This page separates reported trial results from statistical interpretation. The numerical results presented here are limited to the ClinicalTrials.gov record. ClinicalTrials.gov provides the official trial registry record.
1. Trial at a Glance
HPTN 052 was a completed phase 3 randomized parallel trial in HIV infections with a primary purpose of prevention. The registry reports 3526.0 participants, 2 arms, 2 registered primary endpoints, and 2 posted statistical analyses, both using Cox proportional-hazards models with hazard ratios as the effect measure.
| Feature | HPTN 052 |
|---|---|
| Trial name | HPTN 052 |
| Brief title | Preventing Sexual Transmission of HIV With Anti-HIV Drugs |
| Phase | Phase 3 |
| Condition | HIV Infections |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Single |
| Primary purpose | Prevention |
| Enrollment | 3526.0 |
| Status | Completed |
| Start | 2005-02 |
| Primary completion | 2015-05 |
| Lead sponsor | National Institute of Allergy and Infectious Diseases (NIAID) |
| Sponsor type | NIH |
| ClinicalTrials.gov | NCT00074581 |
2. Clinical Question
The registered trial evaluated HIV infection rates among partners of randomized HIV-infected index cases, comparing the Early-ART and Delayed-ART arms. The primary analysis included incident HIV infections occurring in HIV-negative partners at enrollment and, for the linked-infection endpoint, required confirmation by genotyping that the viral envelope sequence in the index case matched that of the partner.
Population
Partners who were HIV-negative at enrollment of randomized HIV-infected index cases, with the primary endpoint specifically requiring infections to be linked to the index partner by genotyping.
Intervention strategy
Early antiretroviral therapy, represented in the statistical analyses as the Early-ART arm.
Comparator
Delayed antiretroviral therapy, represented in the statistical analyses as the Delayed-ART arm.
Primary question
How does early ART compare with delayed ART with respect to partner HIV infection rates throughout the study?
3. Trial Design
Interventions recorded in the registry
The registry lists the following drug interventions: Atazanavir, Didanosine, Efavirenz, Emtricitabine/Tenofovir disoproxil fumarate, Lamivudine, Lopinavir/Ritonavir, Nevirapine, Stavudine, Tenofovir disoproxil fumarate, and Zidovudine/Lamivudine.
4. Registered Primary Endpoints
| Endpoint | Time frame | Registry definition |
|---|---|---|
| Linked Partner HIV Infection Rates in Early-ART and Delayed-ART Arms | Throughout study | Incident HIV infections occurring in the partners (HIV-negative at enrollment) of randomized HIV-infected index (HIV-positive at enrollment) cases are assessed, by arm. Only acquisition from the index partner were included in the primary analysis, therefore, each endpoint was required to be confirmed (by genotyping) such that the viral envelop sequence in the index case matched that of the partner. |
| All Partner HIV Infection Rates in Early-ART and Delayed-ART Arms | Throughout study | All incident HIV infections occurring in the partners (HIV-negative at enrollment) of randomized HIV-infected index (HIV-positive at enrollment) cases are assessed, by arm. |
Both registered primary endpoints are time-to-event outcomes in the posted statistical analyses. The registry reports an outcome unit of event rate per 100 person-yr, while the formal effect measure reported for each analysis is a hazard ratio.
5. Analysis Populations and Event Definitions
| Analysis | Population / definition reported in registry data |
|---|---|
| Linked partner infection analysis | Partner HIV infections attributable to the randomized HIV-infected index partner, confirmed by genotyping through matching viral envelope sequences. |
| All partner infection analysis | All partners who were HIV-negative at enrollment of randomized HIV-infected index cases, by randomized arm. |
The distinction between these two endpoints is statistically important. The linked endpoint restricts events to infections confirmed as originating from the index partner. The all-infection endpoint includes all incident HIV infections among the specified partners, regardless of whether the infection was established as linked to the index partner.
