months (IQR: 8--10 months); among the 48 children who achieved viral suppression between ages 1--5 years, the median age at viral suppression was 44 months (IQR: 31--59 months). The median number of analyzed specimens for HIV-1 antibodies per participant over follow--up was nine specimens (IQR: 5--12). Twenty children (33%) had one viral load blip; nine (15%) had two blips; and five (8%) had three to five blips.
HIV-1-specific antibody trajectories following virologic suppression on antiretroviral therapy
Among participants with viral suppression at less than 1 year of age, initial antibody levels were similar to PHEU controls as evidenced by the intersection of the line fits for the two groups (Fig. 1). However, antibodies did not decrease by age at the same rate among participants with viral suppression at less than 1 year of age as the PHEU controls; antibodies to all HIV-1 proteins among PHEU controls decreased to undetectable by 1–2 years of age. Among children with viral suppression at less than 1 year of age, antibodies to gp160 and gp41 remained stable, antibodies to p17, p24, and RT decreased over time; antibodies to p31 were low and remained low. Antibody levels for children with viral suppression between 1 and 5 years of age were higher than those for children with earlier viral suppression and remained stable or increased over time. Antibody trajectories were similar among participants who experienced viral load blips and those who did not (data not shown).
Relationship between quantitative HIV-1-specific antibody levels and HIV-1 DNA levels
HIV-1 antibody levels for each of the proteins tended to be lower when HIV-1 DNA levels were less than 100 copies per million PBMCs, compared with time points where HIV-1 DNA levels were at least 100 copies per million PBMCs in both the full analysis population (eFigure 1, https://links.lww.com/QAD/B694) and the population restricted to no viral load blips (eFigure 2, https://links.lww.com/QAD/B694). Although fewer time points included HIV-1 DNA levels less than 10 copies per million PBMCs, this trend of lower HIV-1 antibody levels with low HIV-1 DNA levels remained in both analytic populations (eFigures 3–4, https://links.lww.com/QAD/B694).
ROC curves for gp41, gp160, p31, and RT showed fair discrimination for HIV-1 DNA levels less than 100 copies per million PBMCs with C-statistics ranging from 0.77 for gp160 and 0.76 for gp41 to 0.71 for p31 (Fig. 2 a). Among the restricted population (n = 27), only gp41, gp160, and p31 had fair discrimination of HIV-1 DNA levels less than 100 copies per million PBMCs with C-statistics of 0.75, 0.79, and 0.77, respectively (not shown). For the HIV-1 DNA threshold of less than 10 copies per million PBMCs, gp41, gp160, and p31 showed fair discrimination in ROC analyses of the full analysis population (Fig. 2 b). Among the restricted population, only gp41 and gp160 had fair discrimination for HIV-1 DNA levels less than 10 copies per million PBMCs with C-statistics of 0.70 and 0.75 respectively (not shown).
The ability of the combination of all six HIV-1 antibody levels, cut off at their respective thresholds that maximized sensitivity and specificity, to discriminate between HIV-1 DNA levels less than 100 copies per million PBMCs from higher levels was similar in the full and restricted analytic populations (C-statistic 0.77) (Table 2). The joint ability of all six HIV-1 antibody levels to discriminate between HIV-1 DNA levels less than 10 copies per million PBMCs from higher levels was similar (0.75 in the full analytic population and 0.77 in the restricted population). The stepwise model-building approach highlighted gp160, gp41, and p24 as the most important predictors of HIV-1 DNA levels less than 10 copies per million PBMCs (C-statistic 0.75) as the addition of the rest of the antibodies to the combined model did not increase the overall C-statistic beyond the model including only these three antibodies.
