This page separates information reported in the ClinicalTrials.gov record from statistical interpretation. The registry reports seven outcome measures and four registered primary endpoints, but the ClinicalTrials.gov record does not contain formal statistical-analysis results with effect estimates, confidence intervals, or p-values.
1. Trial at a Glance
RV144 was a randomized, parallel-group, triple-masked, phase 3 prevention trial evaluating ALVAC-HIV vCP1521 + AIDSVAX against ALVAC placebo + AIDSVAX placebo in adults enrolled in Thailand. The trial enrolled 16,402 participants and had four registered primary endpoints evaluated over 42 months.
| Feature | RV144 |
|---|---|
| Trial name | RV144 |
| NCT identifier | NCT00223080 |
| Brief title | HIV Vaccine Trial in Thai Adults |
| Therapeutic area | Infectious Disease |
| Condition | HIV Infection |
| Phase | Phase 3 |
| Status | Completed |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Triple |
| Primary purpose | Prevention |
| Enrollment | 16,402 |
| Arms | 2 |
| Lead sponsor | U.S. Army Medical Research and Development Command |
| Sponsor type | Federal |
2. Clinical Question
The central statistical question was whether the HIV vaccine regimen reduced HIV-1 infection compared with the placebo regimen over the registered 42-month assessment period. The registry also specified analyses of HIV-1 viral load among volunteers who developed HIV infection during the trial.
Population
Adults enrolled in the HIV vaccine trial in Thailand, with 16,402 participants enrolled overall.
Intervention
ALVAC-HIV vCP1521 + AIDSVAX, classified in the registry as a biological intervention.
Comparator
ALVAC placebo + AIDSVAX placebo, classified in the registry as an other intervention.
Primary question
Does the vaccine regimen change the incidence of HIV-1 infection compared with placebo, and among participants who become infected, does it affect HIV-1 viral load?
3. Trial Design
ALVAC-HIV vCP1521 + AIDSVAX
- ALVAC-HIV vCP1521
- AIDSVAX
- Biological intervention
- Primary efficacy outcomes evaluated over 42 months
ALVAC Placebo + AIDSVAX Placebo
- ALVAC placebo
- AIDSVAX placebo
- Other intervention category in the registry
- Primary efficacy outcomes evaluated over 42 months
The design is a randomized parallel-group comparison. Randomization is important because it creates the basis for comparing subsequent HIV infection outcomes between treatment assignments without deliberately assigning participants according to their expected infection risk.
4. Trial Timeline
Trial start
The registry lists October 2003 as the trial start date.
Primary completion
The registry lists July 2006 as the primary completion date.
Registry status
The ClinicalTrials.gov record identifies RV144 as completed and reports results for its registered outcome measures.
5. Endpoints
The registry identifies four primary endpoints. Two concern HIV-1 infection and two concern HIV-1 viral load among volunteers developing HIV infection during the trial.
| Registered primary endpoint | Time frame | Registry definition |
|---|---|---|
| Kaplan-Meier Estimate of HIV-1 Infection Rate in Intent to Treat Population | 42 Months | HIV-1 infection rate. Detection of HIV-1 infection was defined according to the HIV diagnostic algorithm utilizing serologic and nucleic acid technologies. Incidence of HIV infection was compared in the vaccine and placebo-recipient groups. |
| Vaccine Efficacy as Determined by Acquisition of Infection in the Per-protocol Population | 42 Months | Cumulative Number of HIV Infections. Detection of HIV-1 infection was defined according to the HIV diagnostic algorithm utilizing serologic and nucleic acid technologies. Incidence of HIV infection was compared in the vaccine and placebo-recipient groups. |
| Changes in HIV-1 Viral Load in Volunteers Developing HIV Infection During the Trial for the MITT Population | 42 months | Log10 HIV-1 viral loads for diagnostic specimens for subjects with post-HIV infection. The trial quantitated HIV plasma viral load at the time of diagnosis and through the remainder of the follow-up period. Peri infection results were compared in vaccine and placebo recipients who became HIV-infected during the trial. |
| Changes in HIV-1 Viral Load in Volunteers Developing HIV Infection During the Trial for the Per Protocol Population | 42 months | Log10 HIV-1 viral loads for diagnostic specimens for subjects with post-HIV infection. The trial quantitated HIV plasma viral load at the time of diagnosis and through the remainder of the follow-up period. Peri infection results were compared in vaccine and placebo recipients who became HIV-infected during the trial. |
6. Analysis Populations
The registry explicitly distinguishes an Intent to Treat population, a Modified Intent to Treat (MITT) population, and a Per-protocol population across the registered primary endpoints.
