Strategic Antiretroviral Therapy and HIV Testing for Youth in Rural Africa: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Strategic Antiretroviral Therapy and HIV Testing for Youth in Rural Africa (NCT03848728), a randomized na study. The primary outcome is Effectiveness of SEARCH Youth Intervention.

2. Clinical Question

The study addressed whether a combined approach to care could change long-term viral-load suppression among adolescents and young adults living with HIV. It was not a comparison of one isolated counseling or clinic-access component against standard care: the intervention bundled several components. Consequently, the contrast concerns the combination as implemented at intervention clinics, not the separate contribution of each component.

The outcome was framed around viral suppression after two years of follow-up. Its definition and analysis population matter when interpreting the result: the reported endpoint concerns participants remaining eligible under the registry's stated exclusions, and the mean was calculated by averaging percentages across clinics. The reported analysis therefore addresses a clinic-level average outcome, rather than simply describing the proportion of all enrolled individuals who were suppressed.

3. Trial Design

The registry describes a randomized, parallel-group study with no masking. Random assignment took place at the clinic level: clinics were assigned either to continue optimized country standard of care or to provide the SEARCH Youth combination intervention. The registry reports 28 rural HIV clinics in western Kenya and southwest Uganda. Because groups were assigned by clinic, participants within the same clinic share the assigned study condition.

The intervention combined life-stage assessment and counseling, rapid viral-load feedback, structured choice in clinic access, and e-collaboratives chat-based discussions among providers. The comparator was optimized country standard of care. The record does not describe the specific procedures used to generate the clinic assignments, conceal assignment before allocation, or implement the intervention at each site; those details are not reported in the registry record.

Enrollment was 1988. The study is listed as completed, with a start date of 2019-03-14 and primary completion date of 2023-06-23. Participants were to take part for at least 2 years. The primary outcome timeframe was 2 years, and its stated analysis population excluded participants who enrolled on or after Dec 1, 2019, transferred care, outmigrated, or withdrew.

The record identifies the intervention as an “OTHER” intervention and lists phase as NA. It does not report a drug-versus-drug comparison. The study question concerned the effect of a clinic-based combination intervention on suppression, with the sponsor's analysis comparing the intervention and control groups. How many clinics were assigned to each group, whether assignment was stratified, and the detailed operational definition of optimized country standard of care are not reported in the registry record.

Randomization supports a comparison of assigned clinic strategies, but clinic assignment also shapes the statistical analysis: individual participants are not independent assignment units. The reported analysis is therefore more relevant to this design than an unadjusted comparison that treats every participant as independently randomized.

4. Participant Flow and Accounting

The registry reports how participants moved through each stage of the study. All counts below are exactly as posted; missing counts are not inferred from other stages.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: Study participants at year 2, excluding those who enrolled on/after Dec 1, 2019, transferred care, outmigrated, or withdrew.

5. Baseline Characteristics

The table reproduces the baseline characteristics posted in the registry, which describe the participants analysed at baseline and help show whether the randomized groups were comparable.

Population: Population includes participants who were included in the primary analysis, enrolled before December 1, 2019.

6. Primary Endpoint

The registry lists the primary outcome as Effectiveness of SEARCH Youth Intervention, measured over 2 years.

The primary endpoint was the percentage of study participants with virologic suppression, defined as HIV RNA <400 c/mL, at 2 years of follow-up. The registry describes the reported mean as the average percentage across the clusters, meaning clinics. The endpoint therefore reflects clinic-level percentages averaged across clinics, rather than a simple pooled percentage across all participants.

Viral load is used in this study as an outcome that indicates whether HIV RNA is below the stated threshold at the specified follow-up time. A fixed threshold makes the outcome interpretable as a binary status for each assessed participant, while the clinic-level percentage expresses how frequently that status was observed within each clinic. The contrast between arms then addresses differences in these clinic-level averages.

The endpoint is relevant to the stated study aim of evaluating long-term viral-load suppression. Its timepoint is important: it describes status at the two-year assessment, not the full pattern of viral load over the entire follow-up period. The registry's analysis population also limits the scope of the endpoint to year-2 participants after the stated exclusions. The record does not report how participants with unavailable viral-load measurements were handled, or whether another endpoint definition was used in sensitivity analyses.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The registry reports one primary outcome analysis. The outcome is a percentage with virologic suppression, and the registry describes the mean as the average percentage across clinics. Although the individual outcome is a suppressed/not-suppressed status, the reported endpoint is summarized at the clinic level. This distinction matters because the study assigned clinics, not individual participants, to study groups.