6. Statistical Methodology
Cox proportional-hazards model
The registry identifies Regression, Cox as the reported method for both primary analyses. The statistical method is a Cox proportional-hazards model, placing the analysis within the survival-analysis family.
For a two-group comparison, the hazard ratio is the ratio of the modeled hazard in one group to that in the reference group. In the posted analyses, Early-ART is the numerator and Delayed-ART is the denominator.
Hazard ratio
A hazard ratio is a relative time-to-event measure. An HR below 1 indicates a lower estimated instantaneous event rate in the numerator group relative to the denominator group, under the fitted Cox model.
Thus, an HR of 0.31 corresponds to an estimated instantaneous event rate that is 31% of the corresponding rate in the Delayed-ART group, or equivalently a 69% lower estimated hazard.
Two-sided confidence intervals
Both posted analyses report 95% two-sided confidence intervals. The interval describes statistical uncertainty around the estimated hazard ratio under the analysis framework; it is not a range containing the individual-level effects experienced by every participant.
Superiority hypothesis
The registry identifies the hypothesis type for both analyses as superiority. This is distinct from a non-inferiority framework: the objective is not to establish that the treatment is no worse than a prespecified margin, but to evaluate a superiority comparison between Early-ART and Delayed-ART.
7. Statistical Methods Explained
What does the Cox proportional-hazards model contribute?
The Cox model is designed for time-to-event data, where both whether an event occurs and when it occurs contribute information. It produces a hazard ratio that summarizes the relative event rate between groups over follow-up, subject to the model's assumptions.
Why is the hazard ratio appropriate for a partner HIV infection endpoint?
The registered endpoints are evaluated throughout the study and are expressed in terms of event rates per 100 person-years. A time-to-event framework therefore preserves information about the timing of infection and accommodates differing amounts of follow-up through the survival-analysis framework.
What does an HR of 0.07 mean?
The posted linked-infection analysis compares Early-ART with Delayed-ART and reports HR 0.07. On the hazard-ratio scale, this means the estimated instantaneous infection hazard in Early-ART was 7% of that in Delayed-ART, corresponding to a 93% lower estimated hazard.
What does an HR of 0.31 mean?
The all-infection analysis reports HR 0.31. This means the estimated instantaneous infection hazard in Early-ART was 31% of that in Delayed-ART, corresponding to a 69% lower estimated hazard.
Why does the confidence interval matter?
The point estimate is only one representation of the evidence. For the linked endpoint, the 95% CI is 0.02–0.22. For all partner infections, the 95% CI is 0.19–0.53. These intervals communicate the statistical precision of the estimated hazard ratios. A narrower interval generally indicates greater precision than a wider interval, although precision also depends on the underlying design and information available.
Why does the p-value not measure the size of the effect?
A p-value addresses compatibility with a null hypothesis under the statistical model; it is not a percentage reduction, probability that the treatment works, or measure of clinical magnitude. The magnitude of the estimated effect is conveyed by the hazard ratio, while the confidence interval describes uncertainty around that estimate.
What assumption deserves attention with a Cox model?
The proportional-hazards model assumes that the relative hazard represented by the model is appropriately characterized over time. A single hazard ratio can be less descriptive if proportional hazards do not adequately represent how the two event rates evolve during follow-up. The ClinicalTrials.gov record does not report a formal proportional-hazards diagnostic.
8. Primary Result: Linked Partner HIV Infection Rates
The first posted primary analysis concerns Linked Partner HIV Infection Rates in Early-ART and Delayed-ART Arms. The time frame is throughout study. The outcome unit is event rate per 100 person-yr, and the reported analysis uses a Cox proportional-hazards model.