| Population | Role in the registered endpoints |
|---|---|
| Intent to Treat | Used for the registered Kaplan-Meier estimate of HIV-1 infection rate. |
| Per-protocol | Used for the registered vaccine-efficacy endpoint based on acquisition of infection and for one registered viral-load endpoint. |
| Modified Intent to Treat | Used for one registered viral-load endpoint among volunteers developing HIV infection during the trial. |
7. Posted Outcome Measures and Analysis Status
The ClinicalTrials.gov record states that results have been posted and that seven outcome measures are posted. However, the ClinicalTrials.gov record contains no formal statistical-analysis entries: there are no posted hazard ratios, risk ratios, vaccine-efficacy estimates, confidence intervals, p-values, or formal model results in the posted results.
What the registry does establish
- HIV-1 infection was compared between vaccine and placebo recipients.
- The first infection endpoint used a Kaplan-Meier estimate in the Intent to Treat population.
- The second infection endpoint used cumulative HIV infections in the Per-protocol population.
- Two primary endpoints evaluated log10 HIV-1 viral loads among volunteers who developed HIV infection.
- The viral-load outcomes were evaluated in MITT and Per-protocol populations.
- All four registered primary endpoints had a 42-month time frame.
What the ClinicalTrials.gov record does not establish
- A numerical vaccine-efficacy estimate.
- A numerical hazard ratio or risk ratio.
- A confidence interval for an infection comparison.
- A p-value for any primary endpoint.
- A numerical difference in viral load between randomized groups.
- A formal multiplicity-adjusted conclusion.
8. Planned Analysis
Because the ClinicalTrials.gov record contains registered endpoints but no formal statistical-analysis entries, the appropriate approach is to describe how these endpoint types would ordinarily be analyzed without attributing an unreported numerical result to RV144.
HIV-1 infection rate: Kaplan-Meier analysis
The first primary endpoint explicitly specifies a Kaplan-Meier estimate of HIV-1 infection rate in the Intent to Treat population over 42 months. A Kaplan-Meier framework is appropriate for time-to-first-infection data when participants can have different amounts of follow-up and some participants remain infection-free at the end of observation.
The Kaplan-Meier estimator updates the estimated event-free probability at each observed event time, accounting for the number at risk immediately before that time and the number of events at that time.
For an HIV-acquisition endpoint, the event is development of HIV-1 infection according to the trial's diagnostic algorithm. Participants who remain uninfected at their last evaluable follow-up would generally contribute information until censoring.
Acquisition of infection: cumulative infection comparison
The second primary endpoint is defined as vaccine efficacy as determined by acquisition of infection in the Per-protocol population and specifies the cumulative number of HIV infections. A conventional analysis of such an endpoint would compare infection risk between randomized groups and estimate a relative treatment effect with an appropriate confidence interval.
When vaccine efficacy is expressed from a risk ratio, VE is one minus the ratio of infection risk in the vaccine group to infection risk in the comparison group. This formula is presented here as a statistical concept; the registry-reported RV144 data do not provide the numerical RR or VE.
Viral-load endpoints
The third and fourth primary endpoints concern changes in log10 HIV-1 viral load among volunteers developing HIV infection during the trial. Because viral load is recorded on a logarithmic scale, an analysis would ordinarily compare the longitudinal or peri-infection measurements between vaccine and placebo recipients using a model appropriate to the repeated measurements and timing of specimens.
