This page separates reported trial information from statistical interpretation. The ClinicalTrials.gov record contains posted outcome measures and serious-adverse-event counts, but no formal statistical-analysis records with treatment-effect estimates, confidence intervals, or p-values.
1. Trial at a Glance
ACTG A5202 was a randomized, parallel-group, phase 3 treatment trial in HIV-infected treatment-naive adults. The study evaluated two antiretroviral backbone strategies—emtricitabine/tenofovir disoproxil fumarate and abacavir/lamivudine—in combination with either efavirenz or ritonavir-boosted atazanavir.
| Feature | ACTG A5202 |
|---|---|
| Trial name | ACTG A5202 |
| NCT identifier | NCT00118898 |
| Phase | Phase 3 |
| Condition | HIV infections |
| Population | HIV-infected treatment-naive adults |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | None |
| Primary purpose | Treatment |
| Enrollment | 1864 |
| Arms | 4 |
| Trial status | Completed |
| Start | 2005-09 |
| Primary completion | 2009-11 |
| Lead sponsor | Advancing Clinical Therapeutics Globally for HIV/AIDS and Other Infections |
| Sponsor type | Network |
2. Clinical Question
The study addressed a comparative treatment question in adults with HIV infection who had not previously received treatment. The factorial structure combined one of two antiretroviral agents with one of two nucleoside/nucleotide reverse transcriptase inhibitor backbones, producing four randomized treatment combinations.
Population
HIV-infected treatment-naive adults.
Efavirenz strategy
Efavirenz given with either emtricitabine/tenofovir disoproxil fumarate or abacavir/lamivudine.
Atazanavir strategy
Ritonavir-boosted atazanavir given with either emtricitabine/tenofovir disoproxil fumarate or abacavir/lamivudine.
Primary statistical question
How do the randomized treatment combinations differ with respect to time to virologic failure, grade 3/4 safety events, and time to treatment modification?
3. Trial Design
The registry describes the study as randomized, parallel, and unmasked. The intervention set contains seven listed intervention products, including active drugs and two placebo products used to support the study's treatment comparisons.
Efavirenz + FTC/TDF
- Efavirenz
- Emtricitabine/tenofovir disoproxil fumarate
- Placebo for abacavir/lamivudine
Efavirenz + ABC/3TC
- Efavirenz
- Abacavir/lamivudine
- Placebo for emtricitabine/tenofovir disoproxil fumarate
RTV-boosted atazanavir + FTC/TDF
- Atazanavir with ritonavir
- Emtricitabine/tenofovir disoproxil fumarate
- Placebo for abacavir/lamivudine
RTV-boosted atazanavir + ABC/3TC
- Atazanavir with ritonavir
- Abacavir/lamivudine
- Placebo for emtricitabine/tenofovir disoproxil fumarate
4. Primary Endpoints
The registry identifies three primary endpoints. Two are explicitly time-to-event outcomes, while the safety endpoint is defined as time from treatment dispensation to the first qualifying grade 3/4 safety event.
| Endpoint | Registry time frame | Definition | Type |
|---|---|---|---|
| Time From Randomization to Virologic Failure | Follow-up time was variable; median follow-up was 138 weeks | Blood samples were obtained at visit weeks 16 and 24 and every 12 weeks thereafter. Virologic failure was defined as a confirmed plasma HIV-1 RNA level ≥1000 copies/mL at or after 16 weeks after randomization and before 24 weeks, or ≥200 copies/mL at or after 24 weeks. | Time-to-event |
| Time From Treatment Dispensation to a Grade 3/4 Safety Event | All follow-up while on initially assigned regimen; the registry describes the median (25th, 75th percentile) follow-up while on the initial regimen was 120 (54, 156) weeks and the range was 0 to 205 weeks. | A grade 3 or 4 sign, symptom, or laboratory abnormality at least one grade higher than baseline. Total bilirubin and creatine kinase were excluded. Grading used the Division of AIDS 2004 Severity of Adverse Events Tables. | Other / unclear; time-to-event framework is appropriate |
| Time From Treatment Dispensation to Treatment Modification | Follow-up time was variable; median follow-up was 138 weeks | First modification of the regimen, including a permanent discontinuation, switch, or substitution. | Other / unclear; time-to-event |
5. Planned Analysis
ClinicalTrials.gov reports that results have been posted for all three registered primary endpoints. However, the ClinicalTrials.gov record contains no formal statistical analyses. That means there are no formal treatment-effect estimates, confidence intervals, or p-values in the registry record that can be reproduced on this page.
