← Clinical Trials
HIV Infections Phase 2/3 Vaccine Prevention NCT02968849

HVTN 702: Complete Statistical Analysis of ALVAC/Bivalent gp120/MF59 in HIV Infections

An independent statistical review of the randomized phase 2/3 HVTN 702 study evaluating ALVAC-HIV (vCP2438) and bivalent subtype C gp120/MF59 versus placebo for prevention of HIV-1 infection in South Africa.

HVTN 702  ·  Phase 2/3  ·  Randomized  ·  Quadruple-masked  ·  Prevention
Scope of this record

This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record. Numerical trial results on this page are limited to the ClinicalTrials.gov record for HVTN 702.

1. Trial at a Glance

HVTN 702 was a randomized, parallel, quadruple-masked phase 2/3 prevention study evaluating ALVAC-HIV (vCP2438) and bivalent subtype C gp120/MF59 against placebo in participants at risk for HIV infection. The registry reports 5404 participants enrolled and two study arms.

5404
Enrollment
Participants
2
Study Arms
Vaccine vs placebo
1.02
24-Month HR
95% CI 0.81–1.30
1.05
36-Month HR
95% CI 0.85–1.28
FeatureHVTN 702
Trial nameHVTN 702
Brief titlePivotal Phase 2b/3 ALVAC/Bivalent gp120/MF59 HIV Vaccine Prevention Safety and Efficacy Study in South Africa
Phase2/3
StatusCompleted
ConditionHIV Infections
Primary purposePrevention
AllocationRandomized
Design modelParallel
MaskingQuadruple
Enrollment5404
Lead sponsorNational Institute of Allergy and Infectious Diseases (NIAID)
Sponsor typeNIH
Start2016-10-26
Primary completion2021-11-16
ClinicalTrials.govNCT02968849

2. Clinical Question

The central statistical question was whether participants assigned to the ALVAC-HIV (vCP2438) and bivalent subtype C gp120/MF59 vaccine regimen experienced a different incidence of diagnosed HIV-1 infection than participants assigned to placebo during the prespecified follow-up periods.

Population

Participants enrolled in the HVTN 702 HIV vaccine prevention study in South Africa, with efficacy analyses reported in the MITT cohort.

Intervention

ALVAC-HIV (vCP2438) and Bivalent Subtype C gp120/MF59, both identified in the registry as biological interventions.

Comparator

Placebo, identified in the registry as a biological intervention.

Primary question

How does the hazard of diagnosed HIV-1 infection in the vaccine group compare with that in the placebo group over the specified follow-up periods?

3. Trial Design

01
Enroll5404 participants
02
RandomizeTwo parallel arms
03
MaskQuadruple masking
04
FollowHIV-1 infection and safety
05
AnalyzeTime-to-event comparison
Allocation
Randomized allocation with a parallel design.
Masking
Quadruple masking.
Purpose
Prevention of HIV infection.
Statistical framework
Time-to-event analysis using sex-stratified log-rank testing and sex-stratified Cox proportional-hazards modeling, with Wald analyses also reported.
VACCINE ARM

ALVAC / Bivalent gp120/MF59

  • ALVAC-HIV (vCP2438)
  • Bivalent Subtype C gp120/MF59
  • Both classified as biological interventions in the registry
CONTROL ARM

Placebo

  • Placebo
  • Classified as a biological intervention in the registry

4. Trial Timeline

2016-10-26

Study start

The registry records the study start date as 2016-10-26.

2020-01-23

Primary-outcome supersession

The registry definition for the later HIV-1 infection endpoint states that a 23 January 2020 DSMB finding that monitoring boundaries for non-efficacy had been met resulted in the earlier primary outcome being superseded by the later primary outcome.

2021-11-16

Primary completion

The registry records primary completion on 2021-11-16.

5. Primary Endpoints

The registry lists 12 primary endpoints. Two are HIV-1 infection incidence endpoints with formal statistical analyses reported in the ClinicalTrials.gov record. The remaining primary endpoints concern reactogenicity, adverse events, serious adverse events, adverse events of special interest, chronic medical conditions, early study termination, and study product discontinuation.