The exact model, covariance structure, handling of missing viral-load measurements, and hypothesis-testing procedure are not specified in the ClinicalTrials.gov record. Those details should therefore not be attributed to the RV144 analysis.
9. Statistical Methodology
Kaplan-Meier estimation
Kaplan-Meier estimation is directly named in the first registered primary endpoint. Its purpose is to estimate the time-to-infection distribution while accounting for censoring.
Here, S(t) represents the probability of remaining free of the event beyond time t. For an infection endpoint, the complementary quantity describes the estimated probability of infection by a specified time under the relevant event and censoring definitions.
Time-to-event comparison
When two randomized groups are followed for acquisition of infection, a time-to-event comparison can account for both when infection occurs and how long participants remain under observation without infection. A log-rank test is a common group-comparison procedure for Kaplan-Meier curves, while a Cox proportional-hazards model is a common way to estimate a hazard ratio.
Neither a log-rank p-value nor a Cox hazard ratio is reported in the ClinicalTrials.gov record for RV144. They therefore should be understood here as the standard analytical framework for this endpoint type rather than as reported RV144 results.
Risk and vaccine efficacy
The per-protocol endpoint is based on acquisition of infection and cumulative numbers of infections. A risk-based analysis focuses on the probability of infection over a specified period rather than the instantaneous event rate. Vaccine efficacy can then be expressed as the relative reduction in infection risk.
A vaccine-efficacy estimate is a relative measure. It does not by itself give the absolute probability that an individual participant will acquire HIV-1. Absolute infection risk in the comparison population is needed to understand the corresponding absolute risk difference.
Log10 viral load
The registry specifies log10 HIV-1 viral loads. A logarithmic transformation is commonly used for viral-load measurements because biological concentrations can span several orders of magnitude and logarithmic values can make distributions more suitable for statistical modeling.
A change on the log10 scale represents a multiplicative change on the original viral-load scale. The ClinicalTrials.gov record does not provide the numerical viral-load changes or between-group estimates needed to quantify the effect in RV144.
10. Blinding and Its Statistical Role
RV144 is recorded as triple-masked. Blinding is a design feature intended to reduce the possibility that knowledge of treatment assignment influences trial conduct, outcome assessment, participant behavior, or other processes that could affect observed results.
Why masking matters
When participants and relevant trial personnel do not know treatment assignment, expectations about the intervention are less able to influence trial procedures and observations.
What masking does not do
Masking does not replace randomization or statistical adjustment. It is a design safeguard that operates alongside the randomized comparison.
The ClinicalTrials.gov record identifies the trial as triple-masked but does not specify the identities of all three masked groups. This page therefore does not assign particular masking roles beyond the registry's stated designation.
11. Randomization and the Parallel-Group Design
The trial used randomized allocation and a parallel design. Each participant is therefore understood within the statistical comparison defined by their randomized arm rather than being used as their own control through crossover.
| Design feature | Statistical implication |
|---|---|
| Randomized | Creates the basis for an unbiased comparison of treatment assignments, subject to implementation and follow-up. |
| Parallel | Participants remain associated with their randomized study arm rather than moving through sequential treatment conditions as a crossover design. |
| Triple-masked | Reduces opportunities for treatment knowledge to influence trial conduct or assessment. |
| Prevention purpose | The central efficacy event is acquisition of HIV-1 infection rather than treatment response after established infection. |
12. Safety Results
The ClinicalTrials.gov record reports serious adverse events by randomized arm as affected participants divided by participants at risk.
Serious adverse events
Vaccine arm vs placebo arm
| Safety measure | Vaccine | Placebo |
|---|---|---|
| Serious adverse events, affected / at risk | 1,175 / 8,197 | 1,219 / 8,198 |
The ClinicalTrials.gov record does not provide a statistical-analysis entry for serious adverse events. Consequently, this page reports the affected and at-risk counts exactly as reported in the registry rather than calculating an unreported risk ratio, confidence interval, or p-value.