For an endpoint defined as time from randomization to virologic failure, the natural statistical framework is a survival analysis. A Kaplan-Meier estimator would describe the distribution of time to failure, while a log-rank test could compare randomized groups and a Cox proportional-hazards model could estimate a hazard ratio. Those methods are described here as the appropriate analytical framework for the endpoint; the ClinicalTrials.gov record does not establish that any particular one was used for the posted result.
The same time-to-event logic applies to time to treatment modification. The event is the first regimen modification, while participants who remain on their initially assigned regimen without modification contribute follow-up until censoring. The safety endpoint can likewise be analyzed as time to the first qualifying grade 3/4 event while on the initially assigned regimen.
This framework uses both whether an event occurred and when it occurred. It is therefore more informative than simply calculating the percentage of participants who experienced an event by an arbitrary fixed date when follow-up times differ.
6. Results Reporting Status
The trial has Results Posted: Yes and lists 22 outcome measures posted. Results are therefore available in the registry record, but the ClinicalTrials.gov record does not include formal statistical-analysis objects for the three primary endpoints.
| Primary endpoint | Results posted? | Formal statistical analysis reported? | What can be reported here |
|---|---|---|---|
| Time From Randomization to Virologic Failure | Yes | No | Endpoint definition and follow-up framework; no formal comparative estimate posted |
| Time From Treatment Dispensation to a Grade 3/4 Safety Event | Yes | No | Endpoint definition plus reported serious-adverse-event counts separately below |
| Time From Treatment Dispensation to Treatment Modification | Yes | No | Endpoint definition and appropriate time-to-event framework |
7. Safety Results
The ClinicalTrials.gov record provides serious adverse events by arm as affected participants divided by the corresponding at-risk denominator.
| Regimen | Serious adverse events | Affected / at risk |
|---|---|---|
| EFV, FTC/TDF, and placebo ABC/3TC | Serious adverse events | 48/461 |
| EFV, placebo FTC/TDF, and ABC/3TC | Serious adverse events | 62/461 |
| RTV-boosted ATV, FTC/TDF, and placebo AB | Serious adverse events | 58/464 |
| RTV-boosted ATV, placebo FTC/TDF, and AB | Serious adverse events | 71/462 |
The bars above are scaled only to the largest reported affected count, 71; they are a visual representation of the raw affected-participant counts and are not estimates of event risk. No statistical comparison should be inferred from the visual display.
The reported serious-adverse-event counts describe how many participants were affected in each safety group relative to the registry-reported at-risk denominator. They do not by themselves establish that one regimen has a higher or lower safety-event rate.
A formal comparison would need to account for the endpoint definition, follow-up while on the initially assigned regimen, censoring, and the prespecified analysis population. For a time-to-first-event endpoint, simply comparing raw counts can discard information about when events occurred.
The distinction between a serious adverse event and the registered primary safety endpoint is also important: the registered endpoint is time to a grade 3/4 safety event, whereas the registry-reported arm-level summary is specifically labelled serious adverse events. They should not be treated as interchangeable outcomes.
8. Statistical Methodology
Time-to-event analysis
Two of the three primary endpoints are explicitly described as time-to-event outcomes, and the safety endpoint is also framed as time from treatment dispensation to the first qualifying event. Time-to-event methods are designed for studies in which participants may have different amounts of observable follow-up.
Here, di represents events at a particular event time and ni represents participants at risk immediately before that time. The resulting curve estimates the probability of remaining event-free beyond time t.
