This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record.
1. Trial at a Glance
CAPRISA 004 was a completed phase 2 randomized study evaluating the safety and effectiveness of a candidate vaginal microbicide for prevention of HIV infection. The registry describes two study arms, double masking, a prevention purpose, and enrollment of 889 participants.
| Feature | CAPRISA 004 |
|---|---|
| Study | CAPRISA 004 |
| NCT identifier | NCT00441298 |
| Official title | Safety and Effectiveness Study of a Candidate Vaginal Microbicide for Prevention of HIV |
| Phase | Phase 2 |
| Status | COMPLETED |
| Condition | HIV Infections |
| Allocation | RANDOMIZED |
| Design model | SINGLE_GROUP |
| Masking | DOUBLE |
| Primary purpose | PREVENTION |
| Enrollment | 889 |
| Number of arms | 2 |
| Lead sponsor | Centre for the AIDS Programme of Research in South Africa |
| Sponsor type | NETWORK |
2. Clinical Question
The central statistical question is whether the incidence of HIV infection differs between participants assigned to tenofovir gel and those assigned to placebo during the study period.
Population
Participants enrolled in the phase 2 study for prevention of HIV infection. The registry lists the condition as HIV Infections and the primary purpose as PREVENTION.
Intervention
Tenofovir gel, identified in the registry as a drug intervention.
Comparator
Placebo, specified as Universal HEC placebo and identified in the registry as a drug intervention.
Primary question
Does the incidence of HIV infection differ between the tenofovir and placebo arms?
3. Trial Design
Tenofovir gel
- Intervention type: drug
- Candidate vaginal microbicide intervention
Placebo
- Intervention type: drug
- Universal HEC placebo
4. Primary Endpoint
| Endpoint | Registry definition | Time frame |
|---|---|---|
| Change in HIV status compared between arms (tenofovir vs placebo) | The effectiveness of tenofovir against HIV infection will be measured by comparing the incidence of HIV in the tenofovir arm with that in the placebo arm. | Baseline and monthly HIV testing for the duration of the study, an expected average of 18 months |
The endpoint is fundamentally an incidence comparison. Participants begin with HIV-status assessment at baseline and undergo monthly HIV testing during the study. The statistical target is the difference in occurrence of HIV infection between the randomized study arms over the specified follow-up period.
The registry defines effectiveness through this between-arm incidence comparison rather than through a continuous laboratory measurement or a tumor-response endpoint.
5. Statistical Methodology
Incidence as the primary statistical target
For an endpoint defined by new HIV infection during follow-up, the basic statistical comparison concerns how frequently new infections occur in each randomized group. Because participants are tested repeatedly over time, the analysis has a longitudinal time component even though the clinical event itself is binary: a participant either develops HIV infection during follow-up or does not.
A simple analysis can summarize the number of incident infections and the amount of follow-up contributed by each group. Depending on the prespecified statistical analysis plan and the available event-time information, incidence can be expressed as an incidence proportion or an incidence rate. A rate-based comparison is particularly useful when participants contribute different amounts of observation time.
Time-to-event analysis
When the timing of HIV infection and the timing of censoring are available, a time-to-event framework can use more of the information collected during monthly testing. Kaplan-Meier estimation can describe the probability of remaining free of HIV infection over time, while a log-rank comparison can test differences between randomized groups. A Cox proportional-hazards model can estimate a hazard ratio if the proportional-hazards assumption is appropriate.
Here, T represents time to HIV infection. The survival function is interpreted as the probability of remaining free of the event through time t.
Why randomization matters
Randomized allocation provides the foundation for a between-arm causal comparison. If the groups are analyzed according to their randomized assignments, differences in HIV incidence can be interpreted within the framework created by the trial's randomization rather than as an observational comparison between self-selected treatment groups.
Repeated HIV testing
Monthly testing is important statistically because the event is detected through scheduled assessments. The analysis must therefore respect the fact that the exact time of infection may be known only within the testing schedule. The appropriate event-time convention would normally be specified prospectively in the statistical analysis plan.