Hazard ratio for linked partner HIV infection
95% CI: 0.02–0.22 · P < 0.0001
Early-ART vs Delayed-ART · Two-sided 95% CI · Superiority
| Analysis feature | Reported result |
|---|---|
| Endpoint | Linked Partner HIV Infection Rates in Early-ART and Delayed-ART Arms |
| Time frame | Throughout study |
| Outcome unit | Event rate per 100 person-yr |
| Groups compared | Early-ART vs Delayed-ART |
| Method | Regression, Cox |
| Effect measure | Hazard Ratio (HR) |
| Estimate | 0.07 |
| 95% CI | 0.02–0.22 |
| P-value | <0.0001 |
| Hypothesis | Superiority |
The reported HR of 0.07 compares the Early-ART arm with the Delayed-ART arm. In the registry's analysis notes, this is interpreted as a 93% lower estimated risk of infection among linked partner infections during the entire study.
The HR should not be read as saying that 93% of partners were protected, that 93% of infections were prevented in every individual, or that the probability of infection for an individual participant was exactly 93% lower. It is a model-based relative measure of the event hazard.
The 95% CI of 0.02–0.22 indicates uncertainty around the estimated hazard ratio. It does not describe the range of individual patient effects. The fact that the interval remains below 1 is consistent with a lower estimated hazard in Early-ART under the fitted model.
The P < 0.0001 result addresses the statistical evidence against the relevant null hypothesis; it does not quantify the magnitude of the treatment effect. The HR and its confidence interval are needed to describe magnitude and precision.
Because this is a Cox analysis, interpretation also depends on the appropriateness of the proportional-hazards assumption and on how censoring and event timing were handled. The ClinicalTrials.gov record does not provide additional diagnostics or censoring details.
9. Primary Result: All Partner HIV Infection Rates
The second posted primary analysis evaluates All Partner HIV Infection Rates in Early-ART and Delayed-ART Arms throughout the study. Unlike the linked endpoint, this outcome includes all incident HIV infections among the specified HIV-negative partners.
Hazard ratio for all partner HIV infection
95% CI: 0.19–0.53 · P < 0.0001
Early-ART vs Delayed-ART · Two-sided 95% CI · Superiority
| Analysis feature | Reported result |
|---|---|
| Endpoint | All Partner HIV Infection Rates in Early-ART and Delayed-ART Arms |
| Time frame | Throughout study |
| Outcome unit | Event rate per 100 person-yr |
| Analysis population | Population includes all partners (HIV-negative at enrollment) of randomized HIV-infected index (HIV-positive at enrollment) cases, by arm. |
| Groups compared | Early-ART vs Delayed-ART |
| Method | Regression, Cox |
| Effect measure | Hazard Ratio (HR) |
| Estimate | 0.31 |
| 95% CI | 0.19–0.53 |
| P-value | <0.0001 |
| Hypothesis | Superiority |
The reported HR of 0.31 means that the estimated instantaneous infection hazard in the Early-ART arm was 31% of the corresponding hazard in the Delayed-ART arm. The registry analysis notes characterize this as a 69% lower estimated risk of infection among all partner infections during the entire study.
This does not mean that 69% of all partners necessarily avoided infection because of treatment, nor does it imply that every participant experienced the same proportional reduction. The hazard ratio is a relative, model-based summary of event rates over time.
The 95% CI of 0.19–0.53 expresses uncertainty around the estimated HR. It does not represent the range of individual responses or individual infection probabilities. Its lower and upper limits remain below 1, which is consistent with a lower estimated hazard in Early-ART under the reported Cox analysis.
The P < 0.0001 indicates strong statistical evidence against the null hypothesis specified for the superiority analysis, but a p-value is not an effect-size measure. The HR and confidence interval provide the information needed to understand relative magnitude and precision.
The analysis remains subject to the assumptions of the Cox model, including the proportional-hazards framework. The ClinicalTrials.gov record does not report a formal assessment of that assumption.