In the full analysis population, the Random Forest method identified gp41 and gp160 as the variables of most importance for predicting HIV-1 DNA levels less than 100 copies per million PBMCs, followed by p31, p17, and RT (Fig. 3a). In comparison, p31, gp160, and gp41 were identified as variables of most importance for predicting HIV-1 DNA levels less than 100 copies per million PBMCs among the population with no viral blips. Area under the curve (AUC) estimates utilizing all six HIV-1 antibodies for prediction of HIV-1 DNA levels less than 100 copies per million PBMCs with Random Forests were 0.81 in the full analytic population and 0.78 in the population restricted to no viral blips. Results were similar for predicting the lower HIV-1 DNA threshold of less than 10 copies per million PBMCs. gp160 was identified as the variable of most importance in the full and restricted study populations, followed by RT and gp41 (Fig. 3b). AUC estimates utilizing all six HIV-1 antibodies for prediction of HIV-1 DNA levels less than 10 copies per million PBMCs with Random Forests were 0.74 in the full analytic population and 0.70 in the restricted population.
Developing feasible ways to identify children with low proviral reservoirs for participation in clinical trials aimed at ART-free remission and cure is important for broader application of HIV cure trials in low and middle-income countries. HIV-1 antibody tests are relatively inexpensive and do not require sophisticated laboratory setups, making testing relatively easy to implement in limited-resource settings, where most new pediatric HIV-1 infections now occur. We and others have previously shown lower residual blood cell-associated HIV-1 DNA levels [4–7,20,21] and low or absent HIV-1-specific antibody levels [13,22,23] in perinatal infection when viral suppression occurs following ART initiation in the first few months of life. These findings led to the concept that HIV-1-specific antibody levels may be a marker for the extent of exposure to HIV-1 replication which, in turn, has been related to the size of the DNA and latent reservoirs following ART. Our previous study found that gp160 was associated with HIV-1 DNA levels . We sought to examine HIV-1 antibody decay rates after sustained viral suppression on ART among children with PHIV and determine whether the quantitative levels of HIV-1 antibodies to a specific antigen or set of antigens would allow us to identify children who had low HIV-1 DNA levels defined as less than 100 copies per million PBMCs or less than 10 copies per million PBMCs. Recently, Rocca, Palma, and colleagues reported that an aggregate western blot score could estimate the size of the HIV-1 DNA reservoir and the timing of ART initiation in children who have achieved long-term viral suppression .
The HIV-1 quantitative antibody level trajectories for the HIV-infected children compared with the trajectories of the PHEU controls suggest that most children who achieved viral suppression had been exposed to virus long enough to generate their own HIV-1 antibodies. Differences in the levels of individual HIV-1 antibodies between the viral suppression less than 1 year group and viral suppression 1--5 years group likely reflect the timing of ART with respect to the acquisition of infection and the timing of the generation of HIV-1-specific IgG antibodies in primary infection . The average age at initiation of ART and virologic suppression in the viral suppression less than 1-year group were 2 and 8 months, respectively, compared with 21 and 44 months in the viral suppression 1--5 years’ group. IgG antibodies to HIV-1 gp41 and gp120 are detected 13 and 28 days, respectively, following initial detection of HIV-1 RNA in primary adult infection, whereas antibodies to p24 and p17 are detected a median of 18 and 33 days .
HIV-1 antibody levels to p31 were either at the lower dynamic range limit or simply very low for early viral suppression participants throughout the study; by contrast 60% of children with viral suppression between ages 1 and 5 years were consistently p31 antibody positive. Indeed, half of the participants who had suppressed viral load less than 1 year of age were undetectable at the beginning of the study. This was not surprising given that the p31 antibodies are produced around 7 weeks after detectable viremia  and appear to be lost in the absence of viral production .
We used two methods in this exploratory analysis to determine, which HIV-1 antibody levels could be used to predict specimens having HIV-1 DNA levels less than 10 or less than 100 with sufficient accuracy. They included ROC curves with a stepwise model-building approach for evaluating joint prediction, and the random forests ensemble learning method, which is a nonparametric statistical tool that often shows improved ability over stochastic regression modeling in prediction settings. The set of HIV-1 antibody levels identified as the strongest predictors of HIV-1 DNA were fairly consistent across the different methods, with HIV-1 antibody levels to gp160 and gp41 often identified. Identification of gp160 results are consistent with our previous study . None of the methods explored; however, achieved a level of discrimination sufficient to act as surrogate for the gold standard, the droplet digital PCR for PBMC-associated HIV-1 DNA level. Measurement of quantitative antibodies may otherwise be useful in identifying children on ART whose samples will test below the target 100 or 10 HIV-1 DNA copies per million PBMC thresholds.