The two arm-specific fractions describe the number of participants affected by serious adverse events relative to the number at risk. A formal comparative safety analysis would require a prespecified definition of the safety population and endpoint, followed by an appropriate comparison of event proportions or event rates. The ClinicalTrials.gov record does not specify such an analysis.
13. Primary Endpoint-by-Endpoint Statistical Interpretation
Kaplan-Meier Estimate of HIV-1 Infection Rate
The first endpoint asks for a time-to-infection estimate in the Intent to Treat population. The statistical advantage of Kaplan-Meier estimation is that it uses the observed timing of infection rather than treating every participant as though they had exactly the same follow-up duration.
The registry's endpoint is designed to describe HIV-1 infection over 42 months while accounting for the timing of infection and censoring. The ClinicalTrials.gov record does not contain the resulting Kaplan-Meier estimate, a treatment-effect estimate, a confidence interval, or a p-value. Therefore no numerical vaccine effect can be inferred from the endpoint name alone. If a hazard ratio were reported in a complete statistical analysis, its interpretation would depend on the proportional-hazards model and its assumptions; a hazard ratio would not be equivalent to a percentage of participants protected.
Vaccine Efficacy by Acquisition of Infection
The second endpoint moves from a time-to-event description to cumulative HIV infections in the Per-protocol population. This distinction is important because the analysis population changes from ITT to per-protocol.
A vaccine-efficacy estimate would quantify the relative reduction in infection risk associated with the vaccine regimen under the prespecified population and follow-up framework. It would not mean that the same percentage of every individual participant's personal infection probability was reduced. The confidence interval would describe statistical uncertainty around the estimated relative effect, while a p-value would address evidence against a specified null hypothesis rather than measure the size of the effect. No numerical VE, confidence interval, or p-value is reported here.
Changes in HIV-1 Viral Load: MITT Population
This endpoint evaluates log10 HIV-1 viral loads among volunteers who developed HIV infection during the trial. The registry states that viral load was quantitated at diagnosis and through the remainder of follow-up and compared between vaccine and placebo recipients who became infected.
This endpoint addresses a different question from infection prevention: among participants who became HIV-infected, does the randomized vaccine assignment correspond to a difference in viral-load measurements? It therefore should not be interpreted as another measure of vaccine efficacy against acquisition. The ClinicalTrials.gov record contains no numerical viral-load estimates, confidence intervals, or p-values.
Changes in HIV-1 Viral Load: Per-protocol Population
The fourth primary endpoint uses the same viral-load concept but evaluates it in the Per-protocol population.
The difference between this endpoint and the MITT viral-load endpoint is primarily the analysis population. Comparing results across the two populations can be informative when both are reported, but the ClinicalTrials.gov record does not provide numerical estimates with which to make such a comparison.
14. Statistical Methods Explained
Why was a Kaplan-Meier method specified?
HIV acquisition is a time-to-event outcome. Participants may become infected at different times, while others may remain infection-free when their follow-up ends. Kaplan-Meier estimation allows those different observation times to contribute information without requiring every participant to have an observed infection event.
What does the Intent to Treat population contribute?
The ITT framework preserves randomized assignment as the basis of the primary comparison. This is especially important in a prevention trial because excluding participants after randomization can introduce selection related to post-randomization behavior or follow-up.
Why is there also a Per-protocol endpoint?
A per-protocol analysis addresses the treatment comparison within a protocol-defined population. It can provide a complementary perspective, but because eligibility for the analysis can depend on post-randomization events or adherence-related criteria, it does not carry exactly the same causal protection from randomization as an ITT analysis.
Why does the viral-load endpoint use log10 measurements?
Viral-load values can vary over large ranges. The log10 transformation expresses these measurements on a scale where multiplicative differences become additive differences, which can make statistical modeling more practical and clinically interpretable.