10. Comparing the Two Primary Endpoints
| Feature | Linked partner infections | All partner infections |
|---|---|---|
| Endpoint | Linked Partner HIV Infection Rates in Early-ART and Delayed-ART Arms | All Partner HIV Infection Rates in Early-ART and Delayed-ART Arms |
| Time frame | Throughout study | Throughout study |
| Event definition | Incident partner infection confirmed as acquired from the index partner by genotyping | All incident HIV infections among the specified partners |
| Effect measure | Hazard ratio | Hazard ratio |
| HR | 0.07 | 0.31 |
| 95% CI | 0.02–0.22 | 0.19–0.53 |
| P-value | <0.0001 | <0.0001 |
The two endpoints answer related but different questions. The linked endpoint focuses specifically on HIV acquisition confirmed as coming from the randomized index partner. The all-infection endpoint broadens the event definition to all incident HIV infections among the specified partners. Consequently, their hazard ratios should not be treated as interchangeable estimates of exactly the same outcome.
11. Kaplan-Meier Estimation and Time-to-Event Thinking
The ClinicalTrials.gov record identifies both primary endpoints as time-to-event outcomes and report Cox proportional-hazards models. A natural companion to a Cox analysis is Kaplan-Meier estimation, which describes the estimated event-free survival function over time.
Here, di represents events at time ti and ni represents the number at risk immediately before that time.
Kaplan-Meier estimation and Cox regression answer related questions in different ways. Kaplan-Meier methods describe the event-free probability over time, while the Cox model summarizes the relative hazard between groups. A hazard ratio therefore should not be mistaken for a direct comparison of cumulative infection probabilities at one particular time point.
12. P-Values and Confidence Intervals
Linked endpoint
HR 0.07 with a two-sided 95% CI of 0.02–0.22 and P < 0.0001.
All-infection endpoint
HR 0.31 with a two-sided 95% CI of 0.19–0.53 and P < 0.0001.
What the CI provides
The confidence interval communicates uncertainty around the estimated hazard ratio and gives more information about precision than the p-value alone.
What the p-value provides
The p-value addresses evidence against the relevant null hypothesis; it does not measure effect size or clinical importance.
13. Safety
The ClinicalTrials.gov record reports serious adverse events by randomized arm as affected participants over participants at risk:
| Arm | Serious adverse events affected / at risk |
|---|---|
| Early-ART | 526/886 |
| Delayed-ART | 522/877 |
These are the ClinicalTrials.gov record. The registry extract does not provide a formal effect estimate, confidence interval, or p-value for the serious-adverse-event comparison, so no comparative inferential statistic is added here.
14. Trial Timeline
Trial start
The registry records the study start as 2005-02.
Randomized prevention trial
The registry identifies HPTN 052 as a phase 3 randomized, parallel trial with a primary purpose of prevention.
Primary completion
The registry records primary completion as 2015-05.
Results posted
ClinicalTrials.gov reports that results have been posted, including 2 outcome measures and 2 statistical analyses.
15. Multiplicity, Interim Analysis, and Other Design Issues
The ClinicalTrials.gov record identifies 2 registered primary endpoints and 2 statistical analyses, both with superiority hypotheses. However, the ClinicalTrials.gov record do not report an alpha-allocation strategy, multiplicity adjustment procedure, interim-analysis schedule, stopping boundaries, or a formal hierarchy between the two primary endpoints.
| Design topic | What the ClinicalTrials.gov record supports |
|---|---|
| Primary endpoints | 2 registered primary endpoints |
| Statistical analyses | 2 posted statistical analyses |
| Hypothesis type | Superiority for both posted analyses |
| Multiplicity adjustment | Not reported in the ClinicalTrials.gov record |
| Interim analysis | Not reported in the ClinicalTrials.gov record |
| Alpha spending | Not reported in the ClinicalTrials.gov record |
| Non-inferiority margin | Not applicable to the reported superiority framework; no margin is reported |
| Bayesian methods | Not reported |
| Imputation / missing-data method | Not reported |
| Stratification variables | Not reported in the ClinicalTrials.gov record |
This distinction matters because a reader should not infer an unreported multiplicity or interim-monitoring strategy simply because a trial has more than one primary endpoint. The appropriate interpretation is limited to what the registry data document.
16. Crossover and Factorial Design
Crossover
The ClinicalTrials.gov record does not report a crossover design or crossover rate. No crossover effect on the hazard-ratio estimates is therefore inferred.