Primary endpointTime frameEndpoint typeFormal analysis in the ClinicalTrials.gov record
Incidence Rate of HIV-1 Infection Diagnosed After Enrollment (Concurrent With First Vaccination) Through 24 Months After Enrollment Measured through 24 months after first vaccination Count / rate Yes
Number of Participants Reporting Local Reactogenicity Signs and Symptoms: Pain and/or Tenderness Measured through 3 full days following each vaccination Binary No statistical analysis reported
Number of Participants Reporting Local Reactogenicity Signs and Symptoms: Erythema and/or Induration Measured through 3 full days following each vaccination Binary No statistical analysis reported
Number of Participants Reporting Systemic Reactogenicity Signs and Symptoms Measured through 3 full days following each vaccination Binary No statistical analysis reported
Number of Participants Reporting Adverse Events (AEs), by Relationship to Study Product Measured through 30 days after each vaccination Binary No statistical analysis reported
Number of Participants Reporting Adverse Events (AEs), by Severity Grade Measured through 30 days after each vaccination Binary No statistical analysis reported
Number of Participants Reporting Serious Adverse Events (SAEs) Measured through 12 months after last vaccination Binary No statistical analysis reported
Number of Participants Reporting Adverse Events of Special Interest (AESIs) Measured through 12 months after last vaccination Binary No statistical analysis reported
Number of Participants Reporting New Chronic Medical Conditions Measured through 12 months after last vaccination Binary No statistical analysis reported
Number of Participants With Early Study Termination Associated With an AE or Reactogenicity Measured through study completion (through 6 months after confirmation of HIV-1 diagnosis for participants who acquired HIV and through 12 months after last vaccination for the rest) Binary No statistical analysis reported
Number of Participants With Study Product Discontinuation Associated With an AE or Reactogenicity Measured through study completion (through 6 months after confirmation of HIV-1 diagnosis for participants who acquired HIV and through 12 months after last vaccination for the rest) Binary No statistical analysis reported
Incidence Rate of HIV-1 Infections Diagnosed Following Enrollment and Throughout All Participant Follow-Up Measured through 36 months after first vaccination Count / rate Yes
Registry endpoint transition: the registry definition states that the 23 January 2020 DSMB finding that monitoring boundaries for non-efficacy had been met caused the original HIV-1 infection primary outcome to be superseded by the endpoint covering HIV-1 infections diagnosed following enrollment throughout all participant follow-up.

6. Endpoint Definitions and Measurement

HIV-1 infection incidence

The first infection endpoint concerns the incidence rate of HIV-1 infection diagnosed after enrollment, concurrent with first vaccination, through 24 months. The registry definition states that vaccine efficacy was calculated as 1 minus the hazard ratio for HIV-1 infection, with the hazard ratio estimated using a sex-stratified Cox proportional-hazards model and tested using a sex-stratified log-rank test. The definition also states that vaccine efficacy was measured using a ratio of cumulative incidences.

The later primary infection endpoint covers HIV-1 infections diagnosed following enrollment throughout all participant follow-up and has a time frame of 36 months after first vaccination. Its registry definition states that vaccine efficacy was calculated as 1 minus the hazard ratio, with the hazard ratio estimated using a sex-stratified Cox proportional-hazards model and tested using a sex-stratified log-rank test.

Reactogenicity

Local pain and/or tenderness, local erythema and/or induration, and systemic reactogenicity were measured through 3 full days following each vaccination. The registry states that these outcomes were graded according to the Division of AIDS Table for Grading the Severity of Adult and Pediatric Adverse Events, Version 2.1, July 2017, with the maximum grade observed for each symptom over the time frame presented.

Adverse events

Adverse events were measured through 30 days after each vaccination. For participants reporting multiple adverse events over the time frame, the maximum relationship was counted for the relationship-to-study-product endpoint, while the maximum severity grade was counted for the severity endpoint.

