Child-related factors associated with viral load suppression among HIV-positive children enrolled in the Jua Mtoto Wako Initiative, Turkana County, Kenya: a cross-sectional study

¹Department of Public Health, Amref International University, Nairobi, Kenya

^&Corresponding author
Becky Chepkorir Bureti, Department of Public Health, Amref International University, Nairobi, Kenya

Introduction    

Human immunodeficiency virus (HIV) remains a major public-health challenge despite the expansion of antiretroviral therapy (ART). Children living with HIV require sustained treatment, adherence support, clinical monitoring, and age-appropriate nutritional care to achieve and maintain viral suppression [1-5]. Individual, caregiver, treatment, and health-system factors can affect viral suppression in children, and researchers should interpret results in relation to the population and threshold used in each study. In Kenya, paediatric HIV services operate across settings with different geographic, socioeconomic, and service-delivery challenges. Turkana County has dispersed communities and pastoralist livelihoods, which may complicate clinic attendance, ART continuity, access to food, and follow-up. Children enrolled in community-based programs may receive a combination of facility care and caregiver-facing support, so it is important to understand which clinical characteristics are associated with viral-load outcomes in this setting [3,6-9]. The Jua Mtoto Wako Initiative provides the program context for this analysis. The available manuscript materials describe a community- and family-centered initiative implemented through eight health facilities in Turkana County. Existing studies have examined pediatric viral suppression in Kenya, including national cohort data and work in Turkana and Homa Bay, alongside evidence on family-centered care [3,4,8,10-14]. This analysis describes suppression and nutritional-status patterns among children documented in the initiative's participating facilities. The specific research questions were i) What proportion of included children had a recorded viral load <200 copies/mL? ii) How did suppression prevalence vary by ART adherence, nutritional status, and other child-level clinical characteristics? iii) Was adherence, nutritional status, and selected clinical characteristics associated with suppression after accounting for measured covariates and facility clustering? We hypothesized that poorer adherence, treatment interruption, comorbidity, and nutritional depletion would be associated with lower viral-load suppression. These hypotheses describe expected associations, not causal effects.

Methods     

Study design and setting: this was a cross-sectional analysis of routine paediatric HIV clinical records supplemented by structured caregiver interviews. The study included eight facilities implementing the Jua Mtoto Wako Initiative in Turkana County, Kenya: Lodwar County Referral Hospital, St. Patrick's, Kalokol AIC Health Center, St. Monica, Kerio Health Center, St. Mary's Kalokol, St. Catherine, and Ngitakito Dispensary, from January 2021 to June 2024. The cross-sectional analysis used recorded clinical characteristics and the most recent viral-load result available in the dataset.

Participants and eligibility: the enumerated population comprised 197 children. We included 165 in the analysis and excluded 32 because of incomplete records, missing viral-load information, or loss to follow-up. We analyzed 165/197 (83.8%); this is not described as a response rate. Children were eligible if they were aged 0-14 years, had confirmed HIV infection, were enrolled in the initiative, were receiving ART, and had a documented eligible viral-load result.

Variables and operational definitions: the outcome was viral-load suppression, coded as a recorded result <200 copies/mL versus ≥200 copies/mL, following the categorical definition in the supplied dataset. This threshold differs from the ≤1,000 copies/mL threshold used in some WHO program-monitoring definitions; the approved study protocol and applicable Kenyan guidance support the <200 threshold. Antiretroviral therapy adherence was classified into three categories: good (≥95%), fair (85-94%), and poor (<85%). We based the classification on information documented in clinical records, including pill counts, pharmacy refill records, caregiver recall, and clinician assessment. We defined good adherence as no missed doses, consistent with the adherence assessment documented in the study records. We reviewed the timing of adherence assessment relative to viral-load testing to determine whether adherence information reflected the period before viral-load measurement or was documented after an elevated viral-load result, potentially during enhanced adherence counseling. Where the available records did not allow us to establish the timing, we noted this as a limitation because adherence documentation following an elevated viral load may not accurately reflect adherence before testing. Nutritional status was assessed using mid-upper arm circumference (MUAC) only. Children were categorized as having normal nutritional status (MUAC ≥12.5 cm), moderate malnutrition (MUAC 11.5-12.4 cm), or severe malnutrition (MUAC < 11.5 cm).Treatment interruption was recorded as yes/no; the duration and reason fields were free text and were not consistently structured. Comorbidity was recorded as yes/no, with TB and pneumonia listed as examples in the data form. Antiretroviral therapy regimen was harmonized for this analysis into DTG-based versus non-DTG/other regimens. Duration on ART was calculated in months between the recorded ART start date and the most recent viral-load test date, because many viral-load dates fall after June 2024.