Why clinic assignment matters

In a cluster-randomized study, participants attending the same clinic share an assigned intervention. Their outcomes may be more alike than outcomes from participants at different clinics because they share a care setting and clinic-level practices. Treating each participant as an independent randomized unit can therefore misrepresent the amount of information in the data and the uncertainty around a comparison. The registry specifically warns that a simple individual-level re-analysis would be misleading. The clinic-level design calls for an analysis that respects the way assignment occurred.

The endpoint's clinic averaging also affects interpretation. A mean of clinic percentages gives each clinic a role in the summary according to the stated averaging approach; it is not necessarily identical to pooling all participant outcomes into one overall percentage. The registry does not give the clinic-specific percentages or further detail about weighting, so the precise contribution of each clinic to the reported estimate is not reported in the registry record.

What the risk ratio describes

The sponsor reported the treatment contrast as a risk ratio (RR), estimated with TMLE. A risk ratio compares the outcome level in one group with that in the other on a relative scale. It is not an absolute difference in percentage points. Interpreting an RR fully would generally benefit from knowing the outcome levels in both groups, but the arm-specific suppression percentages are not reported in the registry record. The registry gives the RR, its confidence interval, and the p-value, but not enough information to derive the absolute contrast without making an unreported calculation.

TMLE, or targeted maximum likelihood estimation, is a statistical approach used to estimate effects while combining an outcome model with a model for the process that determines exposure or treatment assignment. That is a general description of the method, not a claim about the exact variables, models, or implementation used in this trial. The detailed TMLE specification, covariates, link functions, and handling of clinic structure are not reported in the registry record. The method name alone does not reveal those choices.

Uncertainty and statistical evidence

The reported 95% confidence interval runs from 1.03 to 1.16 around the RR of 1.10. In general, a confidence interval expresses the range of estimates reasonably compatible with the observed data and the method's assumptions at the stated confidence level. A narrower interval indicates greater precision than a wider interval on the same scale, but precision is not the same as absence of bias. An interval can reflect sampling uncertainty without covering uncertainties arising from incomplete reporting, measurement, or model misspecification.

The reported p-value is 0.002. In general, a p-value evaluates how unusual results at least as incompatible with a null model would be if that model and the analysis assumptions held. It is not the chance that the intervention has no effect, nor does it measure the practical importance of the observed contrast. The registry labels the analysis superiority, but the detailed hypothesis wording and decision rule are not reported in the registry record.

Planning illustrations are not trial results

Any sample-size figures associated with planning illustrations should be distinguished from observed outcomes and from the trial's enrollment. The stated planning approach uses the observed effect as a basis for illustration, a two-sided significance level of 0.05, equal allocation between arms, and a normal approximation; the effect to detect is specified as a fraction of the observed effect. For time-to-event illustrations, the stated quantity is total events; other illustrations use participants per arm. These are planning assumptions, not evidence that the study was originally designed with those assumptions, and they are not observed power. The trial's own sample-size justification and design calculations are not reported in the registry record.

Because the record does not give numerical planning results in the information described for this analysis, those figures cannot be interpreted or compared with enrollment. More generally, planning based on an observed estimate can be sensitive to uncertainty in that estimate. It should not be confused with a prospective design rationale or with a post-study measure of whether the trial was adequately powered.

What remains unspecified

The record does not describe missing-outcome handling, the full model specification, covariate adjustment, clinic-level variance assumptions, or sensitivity analyses. It also does not report whether confidence intervals and p-values account for all features of the cluster-randomized design, beyond identifying the sponsor's method and warning against a simple individual-level re-analysis. These omissions do not negate the reported estimate, but they limit independent reconstruction and detailed evaluation of its assumptions.

10. Analyses Reported by the Sponsor

The registry reports the sponsor's own analysis: a TMLE-estimated risk ratio of 1.10, with a 95% confidence interval from 1.03 to 1.16 and p = 0.002. It does not report a separate recalculated estimate for this comparison. The registry explicitly notes that the study was cluster-randomized and that a simple individual-level re-analysis would be misleading, so only the sponsor's own analysis is shown.

This matters because the clinics, rather than individual participants, were randomized. An independent calculation that ignored that assignment structure could give an inappropriate estimate of uncertainty. The sponsor's analysis is therefore the numerical result to use when describing the posted comparison; there is no competing recalculated result in the registry record to assess for agreement or discrepancy.

The registry does not provide enough model detail to reproduce the TMLE analysis or independently verify each analytic choice. Nor does it report the arm-specific suppression percentages needed to express the reported relative estimate as an absolute percentage-point contrast. The result should be described as the sponsor's reported relative comparison, with the stated interval and p-value, rather than as a newly calculated effect.