Serious adverse events and other longer-term safety endpoints

Serious adverse events and adverse events of special interest were measured through 12 months after last vaccination. The registry definition identifies potential immune-mediated diseases among the adverse events of special interest. New chronic medical conditions were defined as a new onset or exacerbation of a medical condition requiring 2 or more visits to a medical provider during a period of at least 30 days.

7. Statistical Methodology

Time-to-event analysis

The principal efficacy endpoint is a time-to-event outcome: the time from enrollment and first vaccination to diagnosis of HIV-1 infection. This structure matters because participants can have different lengths of follow-up and because some participants may remain infection-free at the end of their observed follow-up.

Core time-to-event quantity
HR = instantaneous event rate in vaccine group ÷ instantaneous event rate in placebo group

For HVTN 702, the reported hazard ratio uses the vaccine group as the numerator and placebo as the denominator. An HR of 1 represents equal estimated hazards; an HR above 1 indicates a higher estimated hazard in the vaccine group; an HR below 1 indicates a lower estimated hazard.

Sex-stratified Cox proportional-hazards model

The registry states that the hazard ratio was estimated using a sex-stratified Cox proportional-hazards model. Stratification allows the baseline hazard to differ across sex strata while estimating the treatment comparison within the specified survival-analysis framework.

The important distinction is between stratification and adjustment with an ordinary regression coefficient. In a stratified Cox model, the strata can have separate baseline hazards while the treatment effect is estimated across the strata. The registry-reported analysis notes specifically identify sex as the stratification factor.

Sex-stratified log-rank test

The reported log-rank analyses tested the hazard ratio using a sex-stratified log-rank test. The log-rank framework compares the observed and expected numbers of events between treatment groups over follow-up, taking the time ordering of events into account.

Why the log-rank test fits this endpoint
At each event time: observed events are compared with events expected under equal survival experience

Unlike a simple comparison of cumulative infection proportions at one fixed date, a survival-analysis test uses the information accumulated across the follow-up period and can accommodate censoring.

Wald / z-test analyses

The ClinicalTrials.gov record also reports Wald analyses for the primary infection endpoints. In the 24-month analysis, the hazard ratio was estimated using Nelson-Aalen cumulative hazard estimates. The same approach was reported for the 36-month analysis.

The Wald framework provides a model-based inferential route for testing and constructing uncertainty around an estimated effect. The ClinicalTrials.gov record reports these analyses separately from the log-rank analyses, so the two sets of estimates should not be silently substituted for one another.

Nelson-Aalen cumulative hazard estimates

The Wald analyses used Nelson-Aalen cumulative hazard estimates. Cumulative hazard describes the accumulated event hazard over time and is distinct from the survival probability itself. This provides a nonparametric foundation for describing event accumulation before forming the reported hazard-ratio estimate in the registry-reported analysis.

8. Primary Efficacy Results Through 24 Months

The first primary efficacy endpoint was the incidence rate of HIV-1 infection diagnosed after enrollment, concurrent with first vaccination, through 24 months after enrollment. The analysis population was the MITT cohort, and the comparison was vaccine versus placebo.

Sex-stratified log-rank analysis

Hazard ratio for HIV-1 infection

1.02

95% CI: 0.81–1.30   ·   P = 0.84

Analysis: sex-stratified log-rank test with a sex-stratified Cox proportional-hazards model for the hazard ratio.

Clinical Biostats interpretation

The estimated hazard ratio of 1.02 means that the estimated instantaneous hazard of diagnosed HIV-1 infection in the vaccine group was 1.02 times that in the placebo group under the reported Cox model. Because the vaccine group is the numerator and placebo is the denominator, the estimate is close to 1.

The estimate does not mean that exactly 2% more participants became infected, nor does it represent a 2% absolute increase in infection risk. A hazard ratio is a relative time-to-event measure rather than a difference in cumulative incidence.

The 95% confidence interval of 0.81–1.30 describes uncertainty around the estimated hazard ratio. Its width shows that the estimate is not known with arbitrary precision. The interval also spans 1, the value corresponding to equal hazards.

