Prevalence and Correlates of Viral Load Suppression and Human Immunodeficiency Virus (HIV) Drug Resistance Among Children and Adolescents in South Rift Valley and Kisumu, Kenya

BACKGROUND: Children and adolescents living with HIV (CALHIV) face unique challenges, including poorer treatment outcomes, risk for drug-resistance mutations (HIVDRMs), and limited drug formulations. We estimated viral suppression (VS) prevalence and evaluated predictors of VS and HIVDRMs in Kenya. METHODS: From 2018-2020, CALHIV 1-19 years on antiretroviral therapy (ART) >6 months were enrolled in this cross-sectional study. Participants underwent viral load (VL) testing; those with VL ≥1000 copies/mL had HIVDRM testing. Sociodemographic questionnaires and medical record abstraction were completed. VS prevalence (VL <1000 copies/mL) was estimated; robust Poisson regression models were used to estimate prevalence ratios (PRs) and 95% CIs for associations between potential predictors of VS. RESULTS: Nine hundred and sixty-nine participants were enrolled. VS prevalence was .80 (95% CI: .78-.83). Being on ART >24 months (adjusted PR [aPR]: 1.22; 95% CI: 1.06-1.41), an integrase strand transfer inhibitor-containing regimen (1.13; 1.02-1.26), and attending a level 3 health facility (1.23; 1.11-1.36) were associated with VS. Missing ≥3 doses of ART in the past month (aPR: .73; 95% CI: .58-.92), having a viremic mother with HIV (.72; .53-.98), and having 3-7 (.90; .83-.97), 8-13 (.89; .82-.97), or ≥14 (.84; .77-.92) compared with <2 adherence counseling referrals were inversely associated with VS. A high proportion (n = 119, 81.5%) of unsuppressed participants had evidence of any major HIVDRM. CONCLUSIONS: HIV treatment programs should target interventions for pediatric patients at risk for treatment failure-namely, those with a caregiver with failed VS and those struggling with adherence. Between December 2018 and March 2020, participants were invited to enroll in this cross-sectional study based on random selection from 13 treatment clinics supported by the Military HIV Research Program (MHRP) and the US President’s Emergency Plan for AIDS Relief (PEPFAR) in Western Kenya. The random sample was approximately weighted by clinic and stratified by time on ART and age. Randomization was conducted in Stata (StataCorp, College Station, TX) using listings of current clients from each clinic and is representative of the general population of children and adolescents on ART. Individuals were eligible for enrollment if they were at least 1 year to 19 years of age, had been on first- or second-line ART for at least 6 months, and had attended at least 1 follow-up ART clinic visit in the last 6 months. Participants aged 13–17 years were required to have been informed of their HIV status. Study staff administered sociodemographic questionnaires to participants aged 18 years and older, caregivers if the participant was younger than 13 years, and both caregivers and participants if the participant was between 13 and 17 years. Responses captured demographics, self- or parent-reported ART adherence, side effects, and support group participation. Participants underwent a blood draw for VL unless results of a VL drawn for routine clinical care within 1 month of study enrollment were available. Medical and pharmacy record abstraction was completed within 3 weeks of the participant’s visit, including duration on ART, ART regimen, referral history, nutritional status assessments, World Health Organization (WHO) staging, maternal VL where available, and tuberculosis (TB) treatment history. Data from medical record abstraction captured information and events prior to enrollment. Referral history captured documented referrals for adherence counselling and other services; however, evidence of completion of the referred services was unavailable. Data were transcribed onto case report forms and entered in the Clinplus platform (Anju Software, Tempe, AZ). HIV VL was measured via nucleic acid amplification methods on the Abbott m2000sp/rt RealTime System testing platform with a 1-mL plasma sample volume and a lower limit of detection of 40 copies/mL. All testing was performed according to the manufacturer’s instructions. Plasma samples from participants with a VL greater than 1000 copies/mL underwent sequencing of the Pol region using a laboratory-validated modification to the ViroSeq HIV-1 Genotyping System v2.0 (Abbott Molecular, Chicago, IL). Sequences were evaluated for major mutations conferring resistance to nucleoside reverse transcriptase inhibitors (NRTIs), non-nucleoside reverse transcriptase inhibitors (NNRTIs), and protease inhibitors (PIs) using the SmartGene Integrated Database Network System (SmartGene, Zug, Switzerland) to access mutation lists from the Stanford HIV Drug Resistance Database, version 8.8.0 (Stanford University, Stanford, CA) [16]. The Kericho laboratory is accredited by the College of American Pathologists (CAP) and runs CAP EQA for HIVDRM testing. HIV-1 subtype was inferred from the consensus evolutionary tree from SmartGene Integrated Database Network System, which utilizes the neighbor-joining method in MEGA4 software version 4 (Tamura, Dudley, Nei, and Kumar 2007). Evolutionary distances were computed using the maximum composite likelihood method in units of the number of base substitutions per site. The tree was then generated by the neighbor-joining method from a nucleotide alignment. Viral suppression was defined using WHO criteria as a VL of less than 1000 copies/mL. The prevalence of VS was estimated using the Wilson score method and reported with 95% confidence intervals (CIs) in the overall sample, and by age group, duration on ART, and first- versus second-line ART. Bivariate analyses were conducted using Pearson’s chi-square and Wilcoxon rank-sum tests. Generalized linear models with a Poisson distribution and robust standard errors were used to estimate unadjusted and adjusted prevalence ratios (aPRs) and 95% CIs for associations between sociodemographic and clinical factors and VS. Factors significant (α = .05) in the unadjusted models and those identified based on a priori and clinical knowledge of the study setting were included in the adjusted model. ART regimen was dropped from the adjusted model due to redundancy with ART class. We tested for multicollinearity using the variable inflation factor. Analyses were restricted to complete cases after creating separate categories for unknown data from participant medical records. For participants with a VL of 1000 copies/mL or higher, the prevalence of specific HIVDRMs and categories of HIVDRMs were calculated by dividing the number of participants with 1 or more mutations by the total number of participants genotyped. For the drug-resistance analyses, we included all participants with available HIVDRM data and did not restrict to complete cases. Analyses were performed in SAS version 9.4 (SAS Institute, Cary, NC) and Stata version 16.1 (StataCorp) software. The study was approved by institutional review boards of the Walter Reed Army Institute of Research and the Kenya Medical Research Institute Scientific and Ethics Review Unit (KEMRI SERU). All participants provided informed consent and assent, as applicable.