11. Multiplicity, Missing Data and Analysis Populations

The registry lists one primary outcome measure and one reported analysis. That alone does not establish that no other outcomes, subgroup analyses, or inferential comparisons were conducted. A multiplicity plan describing how the study handled the possibility of multiple statistical tests is not reported in the registry record. The record also does not state whether the reported p-value was adjusted for multiple testing. The p-value should therefore be read as reported, without assuming a particular multiplicity procedure.

The primary analysis population is described as participants at year 2, excluding those who enrolled on or after Dec 1, 2019, transferred care, outmigrated, or withdrew. This definition makes clear that the reported result is not an unrestricted analysis of every enrolled participant. The reasons for the exclusions are stated, but the number of participants excluded for each reason, the number with unavailable viral-load results, and the arm-specific composition of the analyzed population are not reported in the registry record.

How missing outcome measurements were handled is also not reported in the registry record. The record does not say whether missing viral-load values were imputed, treated as unsuppressed, omitted from calculations, or addressed through sensitivity analyses. These choices can affect a percentage endpoint, so the reported estimate cannot be fully assessed for sensitivity to missingness from the available description.

The registry does not report an intention-to-treat analysis definition, a per-protocol population, or other analysis populations beyond the endpoint's stated exclusions. It also does not report a prespecified statistical analysis plan, interim analyses, or stopping rules. Their absence from the registry record should not be interpreted as proof that no such procedures existed; their details are simply not reported there. Accordingly, the primary result can be described according to its stated population and method, while the handling of multiplicity, missing outcomes, and alternative populations remains uncertain.

12. Safety Results

The registry reports serious adverse events as affected participants divided by participants at risk.

Deaths

Other (non-serious) adverse events

Participants with at least one non-serious event at or above the registry's 0% reporting threshold.

Most frequent serious adverse event terms

Serious events with the highest arm-level rate: spontaneous abortion (0.8%), anemia (0.5%), malaria (0.4%), death, cause unknown (0.3%), aids resulting in death (0.2%).

The safety record lists 1012 participants at risk in intervention clinics, with 48 serious adverse events and 16 deaths. For control clinics, it lists 976 participants at risk, with 13 serious adverse events and 13 deaths. These are reported counts; they should not be treated as a comparison of event rates without an analysis that accounts for the study design and the relevant follow-up and event definitions.

The listed serious-event terms and percentages are spontaneous abortion, 0.8%; anemia, 0.5%; malaria, 0.4%; death, cause unknown, 0.3%; AIDS resulting in death, 0.2%; and pulmonary tuberculosis resulting in death, 0.2%. The record does not assign those listed term percentages to a specific arm in the information. It also does not explain whether individuals could contribute more than one event or whether the event counts and death counts overlap.

The death counts do not, by themselves, establish that deaths were caused by the intervention or by the control condition. The record does not provide a causal attribution analysis or enough detail about exposure time, event ascertainment, or statistical comparison to interpret the counts as evidence of a treatment-related safety difference. They should be read as descriptive safety figures recorded for the two groups, not as proof of causation.

13. What the Result Does, and Does Not, Mean

The reported result describes a two-year comparison of clinic-average suppression percentages in the study's defined analysis population. The sponsor's RR estimate was 1.10, with a 95% confidence interval from 1.03 to 1.16 and p = 0.002. These statistics indicate the direction and relative scale of the reported contrast and convey uncertainty under the sponsor's analysis. They do not provide the underlying suppression percentages for either group, so the absolute difference cannot be read directly from the RR.

The result is specific to a bundled intervention delivered through clinics and compared with optimized country standard of care. It does not isolate the effects of life-stage counseling, rapid viral-load feedback, flexible or structured clinic access, or provider chat discussions. It also does not establish what would happen under a different intervention, at different sites, or in a population outside the stated ages and locations.

The analysis population excludes people who enrolled on or after Dec 1, 2019, transferred care, outmigrated, or withdrew. The registry does not explain how unavailable outcome measurements were handled. Those details constrain interpretation of who is represented by the endpoint and how missing results might influence it. The reported outcome is suppression at the two-year assessment, not a description of viral-load status at every point during follow-up.

A statistically informative result is not automatically a statement about practical importance. The confidence interval and p-value address statistical uncertainty under an analysis; they do not state whether the size of the difference is important to participants or clinics. The registry does not supply arm-specific percentages or a clinical threshold for judging the magnitude. Nor does the result establish causation for the recorded safety events.