The P = 0.84 value is evidence about compatibility with the null comparison under the specified testing framework; it is not a measure of the size of the observed effect. A p-value should not be interpreted as the probability that the vaccine is effective or ineffective.

Because this is a Cox-model hazard ratio, interpretation also depends on the proportional-hazards framework. The ClinicalTrials.gov record does not provide a formal assessment of that assumption, so the single HR should not be treated as a complete description of how infection hazards behaved at every point in follow-up.

Wald analysis

Hazard ratio from Nelson-Aalen cumulative hazard estimates

1.03

95% CI: 0.81–1.31   ·   P = 0.83

Analysis: Wald; hazard ratio estimated using Nelson-Aalen cumulative hazard estimates.

Clinical Biostats interpretation

The Wald analysis produces an HR of 1.03, again with vaccine as the numerator and placebo as the denominator. This is a closely related estimate to the log-rank/Cox analysis, but it comes from the separately reported Nelson-Aalen-based approach.

The 95% CI of 0.81–1.31 indicates uncertainty around the estimate and includes 1. The P = 0.83 value addresses the corresponding statistical test; it does not quantify the clinical importance of the difference.

The close correspondence between the two reported 24-month estimates—1.02 and 1.03—does not justify combining them into a new estimate. They are separate registry-reported analyses and should be retained as such.

9. Primary Efficacy Results Through 36 Months

The later primary endpoint was the incidence rate of HIV-1 infections diagnosed following enrollment and throughout all participant follow-up, measured through 36 months after first vaccination. The analysis population was the MITT cohort.

Sex-stratified log-rank analysis

Hazard ratio for HIV-1 infection

1.05

95% CI: 0.85–1.28   ·   P = 0.66

Analysis: sex-stratified log-rank test with a sex-stratified Cox proportional-hazards model for the hazard ratio.

Clinical Biostats interpretation

The estimated HR of 1.05 means that the estimated instantaneous hazard of diagnosed HIV-1 infection in the vaccine group was 1.05 times that in the placebo group under the reported Cox model.

This is a relative hazard estimate, not a statement that 5% more participants became infected and not an absolute risk difference. It also does not mean that every participant had the same individual-level change in risk.

The 95% CI of 0.85–1.28 expresses uncertainty around the estimated hazard ratio. It includes 1, so the interval is compatible with equal hazards as well as with hazard ratios on either side of 1 within the interval.

The P = 0.66 value is not an effect-size measure. It describes the statistical evidence generated by the specified test; it should not be translated into a probability that the treatment works or does not work.

As with the 24-month analysis, the interpretation of the single HR depends on the Cox proportional-hazards framework and on how censoring and follow-up contribute information to the analysis.

Wald analysis

Hazard ratio from Nelson-Aalen cumulative hazard estimates

1.00

95% CI: 0.81–1.23   ·   P = 0.98

Analysis: Wald; hazard ratio estimated using Nelson-Aalen cumulative hazard estimates.

Clinical Biostats interpretation

The Wald analysis gives an HR of exactly 1.00 in the ClinicalTrials.gov record. With vaccine as numerator and placebo as denominator, an HR of 1.00 corresponds to equal estimated hazards under this analysis.

The 95% CI of 0.81–1.23 provides the uncertainty range around that estimate and includes values below and above 1. The P = 0.98 value indicates very little statistical evidence against the equal-hazard reference under this specific Wald analysis, but the p-value itself is not a measure of the magnitude or clinical importance of an effect.

The fact that the 36-month log-rank/Cox estimate is 1.05 while the separately reported Wald estimate is 1.00 is not a contradiction. They are results from different statistical procedures applied to the same endpoint and should be reported using their stated methods.

10. Primary Efficacy Results at a Glance

Endpoint / analysisPopulationHR95% CIP-value
24 months · sex-stratified log-rank / Cox MITT cohort 1.02 0.81–1.30 0.84
24 months · Wald / Nelson-Aalen MITT cohort 1.03 0.81–1.31 0.83
36 months · sex-stratified log-rank / Cox MITT cohort 1.05 0.85–1.28 0.66
36 months · Wald / Nelson-Aalen MITT cohort 1.00 0.81–1.23 0.98
Educational note: the ClinicalTrials.gov record provides hazard ratios, confidence intervals, and p-values but does not provide the underlying individual event and censoring times required to reconstruct a valid Kaplan-Meier curve. A graphical survival curve should therefore not be fabricated from these summary statistics.