This was a first-pass, exploratory study to evaluate, which antibodies were the most predictive of low viral reservoir and what the joint predictive capacity was when antibodies were combined. As such, it was limited by the availability of samples, lack of a validation data set, and variable sample collection across both participants and age. We elected to retain data for those participants who experienced isolated viral blips, although viral blips may influence the HIV-1 antibody levels and HIV-1 DNA levels. However, in sensitivity analyses restricting to samples from participants who did not experience viral blips, prediction did not markedly improve across any of the various methods we explored. Due to the lack of a validation data set, there remains a risk of overfitting prediction models. However, the same two antibodies (gp160 and gp41) were almost always identified as the strongest predictors of HIV-1 DNA levels.
Nevertheless, we observed markedly lower HIV-1 antibody levels over time among children with PHIV on ART who achieved viral suppression at less than 1 year of age compared with children with later viral suppression and identified gp41 and gp160 as consistent and important predictors of low HIV-1 DNA levels in children. Future studies using a larger dataset along with a validation dataset would be helpful in crafting an algorithm to identify individuals who would be good candidates for enrollment into HIV-1 cure or remission clinical trials.
We thank the children and families for their participation in PHACS, and the individuals and institutions involved in the conduct of PHACS. The study was supported by the Eunice Kennedy Shriver National Institute of Child Health and Human Development, with co-funding from the National Institute on Drug Abuse, the National Institute of Allergy and Infectious Diseases, the Office of AIDS Research, the National Institute of Mental Health, the National Institute of Neurological Disorders and Stroke, the National Institute on Deafness and Other Communication Disorders, the National Heart Lung and Blood Institute, the National Institute of Dental and Craniofacial Research, and the National Institute on Alcohol Abuse and Alcoholism, through cooperative agreements with the Harvard University T.H. Chan School of Public Health (HD052102, 3 U01 HD052102-05S1, 3 U01 HD052102-06S3) (Principal Investigator: George Seage; Project Director: Julie Alperen) and the Tulane University School of Medicine (HD052104, 3U01 HD052104-06S1) (Principal Investigator: Russell Van Dyke; Co-Principal Investigator: Kenneth Rich; Project Director: Patrick Davis). Data management services were provided by Frontier Science and Technology Research Foundation (PI: Suzanne Siminski), and regulatory services and logistical support were provided by Westat, Inc (PI: Julie Davidson).
The following institutions, clinical site investigators, and staff participated in conducting PHACS AMP in 2012, in alphabetical order: Baylor College of Medicine: William Shearer, Mary Paul, Norma Cooper, Lynette Harris; Bronx Lebanon Hospital Center: Murli Purswani, Mahboobullah Baig, Anna Cintron; Children's Diagnostic & Treatment Center: Ana Puga, Sandra Navarro, Doyle Patton, Deyana Leon; Children's Hospital, Boston: Sandra Burchett, Nancy Karthas, Betsy Kammerer; Ann & Robert H. Lurie Children's Hospital of Chicago: Ram Yogev, Margaret Ann Sanders, Kathleen Malee, Scott Hunter; Jacobi Medical Center: Andrew Wiznia, Marlene Burey, Molly Nozyce; St. Christopher's Hospital for Children: Janet Chen, Latreca Ivey, Maria Garcia Bulkley, Mitzie Grant; St. Jude Children's Research Hospital: Katherine Knapp, Kim Allison, Megan Wilkins; San Juan Hospital/Department of Pediatrics: Midnela Acevedo-Flores, Heida Rios, Vivian Olivera; Tulane University Health Sciences Center: Margarita Silio, Medea Jones, Patricia Sirois; University of California, San Diego: Stephen Spector, Kim Norris, Sharon Nichols; University of Colorado Denver Health Sciences Center: Elizabeth McFarland, Emily Barr, Robin McEvoy; University of Medicine and Dentistry of New Jersey: Arry Dieudonne, Linda Bettica, Susan Adubato; University of Miami: Gwendolyn Scott, Patricia Bryan, Elizabeth Willen.