What would a hazard ratio mean for HIV acquisition?
If a Cox model were used, the hazard ratio would compare the estimated instantaneous rates of HIV acquisition between randomized groups. An HR below 1 would indicate a lower estimated instantaneous acquisition rate in the vaccine group. It would not mean that the percentage of participants infected was reduced by exactly the same percentage.
Why should a p-value not be treated as the effect size?
A p-value addresses compatibility of the observed data with a specified null hypothesis under a statistical model. It depends on both the magnitude of the observed difference and the amount of information available. An effect estimate and its confidence interval are therefore necessary to understand the size and precision of a treatment effect.
15. Censoring and Follow-up in Infection Endpoints
Time-to-infection analysis depends on how participants are followed. A participant who has not developed HIV-1 infection by the end of observable follow-up does not necessarily provide evidence that infection could never occur; instead, the participant contributes infection-free follow-up until the relevant censoring time.
Censoring is not infection
A censored participant has not experienced the event by the last relevant observation. Treating every censored participant as if the event never could occur would generally misrepresent a time-to-event process.
Timing matters
Two participants with the same eventual status can contribute different amounts of information if they have different lengths of observable follow-up.
The ClinicalTrials.gov record does not specify the detailed censoring rules, loss-to-follow-up procedures, or missing-data methods used for the infection analysis. Those elements should therefore not be inferred.
16. Multiplicity and the Four Primary Endpoints
RV144 has four registered primary endpoints: two infection endpoints and two viral-load endpoints. Multiple primary endpoints raise an important statistical-design question: how is the overall type I error controlled when several hypotheses are evaluated?
| Primary endpoint | Question | Population |
|---|---|---|
| HIV-1 infection rate | Time-to-infection comparison | Intent to Treat |
| Vaccine efficacy by acquisition | Cumulative HIV infections | Per-protocol |
| Viral-load change | Log10 viral load among infected volunteers | MITT |
| Viral-load change | Log10 viral load among infected volunteers | Per-protocol |
The distinction is important because four primary endpoints can generate several opportunities for a statistically unusual result. A complete statistical analysis would need to identify the prespecified family of hypotheses and the error-control procedure before interpreting p-values as confirmatory evidence.
17. Non-Inferiority, Crossover, and Bayesian Methods
| Design topic | Supported by reported RV144 data? | Interpretation |
|---|---|---|
| Non-inferiority margin | No | No non-inferiority margin is reported, and the registered endpoints do not identify a non-inferiority framework. |
| Crossover | No | The registry-reported design is parallel; no crossover procedure is reported. |
| Factorial design | No | The design model is reported as parallel rather than factorial. |
| Interim analysis | Not reported | No interim-analysis procedure is reported in the ClinicalTrials.gov record. |
| Missing-data / imputation method | Not reported | No specific imputation method is reported. |
| Stratification | Not reported | The ClinicalTrials.gov record does not identify randomization strata. |
| Bayesian methods | Not reported | No Bayesian model or posterior analysis is included in the ClinicalTrials.gov record. |
These topics are deliberately separated from the registered design rather than filled with assumptions. A statistical-analysis page should distinguish what the registry documents from methods that could merely be plausible for a trial of this type.
18. Safety and Efficacy Are Different Statistical Questions
The serious-adverse-event counts and HIV-infection endpoints answer different questions. Infection endpoints concern the preventive efficacy of the randomized intervention, while serious adverse events concern safety outcomes associated with participation and exposure.
Efficacy
The primary efficacy endpoints focus on HIV-1 acquisition and, among infected participants, changes in HIV-1 viral load.
Safety
The registry-reported safety measure is the number of participants affected by serious adverse events relative to the number at risk in each arm.
A statistically favorable efficacy result would not by itself establish an overall benefit-risk conclusion. Conversely, a safety comparison cannot be substituted for the primary infection endpoint. Each endpoint should be interpreted according to its prespecified clinical and statistical role.