11. Secondary Time-to-Event Results

The registry also reports several secondary HIV-1 infection analyses. These analyses use the same broad time-to-event framework but focus on a different at-risk cohort or prespecified participant subgroup.

Week 26 at-risk cohort: after Month 6.5 through 24 months

AnalysisPopulationHR95% CIP-value
Sex-stratified log-rank / Cox Week 26 at-risk cohort 1.15 0.84–1.58 0.39
Wald / Nelson-Aalen Week 26 at-risk cohort 1.12 0.81–1.54 0.50

The time frame for both analyses was measured after Month 6.5 through 24 months after first vaccination. The registry-reported analysis notes state that the vaccine group is the numerator and the placebo group is the denominator.

Female participants through 24 months

PopulationMethodHR95% CIP-value
Females in MITT cohort Sex-stratified log-rank / Cox 1.03 0.80–1.33 0.82
Males in MITT cohort Sex-stratified log-rank / Cox 0.99 0.50–1.98 0.98

Both analyses were measured through 24 months after first vaccination. The ClinicalTrials.gov record identifies these as secondary analyses rather than the primary treatment comparison.

Female participants by age

PopulationMethodHR95% CIP-value
Females in MITT cohort aged 25 or younger Sex-stratified log-rank / Cox 1.08 0.80–1.47 0.60
Females in MITT cohort older than 25 Sex-stratified log-rank / Cox 0.92 0.58–1.46 0.71

These analyses were also measured through 24 months after first vaccination. They are subgroup-specific estimates, and their confidence intervals reflect the precision of those particular subgroup comparisons.

Subgroup interpretation: the ClinicalTrials.gov record reports separate hazard ratios for females, males, and female age groups, but it does not provide an interaction test. Therefore, differences between the subgroup estimates should not be interpreted as proof that the treatment effect differs between those groups. Comparing whether one subgroup's p-value is smaller than another's is not an interaction test.

12. How to Read the Secondary Hazard Ratios

The secondary results illustrate why the point estimate alone is not enough. For example, the female analysis has an HR of 1.03 with a 95% CI of 0.80–1.33, while the male analysis has an HR of 0.99 with a much wider 95% CI of 0.50–1.98.

The wider interval in the male analysis indicates greater uncertainty in that subgroup estimate. The point estimate being close to 1 does not by itself establish equivalence, and a confidence interval that includes 1 does not prove that there is no treatment effect. It indicates that the registry-reported analysis does not precisely distinguish among a range of possible hazard ratios represented by that interval.

The same principle applies to the female age analyses. An HR of 1.08 among females aged 25 or younger and an HR of 0.92 among females older than 25 are descriptive subgroup estimates. Without a formal interaction analysis reported in the ClinicalTrials.gov record, they should not be treated as evidence that age modifies the treatment effect.

13. Safety Results

The ClinicalTrials.gov record identifies serious adverse events by randomized arm. It reports 103/2704 affected/at risk in the vaccine group and 95/2700 affected/at risk in the placebo group.

Safety measureVaccinePlacebo
Serious adverse events 103/2704 95/2700
Clinical Biostats interpretation

The reported safety figures are participant counts expressed as affected/at risk for each arm. They should be distinguished from the HIV-1 infection hazard ratios because they address a different clinical outcome and use a different data structure.

The ClinicalTrials.gov record does not provide a formal statistical comparison, confidence interval, or p-value for the serious-adverse-event counts. Accordingly, this page does not manufacture a comparative test from the reported counts.

The registry also lists primary safety endpoints covering local reactogenicity, systemic reactogenicity, adverse events by relationship and severity, serious adverse events, adverse events of special interest, new chronic medical conditions, early study termination associated with an AE or reactogenicity, and study product discontinuation associated with an AE or reactogenicity. The registry-reported statistical-analyses dataset does not provide formal comparative estimates for those endpoints.