In short, the posted analysis supports describing a relative difference in the endpoint as estimated by the sponsor's cluster-trial analysis. It should not be extended into an absolute effect, a component-specific effect, or a universal claim about all young people with HIV. The available record leaves important details about missing data and model specification unresolved.

14. Why This Trial Matters Statistically

This trial illustrates why the unit of randomization must guide statistical interpretation. Participants were treated in clinics that were assigned to an intervention or control strategy. The comparison is consequently about clinic-level assignment, and participants within a clinic cannot automatically be regarded as independent randomized units. The registry's warning against a simple individual-level re-analysis highlights a common source of misleading precision in cluster-randomized studies.

The endpoint also makes the scale of analysis important. It was defined as a percentage of participants with suppression, with the mean calculated across clinics. That is distinct from simply combining every participant into a single overall percentage. Understanding which unit is averaged helps readers make sense of the risk ratio and recognize why arm-specific outcome levels would be useful for judging absolute magnitude.

The study combines multiple service-delivery components. Statistically, the randomized contrast can address the intervention package as a whole, but it cannot identify the contribution of each component without additional comparisons or analyses. Clear interpretation therefore depends on keeping the intervention's bundled nature in view.

Finally, the reported result shows the difference between statistical evidence and a complete account of a trial. The RR, confidence interval, and p-value offer a concise description of the sponsor's estimated contrast, while the registry does not report key details such as missing-outcome handling, detailed model specification, or multiplicity procedures. Readers can describe the result accurately while also recognizing what those statistics do not resolve.

15. Limitations

• The posted summary statistics are the only data these results rest on; patient-level data, full protocols and statistical analysis plans are not available.
• Recalculated results are simple, unadjusted calculations from posted summaries. They do not replace the sponsor's prespecified analysis.

Several features specific to this record limit independent appraisal. The trial randomized clinics, and the registry cautions that a simple participant-level re-analysis would be misleading. However, details of the TMLE implementation and how clinic-level dependence was handled are not reported in the registry record. Without those specifications, the analysis cannot be reconstructed from the described result alone.

The primary endpoint is reported as an average of clinic percentages, but clinic-level results, group-specific suppression percentages, and the weighting approach are not reported. That makes it difficult to understand how much individual clinics varied or to translate the relative estimate into an absolute contrast. The result is also confined to the stated two-year assessment and the analysis population after exclusions.

The record does not report counts for each exclusion, the number with missing viral-load outcomes, or the method used to handle missingness. It does not describe sensitivity analyses for those exclusions or for alternate outcome assumptions. The stated exclusions may affect which participants are represented in the result, but the information needed to assess the scale or direction of that effect is not reported.

The intervention is a combination of several service components. The reported comparison cannot distinguish their individual effects. Masking was listed as none, and the registry does not provide further details about outcome ascertainment or how knowledge of clinic assignment may have related to measurement and care practices. The record also does not describe the allocation procedure, clinic-level group counts, or the detailed content of optimized country standard of care.

Finally, the registry does not report a detailed statistical analysis plan, multiplicity handling, interim analyses, or the trial's sample-size justification. These omissions limit assessment of prespecification and analytic robustness. They do not change the numerical result reported by the sponsor, but they narrow the conclusions that can be drawn from the registry description.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Strategic Antiretroviral Therapy and HIV Testing for Youth in Rural Africa — NCT03848728.
• Registry record last updated 2024-10-23; status last verified 2024-09.
• Results first posted 2024-10-23.
• Data retrieved from ClinicalTrials.gov on 2026-11-09.
• Group sizes, means, event counts and sponsor-reported p-values are reproduced as posted. Effect estimates, confidence intervals and p-values not attributed to the sponsor are independently calculated from the posted summary statistics.

19. Record Summary

NCT03848728 is a completed, randomized, parallel-group, unmasked study sponsored by the University of California, San Francisco. It enrolled 1988 participants and assigned rural HIV clinics in western Kenya and southwest Uganda to the SEARCH Youth combination intervention or optimized country standard of care. The intervention included life-stage assessment and counseling, rapid viral-load feedback, structured choice clinic access, and provider e-collaboratives.

The primary endpoint was the percentage with HIV RNA <400 c/mL at 2 years, averaged across clinics, in a population with stated exclusions. The sponsor reported a TMLE risk ratio of 1.10, a 95% confidence interval from 1.03 to 1.16, and p = 0.002. The record cautions that a simple individual-level re-analysis would be misleading. It does not report arm-specific suppression percentages, detailed missing-data handling, or a full statistical model specification.