19. Important Limitations and Interpretation Issues
- No formal statistical-analysis entries reported: the ClinicalTrials.gov record contain no numerical treatment-effect estimates, confidence intervals, or p-values, even though results are reported as posted.
- Multiple primary endpoints: four primary endpoints are registered, but the ClinicalTrials.gov record does not specify how multiplicity was controlled.
- Different analysis populations: the infection endpoints use ITT and Per-protocol populations, while viral-load endpoints use MITT and Per-protocol populations.
- Per-protocol interpretation: because the per-protocol population is defined after randomization, its treatment comparison can be more vulnerable to selection than an ITT analysis.
- Incomplete methodological detail: the ClinicalTrials.gov record does not specify the censoring rules, missing-data strategy, randomization strata, interim-analysis plan, or exact statistical model for viral-load changes.
- No numerical viral-load results: the endpoint definitions establish that log10 viral load was measured, but they do not provide the magnitude or precision of any between-group difference.
- Safety analysis detail: serious-adverse-event counts are reported, but no formal comparative safety analysis is included.
- No non-inferiority framework: there is no registry-reported non-inferiority margin, so non-inferiority logic should not be applied to these endpoints.
20. Why This Trial Matters Statistically
RV144 is a useful statistical teaching case because its registered endpoints illustrate several fundamental issues in prevention-trial analysis: randomized allocation, time-to-infection outcomes, cumulative infection risk, multiple analysis populations, censoring, and log-transformed longitudinal biological measurements.
| Concept | How it appears in RV144 |
|---|---|
| Randomization | Participants were assigned in a randomized parallel-group design. |
| Blinding | The registry identifies the trial as triple-masked. |
| Kaplan-Meier estimation | Explicitly specified for the HIV-1 infection-rate endpoint in the ITT population. |
| Time-to-event analysis | HIV-1 infection is evaluated over a 42-month time frame. |
| Vaccine efficacy | Registered as an acquisition-of-infection endpoint in the Per-protocol population. |
| Analysis populations | ITT, MITT, and Per-protocol populations are explicitly represented. |
| Log transformation | Viral-load endpoints use log10 HIV-1 viral loads. |
| Repeated measurements | Viral load was quantitated at diagnosis and through the remainder of follow-up. |
| Multiplicity | Four primary endpoints are registered, requiring careful interpretation of multiple hypotheses. |
| Safety analysis | Serious adverse events are reported by randomized arm as affected participants over participants at risk. |
21. Related Tutorials
Learn more about the methods used in this trial:
22. Related Statistical Calculators
23. Sources
- ClinicalTrials.gov: RV144, NCT00223080.
- PubMed: PMID 40190112.
- PubMed: PMID 37256916.
- PubMed: PMID 31999736.
- PubMed: PMID 29702672.
- PubMed: PMID 28131393.
Continue through Clinical Biostats
Explore statistical tutorials, clinical-trial methods, and analysis tools that build on the concepts illustrated by RV144.
24. Record Summary
RV144 provides a compact example of how the statistical structure of a prevention trial determines the questions that can be answered. The randomized, triple-masked, parallel-group phase 3 design enrolled 16,402 participants and registered four primary endpoints over 42 months. The first endpoint explicitly uses Kaplan-Meier estimation for HIV-1 infection in the Intent to Treat population; another evaluates cumulative HIV infections in the Per-protocol population; and two evaluate log10 HIV-1 viral loads among participants who develop HIV infection. The ClinicalTrials.gov record also reports serious adverse events by arm.
The key statistical lesson is that these endpoints are not interchangeable. HIV acquisition is a prevention endpoint, time-to-infection analysis accounts for follow-up and censoring, vaccine efficacy is a relative measure of infection risk, and post-infection viral-load analysis addresses a different biological question. The use of ITT, MITT, and Per-protocol populations also means that the estimands associated with the four primary endpoints are not identical.