Data sources and bias: we abstracted demographic and clinical information from routine records. We collected caregiver-related information using a structured, interviewer-administered questionnaire in KoboCollect. Potential biases include selection bias from excluding 32 of 197 enumerated children, information bias from routine documentation of adherence and nutritional status, possible differential documentation of adherence after elevated viral load, and reverse causation because uncontrolled HIV may contribute to wasting. We addressed facility-level differences in care practices and drug supply using facility-clustered standard errors, although we had only eight clusters. Residual confounding may remain, including confounding by regimen, duration on ART, caregiver factors and drug resistance.

Study size: we planned a census approach for the eligible population of 197 children; 165 records met the available completeness criteria, and the excluded 32 were due to missing clinical records (13), missing viral load records (10), and loss to follow-up (9) (Figure 1). This was not a probability sample.

Statistical analysis: frequencies and percentages were calculated for categorical variables. Suppression proportions were calculated as the number suppressed divided by the number with available data in each category. We calculated crude prevalence ratios (cPRs) and 95% confidence intervals for suppression, using normal nutritional status and good adherence as the reference categories. Pearson chi-square tests were used for descriptive comparisons when cell sizes were adequate; fisher's exact test was used for sparse cells. Modified poisson regression with a log link was fitted to estimate adjusted prevalence ratios. The model included age group, sex, orphan status, ART adherence, nutritional status, comorbidity, treatment interruption, regimen group, and duration on ART. Age and sex were included a priori. Complete-case analysis resulted in 165 observations. Facility was used as the clustering variable for robust standard errors; the displayed confidence intervals use a to reference with seven degrees of freedom as a conservative approximation and require independent statistical review, preferably with an agreed small-sample correction or a justified alternative. The software used for the calculations in this study was Python with pandas and statsmodels.

Ethical considerations: the Amref International University Ethics and Scientific Review Committee approved the study. The study was conducted under the National Commission for Science, Technology and Innovation (NACOSTI) research permit No. 272133, with additional authorization from Turkana County. We obtained written informed consent from caregivers who participated in the study. Children who were comfortable and older than ten years provided assent to participate, in accordance with the approved study procedures. The team-maintained confidentiality of participants' information throughout data collection, analysis, and reporting.

Results     

Participant inclusion and overall viral-load suppression: of 197 children enumerated, we included 165 (83.8%) in the analysis and excluded 32 (16.2%) because of missing clinical records (13), missing viral load records (10), and loss to follow-up (9). Among the 165 included children, 122 (73.9%) had a recorded viral load <200 copies/mL and 43 (26.1%) had a result ≥200 copies/mL.

Demographic and clinical characteristics: children were aged 0-14 years; 10 (6.1%) were 0-4 years, 62 (37.6%) were 5-9 years, and 93 (56.4%) were 10-14 years. There were 87 boys and 78 girls. Eighty-four children (50.9%) were not orphaned, 58 (35.2%) were single orphans, and 23 (13.9%) were double orphans. We recorded good adherence in 107 (64.8%), fair adherence in 36 (21.8%), and poor adherence in 22 (13.3%). Nutritional status was recorded as normal for 94 (57.0%), moderate malnutrition for 55 (33.3%), and severe malnutrition for 16 (9.7%). The dataset recorded treatment interruption for 40 children (24.2%) and no interruption for 125 (75.8%). Comorbidity was recorded for 11 children (6.7%), absent for 153 (92.7%), and missing for one. We classified regimen strings as DTG-based for 143 children (86.7%) and non-DTG/other for 22 (13.3%) Table 1.