Funding Support: This work was supported by National Institutes of Health [NIH R01-HD080474 (K.L., D.P.) and UL1-TR001453 (K.L.)] and the Fondazione Penta-Onlus (PENTA Foundation (K.L.)). The Pediatric HIV/AIDS Cohort Study (PHACS) was supported by the Eunice Kennedy Shriver National Institute of Child Health and Human Development with co-funding from the National Institute on Drug Abuse, the National Institute of Allergy and Infectious Diseases, the Office of AIDS Research, the National Institute of Mental Health, the National Institute of Neurological Disorders and Stroke, the National Institute on Deafness and Other Communication Disorders, the National Heart Lung and Blood Institute, the National Institute of Dental and Craniofacial Research, and the National Institute on Alcohol Abuse and Alcoholism, through cooperative agreements with the Harvard University School of Public Health (HD052102, 3 U01 HD052102-05S1, 3 U01 HD052102-06S3) and the Tulane University School of Medicine (HD052104, 3U01HD052104-06S1).
Note: The conclusions and opinions expressed in this article are those of the authors and do not necessarily reflect those of the National Institutes of Health or United States Department of Health and Human Services.
Robin Brody, BA and Anita Gautam, MS determined the quantitative HIV-1 antibody levels.
Author contributions: study design: B.K., K.L., M.M., K.P.; analysis: B.K., K.P.; interpretation of results: B.K., K.L., M.M., K.P.; manuscript draft/edit: B.K., K.L., M.M., D.P., K.P.
Conflicts of interest
There are no conflicts of interest.
1. Luzuriaga K. Early combination antiretroviral therapy limits HIV-1 persistence in children
. Annu Rev Med
2. Persaud D, Palumbo PE, Ziemniak C, Hughes MD, Alvero CG, Luzuriaga K, et al. Dynamics of the resting CD4(+) T-cell latent HIV reservoir in infants initiating HAART less than 6 months of age
3. Hill AL, Rosenbloom DI, Fu F, Nowak MA, Siliciano RF. Predicting the outcomes of treatment to eradicate the latent reservoir for HIV-1
. Proc Natl Acad Sci U S A
4. Ananworanich J, Puthanakit T, Suntarattiwong P, Chokephaibulkit K, Kerr SJ, Fromentin R, et al. Reduced markers of HIV persistence and restricted HIV-specific immune responses after early antiretroviral therapy in children
5. Martinez-Bonet M, Puertas MC, Fortuny C, Ouchi D, Mellado MJ, Rojo P, et al. Establishment and replenishment of the viral reservoir in perinatally HIV-1-infected children initiating very early antiretroviral therapy
. Clin Infect Dis
6. McManus M, Mick E, Hudson R, Mofenson LM, Sullivan JL, Somasundaran M, Luzuriaga K. PACTG 356 Investigators. Early Combination Antiretroviral Therapy Limits Exposure to HIV-1 Replication and Cell-Associated HIV-1 DNA Levels in Infants
. PLoS One
7. Uprety P, Patel K, Karalius B, Ziemniak C, Chen YH, Brummel SS, et al. Human immunodeficiency virus type 1 DNA decay dynamics with early, long-term virologic control of perinatal infection
. Clin Infect Dis
8. van Zyl GU, Bedison MA, van Rensburg AJ, Laughton B, Cotton MF, Mellors JW. Early antiretroviral therapy in South African children reduces HIV-1-infected cells and cell-associated HIV-1 RNA in blood mononuclear cells
. J Infect Dis
9. Luzuriaga K, Tabak B, Garber M, Chen YH, Ziemniak C, McManus MM, et al. HIV type 1 (HIV-1) proviral reservoirs decay continuously under sustained virologic control in HIV-1-infected children who received early treatment