Kaplan-Meier estimation
For virologic failure, Kaplan-Meier estimation would allow the trial to represent the probability of remaining free of the registry-defined virologic-failure event over follow-up. The same approach can be used for treatment modification and other time-to-first-event endpoints.
Log-rank comparison
A log-rank test is a standard way to compare survival or event-free curves between randomized groups. In a four-arm trial, the exact comparison structure matters: a global four-group comparison, pairwise comparisons, or comparisons of the two treatment components answer different questions.
Cox proportional-hazards model
A Cox model can summarize a relative time-to-event effect using a hazard ratio. The basic interpretation is conditional on the model and its assumptions; it is not equivalent to a risk ratio or an absolute difference in event probability.
A hazard ratio describes relative event rates over time under a fitted model. It does not mean that the same percentage of participants experienced a reduction in risk, and it does not by itself provide an absolute clinical effect.
Analysis of the safety endpoint
The registry defines the safety endpoint as time from treatment dispensation to a grade 3/4 event at least one grade higher than baseline, excluding total bilirubin and creatine kinase. The time origin therefore differs from that of virologic failure: treatment dispensation rather than randomization.
Treatment modification
Treatment modification is defined as the first modification of the regimen, including permanent discontinuation, switch, or substitution. This definition makes the endpoint naturally suited to a time-to-first-event analysis.
9. Statistical Methods Explained
Why is randomization important?
Randomization creates the framework for comparing treatment groups without assigning treatment according to participant characteristics. If the randomization process is properly implemented, differences in outcomes can be interpreted within the causal framework created by the assigned treatment groups, subject to follow-up, adherence, missingness, and other trial limitations.
Why use time-to-event analysis for virologic failure?
Participants can experience virologic failure at different times, and some may remain free of failure during their available follow-up. A time-to-event analysis uses the timing of failure and appropriately handles participants whose event has not been observed by the end of follow-up.
What does a Kaplan-Meier curve represent?
A Kaplan-Meier curve estimates the probability of remaining free of the event over time. A downward step occurs when an event is observed; censoring changes the number at risk without being treated as an event.
What would a hazard ratio mean in this trial?
For virologic failure, a hazard ratio below 1 for one randomized regimen relative to another would indicate a lower estimated instantaneous rate of virologic failure under the fitted proportional-hazards model. It would not mean that the same percentage of participants avoided failure or that the absolute difference in failure probability was equal to the hazard-ratio difference.
Why does censoring matter?
Censoring occurs when the event status is not observed beyond a participant's available follow-up. Survival methods can incorporate censored observations, but interpretation depends on assumptions about the relationship between censoring and the underlying event process. The ClinicalTrials.gov record does not provide a detailed missing-data or censoring analysis plan.
Why should four arms not automatically be treated as one two-group comparison?
The study contains four treatment combinations. A statistical comparison can target the four individual regimens, the two third-drug strategies, the two backbone strategies, or possible interaction between those treatment components. Each question has a different estimand and multiplicity structure. The ClinicalTrials.gov record does not specify which factorial contrasts were prespecified.
10. Factorial Structure and Interpretation
The four-arm structure is one of the most statistically informative features of ACTG A5202. It creates a natural two-by-two treatment framework:
| FTC/TDF | ABC/3TC | |
|---|---|---|
| Efavirenz | EFV + FTC/TDF + placebo ABC/3TC | EFV + placebo FTC/TDF + ABC/3TC |
| RTV-boosted atazanavir | RTV-boosted ATV + FTC/TDF + placebo ABC/3TC | RTV-boosted ATV + placebo FTC/TDF + ABC/3TC |
This structure permits several scientifically distinct questions. One can compare the two backbone strategies within the efavirenz stratum, compare them within the ritonavir-boosted atazanavir stratum, or examine the corresponding third-drug comparisons within each backbone.
A factorial analysis can also estimate main effects and test whether the effect of one treatment component depends on the level of the other component. Such dependence is an interaction. However, the ClinicalTrials.gov record does not report a factorial statistical-analysis plan or an interaction estimate, so no interaction conclusion is made here.