Suppression by clinical characteristics: suppression varied across adherence and nutritional-status categories. We recorded suppression for 99/107 (92.5%) children with good adherence, 18/36 (50.0%) with fair adherence, and 5/22 (22.7%) with poor adherence. Suppression was recorded for 84/94 (89.4%) children with normal nutritional status, 32/55 (58.2%) with moderate malnutrition, and 6/16 (37.5%) with severe malnutrition. Figure 2 presents suppression proportions with Wilson 95% confidence intervals. These are descriptive associations and do not establish direction of causality. The category-specific suppression counts and crude prevalence ratios are presented in Table 2.

Adjusted prevalence ratios: the modified poisson model included 165 complete cases. With good adherence as the reference, fair adherence was associated with a lower prevalence of suppression (aPR 0.61, 95% CI: 0.41-0.90), and poor adherence with a lower prevalence (aPR 0.32, 95% CI: 0.17-0.60). With normal nutritional status as the reference, moderate malnutrition had an aPR of 0.81 (95% CI: 0.60-1.11) and severe malnutrition had an aPR of 0.65 (95% CI 0.46-0.92). These confidence intervals use a t reference with seven degrees of freedom to reflect the small number of facility clusters. In this model, adjusted associations for age, sex, orphan status, comorbidity, treatment interruption, regimen group, and duration on ART did not meet the 0.05 significance threshold (Table 3). Facility-specific suppression ranged from 66.7% at St Catherine (3/5) to 84.1% at St Patrick's (37/44); estimates for facilities with small denominators are imprecise. Facility-specific suppression counts and percentages are presented in Table 4.

Sensitivity analysis using <1,000 copies/mL: among the 165 children with a numeric viral-load result, 138 (84.7%) had a result <1,000 copies/mL and 25 (15.3%) had a result ≥1,000 copies/mL.

Discussion     

In this study of 165 children receiving HIV care at eight facilities implementing the Jua Mtoto Wako Initiative in Turkana County, 73.9% achieved viral-load suppression, defined as a viral load of less than 200 copies/mL, consistent with the Kenyan threshold specified for this study. Suppression was more common among children with documented good ART adherence and normal nutritional status. The proportion suppressed declined from 89.4% among children with normal nutritional status to 58.2% among those with moderate malnutrition and 37.5% among those with severe malnutrition. These findings suggest that adherence and nutritional status are important factors to consider when evaluating paediatric HIV treatment outcomes. However, the cross-sectional design limits conclusions about causality and the temporal relationship between these factors and viral suppression. Antiretroviral therapy adherence showed a graded relationship with viral suppression, with children classified as having fair or poor adherence achieving lower suppression than those with good adherence. This finding is consistent with a systematic review of children living with HIV in sub-Saharan Africa, which identified suboptimal ART adherence as one of the most consistently reported factors associated with viral non-suppression [15]. Similarly, Masaba et al. reported improved viral suppression after implementing standardized, enhanced adherence counseling among children and adolescents with suspected treatment failure in Turkana and Homa Bay counties [4]. Their findings highlight the potential value of structured adherence support in paediatric HIV care. Nevertheless, we assessed adherence in the present study using clinical records, including pill counts, pharmacy refill information, caregiver recall, and clinician assessment.