Factorial structure
The design is explicitly identified as parallel, and the ClinicalTrials.gov record does not report a factorial design.
For a randomized time-to-event analysis, treatment switching or crossover can complicate interpretation because the randomized assignment no longer corresponds perfectly to treatment exposure over the full follow-up period. Because crossover information is not provided in the ClinicalTrials.gov recordset, that issue is not incorporated into the reported interpretation here.
17. Missing Data, Censoring, and Analysis Assumptions
Time-to-event analysis naturally encounters participants who have not experienced the event by the time their available follow-up ends. Such observations are commonly handled through censoring rather than treating them as if they experienced the event. The Cox framework also relies on assumptions concerning the relationship between hazards over time.
Censoring
The ClinicalTrials.gov record identifies time-to-event analyses but do not provide the detailed censoring rules used in the statistical analysis.
Missing data
No imputation method is reported in the ClinicalTrials.gov record. No imputation procedure is therefore attributed to HPTN 052 here.
Proportional hazards
The Cox model is reported, but the ClinicalTrials.gov record does not provide a diagnostic assessment of the proportional-hazards assumption.
Event timing
The registered endpoints are evaluated throughout the study, making event timing an integral part of the analysis.
18. Why the Endpoint Definition Matters
The linked and all-infection analyses illustrate an important principle in clinical-trial statistics: the endpoint definition determines what the effect estimate means.
For the linked endpoint, an HIV infection contributes to the primary analysis only when acquisition from the index partner is confirmed through genotyping. This creates a more specific event definition focused on transmission linked to the randomized index case.
The all-infection endpoint instead counts all incident HIV infections occurring in the specified partners. That broader definition captures a different event process. It therefore produces a different estimand, even though the treatment comparison and statistical model are the same.
19. What the Hazard Ratio Does — and Does Not — Mean
The linked-infection HR of 0.07 means the estimated instantaneous infection hazard in Early-ART was 7% of the corresponding hazard in Delayed-ART under the reported Cox model. The registry's analysis notes describe this as a 93% lower estimated risk of infection among linked partner infections during the entire study.
It does not mean that 93% of individuals were protected, that 93% of all possible infections were prevented, or that every individual's infection probability was reduced by exactly 93%.
The all-infection HR of 0.31 means the estimated instantaneous infection hazard in Early-ART was 31% of that in Delayed-ART under the reported Cox model. The registry's analysis notes describe this as a 69% lower estimated risk of infection among all partner infections during the entire study.
Again, this is not an individual-level probability, an absolute risk reduction, or a statement that 69% of participants avoided infection.
The linked endpoint has a 95% CI of 0.02–0.22, while the all-infection endpoint has a 95% CI of 0.19–0.53. These intervals describe uncertainty around their respective hazard-ratio estimates. They do not describe the distribution of treatment effects across individual participants.
Both analyses report P < 0.0001. A p-value provides evidence relative to a null hypothesis under the specified statistical framework; it does not tell the reader how large the treatment effect is. The hazard ratio supplies the relative effect estimate, and the confidence interval supplies information about its precision.
20. Statistical Interpretation vs Clinical Interpretation
Statistical interpretation
Both posted primary analyses report hazard ratios below 1 using Cox proportional-hazards models. The linked endpoint has HR 0.07 (95% CI 0.02–0.22), and the all-infection endpoint has HR 0.31 (95% CI 0.19–0.53); both report P < 0.0001.
Clinical interpretation
The statistical results describe lower estimated infection hazards in Early-ART relative to Delayed-ART for the two registered partner-infection endpoints. Translating those relative effects into individual absolute infection probabilities requires information not contained in the registry-reported statistical summary.
21. Important Limitations and Interpretation Issues
- Endpoint distinction: linked partner infections and all partner infections are different endpoints and should not be treated as identical estimands.
- Hazard-ratio interpretation: HRs are relative, model-based time-to-event measures and are not equivalent to absolute risk reductions or individual probabilities.