The four-arm design contains more information than four unrelated treatment groups. Its structure makes it possible to ask whether treatment-component effects are consistent across the other component. But the existence of a factorial design does not automatically prove that a particular main-effect or interaction analysis was performed.
11. Crossover, Blinding, and Treatment Assignment
The registry identifies the study as unmasked. It also states that EFV and ritonavir-boosted ATV were open-label, and that blinded FTC/TDF versus ABC/3TC were unblinded in the ≥100,000 copies/mL screening viral-load stratum in February 2008.
Open-label components
Efavirenz and ritonavir-boosted atazanavir were open-label according to the registry caveat.
Backbone unblinding
Blinded FTC/TDF versus ABC/3TC were unblinded in the ≥100,000 copies/mL screening viral-load stratum in February 2008.
Why masking matters
When participants or investigators know treatment assignment, subjective treatment decisions and assessment of some outcomes can potentially be influenced by that knowledge.
Time origin matters
Virologic failure begins at randomization, whereas the safety and treatment-modification endpoints begin at treatment dispensation.
The ClinicalTrials.gov record does not describe a treatment crossover procedure. Accordingly, no crossover effect or adjustment is inferred.
12. Multiplicity and Multiple Treatment Comparisons
Three primary endpoints were registered, and the trial has four treatment arms. Those features create multiple potential statistical comparisons. The appropriate interpretation depends on which comparisons were prespecified as confirmatory and how the familywise type I error was controlled.
| Source of multiplicity | Why it matters | What the ClinicalTrials.gov record establishes |
|---|---|---|
| Three primary endpoints | Testing several primary outcomes can increase the chance of at least one false-positive result if each is tested independently at the same nominal level. | Three primary endpoints were registered. |
| Four randomized arms | Multiple pairwise comparisons can create additional type I error if all are treated as confirmatory tests. | Four arms were registered. |
| Factorial treatment structure | Main effects and interactions represent different hypotheses and require a prespecified testing strategy. | The intervention list and four-arm structure support the factorial interpretation; no formal factorial analysis is reported. |
13. Interim Analysis, Missing Data, and Bayesian Methods
Interim analysis
No interim-analysis plan or interim efficacy boundary is included in the ClinicalTrials.gov record. No interim statistical conclusion is therefore reported.
Missing data / imputation
No imputation strategy or detailed missing-data analysis is included in the ClinicalTrials.gov record.
Bayesian methods
No Bayesian analysis is identified in the ClinicalTrials.gov record.
Stratification
No stratification factors are reported in the ClinicalTrials.gov record.
These omissions are important because they define the boundary between what can be reconstructed from the registry data and what would require the protocol, statistical analysis plan, or a publication containing the formal analysis.
14. Limitations
- No formal statistical-analysis records reported: The trial data contain no formal statistical analyses, so treatment-effect estimates, confidence intervals, and p-values cannot be reported from the ClinicalTrials.gov recordset.
- Results are posted but incompletely represented in the registry-reported extract: ClinicalTrials.gov reports 22 outcome measures, but the ClinicalTrials.gov record does not reproduce their numerical results.
- Open-label treatment components: EFV and ritonavir-boosted ATV were open-label according to the registry caveat.
- Partial unblinding: Blinded FTC/TDF versus ABC/3TC were unblinded in the ≥100,000 copies/mL screening viral-load stratum in February 2008.
- Resistance testing: HIV-1 drug resistance testing was not standard of care during enrollment.
- HLA-B*5701 testing: HLA-B*5701 testing was not standard of care during enrollment.
- Endpoint distinction: Serious adverse-event counts reported by arm are not interchangeable with the registered time-to-grade-3/4-safety-event endpoint.
- Factorial interpretation: The four-arm structure supports factorial questions, but the ClinicalTrials.gov record does not identify the prespecified factorial contrasts or interaction analysis.
- Censoring and follow-up: Follow-up time was variable, and detailed censoring rules are not reported beyond the endpoint definitions.
15. Registry-Specific Caveats
Two clinical-practice changes are explicitly identified by the registry as limitations of the enrollment era: HIV-1 drug resistance testing and HLA-B*5701 testing were not standard of care during enrollment.