14. Statistical Methods Explained

Why was a sex-stratified analysis used?

The registry analysis notes explicitly identify a sex-stratified log-rank test and a sex-stratified Cox proportional-hazards model. Stratification allows the analysis to account for different baseline hazards across the specified strata while estimating the vaccine-versus-placebo comparison within the survival-analysis framework.

What does an HR of 1.02 mean?

Because the vaccine group is the numerator and placebo is the denominator, an HR of 1.02 means the estimated instantaneous hazard in the vaccine group was 1.02 times that in the placebo group under the reported model. It is not an absolute risk difference and is not the percentage of participants who experienced infection.

Why is the confidence interval important?

A point estimate such as 1.02 is only one estimate from the available data. The 95% CI of 0.81–1.30 communicates uncertainty around that estimate. The interval should be read as a range generated by the statistical estimation procedure, not as the range of effects experienced by individual participants.

Why does the p-value not measure effect size?

The p-value describes the statistical evidence under the specified null-testing framework. It depends on the observed data and the statistical procedure, but it is not a scale of clinical magnitude. The HR and its confidence interval provide the effect estimate and its precision; the p-value addresses the hypothesis test.

Why use a log-rank test for HIV-1 infection?

HIV-1 infection is recorded as a time-to-event outcome. A log-rank test is designed to compare event-time distributions between groups while incorporating the timing of events and the information contributed by participants who are censored before experiencing the event.

What is the role of the Nelson-Aalen estimate?

The registry-reported Wald analyses use Nelson-Aalen cumulative hazard estimates. The Nelson-Aalen estimator accumulates observed event information over time to estimate cumulative hazard. It therefore provides a different statistical route from the sex-stratified log-rank/Cox analysis, even though both are being used to characterize the vaccine-versus-placebo comparison.

Why should subgroup HRs not be compared by eye?

Separate estimates can differ simply because of sampling variability. An HR of 1.08 in one subgroup and 0.92 in another does not, by itself, establish effect modification. A formal interaction test is generally required to test whether treatment effects differ across subgroups, and no such interaction result is reported in the ClinicalTrials.gov record.

15. Understanding the Hazard Ratio

Reference point
HR = 1  →  equal estimated hazards

Values above 1 correspond to a higher estimated instantaneous hazard in the vaccine group; values below 1 correspond to a lower estimated instantaneous hazard in the vaccine group.

Reported HRStatistical reading
1.02Estimated vaccine-group hazard is 1.02 times the placebo-group hazard.
1.03Estimated vaccine-group hazard is 1.03 times the placebo-group hazard.
1.05Estimated vaccine-group hazard is 1.05 times the placebo-group hazard.
1.00Estimated hazards are equal under the reported analysis.
1.15Estimated vaccine-group hazard is 1.15 times the placebo-group hazard.
1.12Estimated vaccine-group hazard is 1.12 times the placebo-group hazard.
0.99Estimated vaccine-group hazard is 0.99 times the placebo-group hazard.
1.08Estimated vaccine-group hazard is 1.08 times the placebo-group hazard.
0.92Estimated vaccine-group hazard is 0.92 times the placebo-group hazard.

This table is a translation of the reported effect measures into their statistical meaning. It does not convert the hazard ratios into absolute infection probabilities because the ClinicalTrials.gov record does not provide the necessary event and person-time quantities for that calculation.

16. Confidence Intervals and Precision

The confidence intervals provide an important second dimension of interpretation beyond the point estimates. The primary 24-month Cox estimate is 1.02 with a 95% CI of 0.81–1.30. The corresponding Wald estimate is 1.03 with a 95% CI of 0.81–1.31.

At 36 months, the Cox estimate is 1.05 with a 95% CI of 0.85–1.28, while the Wald estimate is 1.00 with a 95% CI of 0.81–1.23.

Point estimate

The hazard ratio is the single numerical estimate produced by the specified analysis.