- Proportional-hazards assumption: the Cox model requires an appropriate proportional-hazards framework, but the ClinicalTrials.gov record does not report a formal diagnostic.
- Confidence intervals: the reported intervals quantify uncertainty around the estimated HR; they do not describe individual-level treatment effects.
- Multiplicity: two registered primary endpoints are documented, but the ClinicalTrials.gov record does not report the multiplicity-control strategy.
- Interim analysis: no interim-analysis schedule or stopping boundary is reported, so none is inferred.
- Missing data: the ClinicalTrials.gov record does not identify an imputation strategy.
- Censoring: the endpoints are time-to-event outcomes, but detailed censoring rules are not included in the ClinicalTrials.gov record.
- Safety inference: serious-adverse-event counts are provided, but no formal comparative safety analysis is reported.
- Absolute risk: the statistical analyses posted on ClinicalTrials.gov report hazard ratios rather than absolute cumulative infection probabilities at specified time points.
22. Why This Trial Matters Statistically
HPTN 052 is a useful teaching case because it demonstrates how a randomized prevention trial can combine a clearly defined event process with time-to-event regression. The trial also illustrates why statistical interpretation cannot be separated from endpoint definition.
| Concept | How it appears in HPTN 052 |
|---|---|
| Randomization | The registry identifies randomized allocation. |
| Parallel design | The design model is parallel. |
| Time-to-event endpoint | Both posted primary analyses are identified as time-to-event outcomes. |
| Hazard ratio | Both primary analyses use hazard ratio as the effect measure. |
| Cox regression | Both analyses use Regression, Cox / Cox proportional-hazards model. |
| Confidence interval | Both analyses report two-sided 95% confidence intervals. |
| P-value | Both analyses report P < 0.0001. |
| Superiority | Both posted analyses are classified as superiority hypotheses. |
| Endpoint specificity | One endpoint requires genotypic confirmation linking infection to the index partner; the other includes all incident partner infections. |
| Safety analysis | Serious adverse events are reported as affected participants over those at risk by arm. |
The central statistical lesson is that an effect estimate is inseparable from its estimand. HR 0.07 and HR 0.31 are not competing answers to one identical question; they correspond to two different registered definitions of partner HIV infection.
23. A Practical Reading Sequence for the Results
This sequence prevents a common statistical error: beginning with the p-value and treating it as the primary description of the result. The endpoint should be understood first, followed by the treatment contrast, effect measure, confidence interval, and finally the hypothesis-test result.
24. Related Tutorials
Learn more about the methods used in this trial:
25. Related Statistical Calculators
26. Sources
- ClinicalTrials.gov: NCT00074581 — HPTN 052.
- PubMed: PMID 30157702.
- PubMed: PMID 27424812.
- PubMed: PMID 24602844.
- PubMed: PMID 21767103.
Continue through the Clinical Biostats statistical pathway
Explore the statistical methods that underlie randomized clinical trials, time-to-event endpoints, hazard ratios, confidence intervals, and hypothesis testing.
27. Record Summary
HPTN 052 provides a clear example of time-to-event statistical analysis in a randomized prevention trial. The ClinicalTrials.gov record documents two registered primary endpoints, both evaluated throughout the study and analyzed using Cox proportional-hazards models. The linked partner HIV infection analysis reports HR 0.07 (95% CI 0.02–0.22; P < 0.0001), while the all-partner-infection analysis reports HR 0.31 (95% CI 0.19–0.53; P < 0.0001). The estimates should be interpreted according to their distinct endpoint definitions rather than treated as interchangeable measures.
The most useful statistical reading combines the endpoint definition, randomized treatment comparison, Cox model, hazard ratio, confidence interval, and p-value. It also keeps the limitations of time-to-event modeling in view, particularly the interpretation of a hazard ratio, the proportional-hazards assumption, censoring, and the absence of additional multiplicity or interim-analysis details in the ClinicalTrials.gov record.