16. Why This Trial Matters Statistically
ACTG A5202 is a useful teaching case because it combines a randomized four-arm structure with several distinct time-to-event endpoints. It also illustrates why the definition of the event, time origin, censoring framework, analysis population, and comparison structure matter as much as the final numerical estimate.
| Concept | How it appears in ACTG A5202 |
|---|---|
| Randomization | 1864 participants were enrolled into a randomized parallel-group phase 3 trial. |
| Four-arm design | Efavirenz and ritonavir-boosted atazanavir were each paired with FTC/TDF or ABC/3TC. |
| Factorial structure | The four regimens form a natural two-by-two treatment-component structure. |
| Time-to-event analysis | Virologic failure and treatment modification are explicitly defined using time from treatment assignment or treatment dispensation. |
| Kaplan-Meier estimation | Appropriate for representing event-free probability over variable follow-up. |
| Hazard ratio | A natural summary measure for randomized time-to-event comparisons, although no HR is reported in the provided statistical-analysis data. |
| Safety analysis | Grade 3/4 safety events are defined using a severity grading system and a baseline-change criterion. |
| Censoring | Relevant because follow-up was variable and participants may remain event-free at the end of observation. |
| Multiplicity | Three registered primary endpoints and four treatment arms create multiple potential hypotheses. |
| Masking | The registry identifies the overall study as unmasked and describes specific open-label and later-unblinded components. |
| Historical context | Resistance and HLA-B*5701 testing were not standard of care during enrollment. |
17. What a Formal Survival Analysis Would Need to Show
A complete statistical analysis of the virologic-failure endpoint would ideally present the event definition, number of events, number censored, Kaplan-Meier estimates, treatment contrasts, confidence intervals, and the statistical test or regression model used for the comparison.
The same logic applies to treatment modification, except that the time origin is treatment dispensation and the event is the first permanent discontinuation, switch, or substitution. For the safety endpoint, the event is the first qualifying grade 3/4 sign, symptom, or laboratory abnormality meeting the registry definition.
A statistical analysis is only meaningful relative to its estimand. Changing the event definition, time origin, censoring rule, or analysis population can change the resulting estimate even when the underlying patients are identical. That is why a registry endpoint should be reproduced precisely before choosing a statistical model.
18. Interpreting the Serious Adverse-Event Counts
The registry-reported safety counts illustrate another important statistical distinction: a numerator and denominator are descriptive, while an effect estimate is comparative.
| Regimen | Affected | At risk | What the ClinicalTrials.gov record establishes |
|---|---|---|---|
| EFV + FTC/TDF | 48 | 461 | 48 participants are reported as affected among 461 at risk. |
| EFV + ABC/3TC | 62 | 461 | 62 participants are reported as affected among 461 at risk. |
| RTV-boosted ATV + FTC/TDF | 58 | 464 | 58 participants are reported as affected among 464 at risk. |
| RTV-boosted ATV + ABC/3TC | 71 | 462 | 71 participants are reported as affected among 462 at risk. |
These values do not tell us how quickly the events occurred, whether participants experienced more than one event, how follow-up differed between groups, or what formal statistical model was prespecified. Those questions matter when the registered primary safety endpoint is explicitly time-to-event.
19. Statistical Interpretation vs Clinical Interpretation
Statistical interpretation
The study is a randomized four-arm phase 3 trial with three registered primary endpoints. Two endpoints are explicitly time-to-event, and the safety endpoint is also defined by time from treatment dispensation to a qualifying event. The ClinicalTrials.gov record does not include formal statistical-analysis estimates.
Clinical interpretation
The trial compares four specified antiretroviral treatment combinations in treatment-naive adults with HIV infection. The ClinicalTrials.gov record identifies the treatment regimens and safety-event counts but does not support an additional comparative clinical conclusion beyond those reported facts.
This separation is important. A registry record can establish what was studied and what outcomes were posted without providing enough statistical detail to reconstruct every treatment-effect estimate. A rigorous analysis should preserve that distinction rather than filling gaps with estimates from another source or from memory.