Confidence interval

The interval communicates uncertainty around that estimate under the statistical framework used.

Reference value

For a hazard ratio, 1 represents equal estimated hazards between the two groups.

P-value

The p-value addresses the corresponding hypothesis test; it is not a measure of effect size or clinical importance.

17. Randomization, Masking, and Causal Interpretation

Randomization is a central design feature because treatment assignment is randomized rather than chosen according to participant characteristics. This creates the foundation for comparing the subsequent infection experience between the vaccine and placebo groups while reducing systematic confounding from measured and unmeasured baseline factors, subject to the usual assumptions of randomized-trial inference.

The trial is described as quadruple-masked. Masking can reduce the influence of knowledge of treatment assignment on participant behavior, clinical assessment, reporting, and other aspects of trial conduct. Its relevance is particularly clear in a prevention study where infection ascertainment and adverse-event reporting occur over time.

Design versus analysis: randomization does not eliminate the need for appropriate time-to-event methods. Participants can contribute different amounts of follow-up, and infection can occur at different times. The Cox and log-rank framework therefore complements the randomized design rather than replacing it.

18. Analysis Populations

The statistical analyses posted on ClinicalTrials.gov identify the MITT cohort for the primary infection analyses. Secondary analyses use the Week 26 at-risk cohort, females in the MITT cohort, males in the MITT cohort, and specified female age subgroups.

Analysis populationUse in the ClinicalTrials.gov record
MITT cohortPrimary HIV-1 infection analyses through 24 and 36 months; female and male subgroup analyses.
Week 26 at-risk cohortSecondary HIV-1 infection analysis after Month 6.5 through 24 months.
Females in MITT cohortSecondary HIV-1 infection analysis through 24 months.
Males in MITT cohortSecondary HIV-1 infection analysis through 24 months.
Females in MITT cohort aged 25 or youngerSecondary HIV-1 infection analysis through 24 months.
Females in MITT cohort older than 25Secondary HIV-1 infection analysis through 24 months.

The distinction between these populations matters. An estimate from the Week 26 at-risk cohort is not directly interchangeable with the estimate from the full MITT cohort, because the populations entering the analyses are defined differently.

19. Multiplicity and Multiple Analyses

The ClinicalTrials.gov record identifies 12 registered primary endpoints and reports 10 statistical analyses in total, including four primary-endpoint analyses and six secondary analyses.

The four primary statistical analyses comprise two methods for the 24-month HIV-1 infection endpoint and two methods for the 36-month HIV-1 infection endpoint. The ClinicalTrials.gov record does not provide a formal multiplicity-adjustment scheme or an alpha-allocation procedure for these reported analyses.

Interpretation caution: the presence of multiple registered endpoints and multiple reported analyses means individual p-values should be interpreted in the context of the overall statistical design. The ClinicalTrials.gov record does not provide enough information to reconstruct a formal familywise-error procedure, so this page does not assign one or perform an unreported multiplicity adjustment.

20. What the Results Do and Do Not Establish

What the HR describes

A relative comparison of the instantaneous event hazard between vaccine and placebo under the specified time-to-event model.

What the HR does not describe

It is not an absolute infection probability, an absolute risk difference, or the proportion of participants who benefit.

What the CI describes

Uncertainty around the estimated hazard ratio under the stated statistical framework.

What the p-value describes

Evidence under the specified hypothesis-testing procedure, not the magnitude or clinical importance of the treatment effect.

The primary hazard-ratio estimates are close to 1: 1.02 and 1.03 through 24 months, and 1.05 and 1.00 through 36 months. The corresponding confidence intervals all include 1. The registry-reported p-values are 0.84, 0.83, 0.66, and 0.98.

Those statements describe the reported statistical results. They should not be expanded into claims about absolute vaccine effectiveness beyond the quantities reported in the ClinicalTrials.gov record. In particular, the ClinicalTrials.gov record does not provide the cumulative infection rates needed to construct an independent absolute-risk comparison here.