Nutritional status was also strongly associated with viral suppression in the descriptive findings. Suppression decreased progressively with increasing severity of recorded malnutrition, consistent with evidence from a systematic review identifying malnutrition among factors associated with unsuppressed viral load in children receiving ART in sub-Saharan Africa, alongside observational evidence linking malnutrition with viral suppression outcomes [16,11]. However, this study assessed nutritional status using MUAC only; it did not use weight-for-age, height-for-age, or BMI-for-age. Furthermore, the authors could not independently verify the age-specific MUAC cut-offs and classification algorithm. Consequently, nutritional-status misclassification is possible, particularly among older children and adolescents. The relationship may also be bidirectional, as nutritional depletion may coexist with poor treatment outcomes, while uncontrolled HIV may contribute to wasting. These findings should therefore be interpreted as associations rather than evidence that malnutrition directly caused viral non-suppression. The adjusted estimates for treatment interruption, comorbidities, ART regimen, and duration on ART were imprecise or did not clearly distinguish an association from the null in the working model. These results should not be interpreted as evidence that these factors are unimportant. The relatively small sample, limited numbers in some categories, potential measurement error, and residual confounding may have reduced the precision of the estimates. In addition, differences across the eight participating facilities may reflect variations in patient characteristics, documentation practices and service delivery. The findings have practical implications for the Jua Mtoto Wako Initiative and similar pediatric HIV programs. Routine adherence assessment, nutritional screening, and timely follow-up of children with elevated viral loads may help identify children who need additional clinical or psychosocial support. Improving documentation of adherence assessment dates, nutritional measurements, and treatment history would also strengthen clinical monitoring and future program evaluations. These recommendations are consistent with the observed associations and published evidence.

Study limitations: we considered several limitations when interpreting the findings. First, the cross-sectional design limits the assessment of temporality and does not establish causal relationships. Second, we derived adherence and nutritional status from clinical records, which may be subject to measurement error or misclassification. Third, using MUAC alone, without other anthropometric indicators, limits assessment of nutritional status across the full age range. Fourth, excluding records with incomplete information may have introduced selection bias if excluded children differed systematically from those included.

Conclusion     

Among the 165 children included in the analysis, 73.9% had a recorded viral load below 200 copies/mL. In the adjusted analysis, adherence and nutritional status were associated with viral load suppression, with good adherence and normal nutritional status showing more favorable outcomes than fair or poor adherence and severe malnutrition, respectively. These findings suggest that adherence and nutritional status are important factors to consider in paediatric HIV care. Integrating adherence support and nutritional assessment into routine clinical management may help address potentially modifiable factors associated with viral load suppression. Further longitudinal studies are recommended to establish the temporal relationships between these factors and viral load suppression and to investigate other determinants of treatment outcomes among children living with HIV.

Competing interests     

The authors declare no competing interests.

Authors' contributions     

Becky Chepkorir Bureti conceived and designed the study, coordinated data collection, analyzed and interpreted the data, and drafted the manuscript. Tom Marwa and Patrick Angala contributed to study design and supervision and critically reviewed the manuscript for important intellectual content. All authors should review and approve this wording and the final manuscript before submission.

Acknowledgments     

The authors acknowledge the participating health facilities and the clinical teams involved in paediatric HIV care.

Tables and figures     

Table 1: demographic and clinical characteristics of HIV-positive children aged 0-14 years enrolled in the Jua Mtoto Wako Initiative across eight health facilities in Turkana County, Kenya (N=165)

Table 2: viral-load suppression by demographic and clinical characteristics among children enrolled in the Jua Mtoto Wako Initiative in Turkana County, Kenya (N=165)

Table 3: modified poisson regression of factors associated with viral-load suppression among HIV-positive children enrolled in the Jua Mtoto Wako Initiative in Turkana County, Kenya (complete-case N=165)

Table 4: viral-load suppression by participating health facility among HIV-positive children enrolled in the Jua Mtoto Wako Initiative in Turkana County, Kenya (N=165)

Figure 1: proportion of children with viral-load suppression by recorded nutritional-status category, with Wilson 95% confidence intervals (N=165)

Figure 2: participant flow from enumeration to analysis

References