20. Important Questions About the Primary Endpoints
Virologic failure
The registry-defined virologic-failure endpoint uses different HIV-1 RNA thresholds at different stages of follow-up: a confirmed level of ≥1000 copies/mL at or after 16 weeks and before 24 weeks, or ≥200 copies/mL at or after 24 weeks. Blood samples were scheduled at weeks 16 and 24 and every 12 weeks thereafter.
Grade 3/4 safety event
The safety endpoint is based on a grade 3 or 4 sign, symptom, or laboratory abnormality that is at least one grade higher than baseline. Total bilirubin and creatine kinase are excluded, and grading used the Division of AIDS 2004 Severity of Adverse Events Tables.
Treatment modification
The endpoint captures the first modification of the assigned regimen, including permanent discontinuation, switch, or substitution. Because only the first modification counts for this endpoint, the statistical question concerns time to the first qualifying modification rather than the total number of modifications over follow-up.
21. Trial History and Follow-Up
Trial start
The registry lists September 2005 as the study start.
Backbone unblinding in a screening viral-load stratum
The registry caveat states that blinded FTC/TDF versus ABC/3TC were unblinded in the ≥100,000 copies/mL screening viral-load stratum in February 2008.
Primary completion
The registry lists November 2009 as the primary completion date.
Results posted
The trial is listed as completed, with results posted and 22 outcome measures reported in the ClinicalTrials.gov record.
22. How to Read a Future Hazard Ratio or Confidence Interval
If a formal analysis reports a hazard ratio for virologic failure, the estimate should be read relative to the specified reference group. An HR below 1 would indicate a lower estimated instantaneous rate of failure for the numerator treatment group under the fitted model; an HR above 1 would indicate a higher estimated rate.
The 95% confidence interval would describe statistical uncertainty around that estimate under the specified model and sampling framework. It would not describe the range of effects experienced by individual patients.
A p-value, if reported, would address the prespecified statistical hypothesis under the relevant testing framework. It would not measure the magnitude of treatment effect, clinical importance, or probability that a treatment is effective.
If a Cox model were used, the hazard ratio would rely on the model's proportional-hazards interpretation. When the relative hazard changes substantially over time, a single HR may provide an incomplete description of the treatment effect. The ClinicalTrials.gov record does not provide enough information to assess that assumption for ACTG A5202.
No non-inferiority margin is reported in the ClinicalTrials.gov record. Therefore, this page does not characterize any ACTG A5202 comparison as a non-inferiority test or infer non-inferiority from the absence of a reported difference.
23. Related Tutorials
Learn more about the methods used in this trial:
24. Related Calculators
25. Sources
- ClinicalTrials.gov: ACTG A5202 — NCT00118898.
- PubMed: PubMed record 38160309.
- PubMed: PubMed record 33901188.
- PubMed: PubMed record 32829410.
- PubMed: PubMed record 30642925.
- PubMed: PubMed record 26009829.
Continue through the Clinical Biostats statistical pathway
Connect randomized-trial design, time-to-event endpoints, survival analysis, and statistical calculators through the broader Clinical Biostats tutorial and tools collection.
26. Record Summary
ACTG A5202 is a statistically instructive randomized phase 3 trial because its four-arm structure combines two antiretroviral treatment components, while its three registered primary endpoints use time-to-event concepts in different ways. Virologic failure is measured from randomization using specific HIV-1 RNA thresholds and scheduled assessments; the safety endpoint begins at treatment dispensation and uses a grade 3/4 severity definition; and treatment modification is defined as the first permanent discontinuation, switch, or substitution.
The ClinicalTrials.gov record reports 1864 enrolled participants, four arms, a median follow-up of 138 weeks for the virologic-failure endpoint, 22 posted outcome measures, and serious-adverse-event counts for each regimen. They do not provide formal statistical-analysis records. The appropriate statistical interpretation is therefore to distinguish the registered estimands and event definitions from the formal treatment-effect estimates that would require the underlying analysis results.