21. Missing Data, Censoring, and Proportional Hazards

Time-to-event analyses generally distinguish participants who experience the event from participants who are censored before an event is observed. The ClinicalTrials.gov record identifies the infection outcomes as time-to-event endpoints and identifies Cox proportional-hazards and log-rank methods, but it does not provide a detailed missing-data or censoring specification.

Accordingly, this page does not infer a particular imputation method or censoring rule beyond what is explicitly present in the ClinicalTrials.gov record.

Why censoring matters
Observed follow-up time ≠ event time for every participant

A survival analysis can use the information contributed by participants up to their censoring time. This is one reason a time-to-event analysis is preferable to simply ignoring participants who do not experience HIV-1 infection during the observed period.

The Cox proportional-hazards model also makes a structural assumption about the relative hazard over time. The ClinicalTrials.gov record does not report a formal proportional-hazards diagnostic, so the HR should be understood as the model-based summary reported by the registry rather than as proof that the proportional-hazards assumption was perfectly satisfied.

22. Limitations

23. Why This Trial Matters Statistically

HVTN 702 is a useful teaching case because it combines randomized prevention-trial design with a time-to-event efficacy endpoint, sex-stratified survival analysis, multiple effect-estimation approaches, subgroup analyses, and a separate safety framework.

ConceptHow it appears in HVTN 702
RandomizationThe trial uses randomized allocation in a parallel two-arm design.
BlindingThe registry describes the study as quadruple-masked.
Time-to-event endpointHIV-1 infection is analyzed according to time from enrollment/first vaccination to diagnosis.
Hazard ratioThe primary efficacy effect measure is the hazard ratio, with vaccine as numerator and placebo as denominator.
Log-rank testPrimary and secondary infection analyses include sex-stratified log-rank tests.
Cox modelThe primary HR estimates are obtained using sex-stratified Cox proportional-hazards models.
Nelson-AalenThe Wald analyses use Nelson-Aalen cumulative hazard estimates.
Wald / z-testWald analyses provide a second reported inferential approach for the primary infection endpoints.
Confidence intervalsEach reported primary HR has a two-sided 95% confidence interval.
Subgroup analysisSecondary analyses examine female and male participants and female age groups.
Safety analysisSerious adverse events are reported separately by randomized arm.
Endpoint evolutionThe registry definition states that the later 36-month HIV-1 infection endpoint superseded the earlier primary outcome after the 23 January 2020 DSMB finding.

24. Statistical Concepts in This Trial

Learn more about the methods used in this trial:

25. Related Statistical Calculators

26. Sources

Continue through Clinical Biostats

Explore the statistical methods behind randomized clinical trials, survival analysis, confidence intervals, and hypothesis testing.

27. Record Summary

HVTN 702 provides a detailed example of statistical analysis for a randomized HIV vaccine prevention study. The ClinicalTrials.gov record combines a randomized, parallel, quadruple-masked design with time-to-event analysis of HIV-1 infection. The primary efficacy analyses use sex-stratified log-rank testing and sex-stratified Cox proportional-hazards models, alongside separately reported Wald analyses using Nelson-Aalen cumulative hazard estimates.

The four reported primary analyses produce hazard ratios of 1.02, 1.03, 1.05, and 1.00, with corresponding two-sided 95% confidence intervals of 0.81–1.30, 0.81–1.31, 0.85–1.28, and 0.81–1.23. The associated p-values are 0.84, 0.83, 0.66, and 0.98.

The most important statistical lesson is that these numbers should be interpreted together rather than individually. A hazard ratio supplies a relative time-to-event estimate; its confidence interval communicates precision; the p-value describes evidence under the specified test; and the analysis population and modeling assumptions determine what the estimate represents. Secondary subgroup estimates require additional caution because a difference between subgroup estimates is not itself an interaction test.

Clinical Biostats methodology: A trial-results page should distinguish registry-reported evidence from statistical interpretation. For HVTN 702, that means preserving the reported hazard ratios, confidence intervals, p-values, analysis populations, and methods while avoiding unsupported reconstruction of event rates, survival curves, subgroup interactions, or unreported multiplicity procedures.