. J Infect Dis
10. Persaud D, Gay H, Ziemniak C, Chen YH, Piatak M Jr, Chun TW, et al. Absence of detectable HIV-1 viremia after treatment cessation in an infant
. N Engl J Med
11. Bitnun A, Samson L, Chun TW, Kakkar F, Brophy J, Murray D, et al. Early initiation of combination antiretroviral therapy in HIV-1-infected newborns can achieve sustained virologic suppression with low frequency of CD4+ T cells carrying HIV in peripheral blood
. Clin Infect Dis
12. Luzuriaga K, Bryson Y, Krogstad P, Robinson J, Stechenberg B, Lamson M, et al. Combination treatment with zidovudine, didanosine, and nevirapine in infants with human immunodeficiency virus type 1 infection
. N Engl J Med
13. Luzuriaga K, McManus M, Catalina M, Mayack S, Sharkey M, Stevenson M, Sullivan JL. Early therapy of vertical human immunodeficiency virus type 1 (HIV-1) infection: control of viral replication and absence of persistent HIV-1-specific immune responses
. J Virol
14. Violari A, Cotton MF, Kuhn L, Schramm DB, Paximadis M, Loubser S, et al. A child with perinatal HIV infection and long-term sustained virological control following antiretroviral treatment cessation
. Nat Commun
15. Persaud D, Patel K, Karalius B, Rainwater-Lovett K, Ziemniak C, Ellis A, et al. Pediatric HIV/AIDS Cohort Study. Influence of age at virologic control on peripheral blood human immunodeficiency virus reservoir size and serostatus in perinatally infected adolescents
. JAMA Pediatr
16. Rocca S, Zangari P, Cotugno N, De Rossi A, Ferns B, Petricone D, et al. EPIICAL Consortium. Human immunodeficiency virus (HIV)-antibody repertoire estimates reservoir size and time of antiretroviral therapy initiation in virally suppressed perinatally HIV-infected children
. J Pediatric Infect Dis Soc
17. McManus M, Henderson J, Gautam A, Brody R, Weiss ER, Persaud D, et al. PACTG 356 Investigators. Quantitative human immunodeficiency virus (HIV)-1 antibodies correlate with plasma HIV-1 RNA and cell-associated DNA levels in children on antiretroviral therapy
. Clin Infect Dis
18. Pediatric HIV/AIDS Cohort Study (PHACS). PHACS active protocols. In: https://phacsstudy.org/About-Us/Active-Protocols
; 2019. Access date: 9/5/2019.
19. Breiman L. Random Forests
. Machine Learning
20. Kuhn L, Paximadis M, Da Costa Dias B, Loubser S, Strehlau R, Patel F, et al. Age at antiretroviral therapy initiation and cell-associated HIV-1 DNA levels in HIV-1-infected children
. PLoS One
21. Tagarro A, Chan M, Zangari P, Ferns B, Foster C, De Rossi A, et al. Early and highly suppressive antiretroviral therapy are main factors associated with low viral reservoir in European perinatally HIV-infected children
. J Acquir Immune Defic Syndr
22. Kuhn L, Schramm DB, Shiau S, Strehlau R, Pinillos F, Technau K, et al. Young age at start of antiretroviral therapy and negative HIV antibody results in HIV-infected children when suppressed
23. Payne H, Mkhize N, Otwombe K, Lewis J, Panchia R, Callard R, et al. Reactivity of routine HIV antibody tests in children who initiated antiretroviral therapy in early infancy as part of the Children with HIV Early Antiretroviral Therapy (CHER) trial: a retrospective analysis
. Lancet Infect Dis
24. Tomaras GD, Haynes BF. HIV-1-specific antibody responses during acute and chronic HIV-1 infection
. Curr Opin HIV AIDS