6. Planned Analysis
The registry says that effectiveness will be assessed by comparing the incidence of HIV infection in the tenofovir arm with the incidence in the placebo arm. The primary endpoint is assessed from baseline with monthly HIV testing for the duration of the study, with an expected average of 18 months of follow-up.
| Analysis component | Statistical interpretation |
|---|---|
| Outcome | New HIV infection during study follow-up |
| Comparison | Tenofovir gel versus placebo |
| Assessment | Baseline and monthly HIV testing |
| Expected average follow-up | 18 months |
| Primary estimand concept | Between-arm difference in HIV incidence |
| Potential time-to-event framework | Kaplan-Meier, log-rank, and Cox methods are standard approaches when event and censoring times are available |
| Posted numerical analysis | None in the ClinicalTrials.gov record |
For an endpoint of this type, the most informative report would generally include the number of HIV infections in each randomized arm, the amount of follow-up, an appropriate measure of relative or absolute incidence, and an uncertainty interval. If infection timing and censoring are available, a time-to-event estimate can additionally account for differences in follow-up duration.
The absence of a posted formal analysis means that the ClinicalTrials.gov record does not supply a numerical treatment-effect estimate from which statistical significance or precision can be assessed.
7. Statistical Methods Explained
Why is HIV incidence an appropriate primary endpoint?
The purpose of the study is prevention of HIV infection. Incidence directly measures the occurrence of new HIV infections during follow-up, making it a natural statistical outcome for comparing preventive effectiveness between the randomized groups.
Why does monthly testing matter?
Repeated testing establishes whether and when a participant becomes HIV positive during the study. More frequent assessments can narrow the interval in which a new infection occurred, which is relevant when the analysis treats infection as a time-to-event outcome.
What is the difference between incidence proportion and incidence rate?
An incidence proportion describes the fraction of participants who experience the event during a defined period. An incidence rate incorporates person-time, making it useful when participants contribute different lengths of observation. The choice should follow the prespecified analysis plan and the structure of the collected data.
Why might Kaplan-Meier methods be useful?
Kaplan-Meier estimation is designed for time-to-event data with censoring. In a prevention trial, some participants may complete follow-up without HIV infection or may stop contributing observations before the planned end of follow-up. Kaplan-Meier methods can retain information from the time each participant was observed.
What would a hazard ratio mean if a Cox model were used?
A hazard ratio would compare the estimated instantaneous rates of HIV infection between the randomized groups over follow-up. A value below 1 would indicate a lower estimated hazard in the numerator group, while a value above 1 would indicate a higher estimated hazard. It would not directly equal the percentage of participants who avoided HIV infection.
Why is the confidence interval important?
A confidence interval communicates statistical uncertainty around an estimated treatment effect. A narrow interval indicates greater precision than a wide interval, all else being equal. The interval does not describe the range of individual participant responses.
Why should a p-value not be interpreted as the size of the treatment effect?
A p-value addresses evidence against a specified null hypothesis under the statistical model. It does not quantify the magnitude or clinical importance of the treatment effect. For a prevention endpoint, an effect estimate together with its confidence interval is needed to describe both direction and precision.
8. Interpretation of the Primary Endpoint
The primary endpoint is designed to determine whether HIV infection occurs at a different incidence in the tenofovir group than in the placebo group during follow-up. A lower incidence in the tenofovir arm would correspond descriptively to fewer observed HIV infections relative to the comparator, while a statistical model would quantify the uncertainty around that comparison.
An incidence comparison alone would not describe every aspect of the intervention's safety, adherence, biological mechanism, or long-term effects. The ClinicalTrials.gov record identifies the primary effectiveness endpoint but does not provide a complete set of additional outcome results.
An absolute incidence difference answers how much the observed occurrence of HIV infection differs between groups. A relative measure, such as an incidence rate ratio or hazard ratio, describes the treatment effect on a relative scale. These quantities answer different questions and should not be treated as interchangeable.
9. Precision, Censoring, and Event Timing
The primary endpoint is assessed repeatedly over an expected average of 18 months. That structure creates several statistical considerations that are important when interpreting a prevention trial.
Censoring
Participants who remain HIV negative through their last observation may contribute follow-up without experiencing the event. A time-to-event analysis must account for that incomplete event information appropriately.
Event timing
Monthly testing means that the exact date of infection may not necessarily be directly observed. The prespecified event-time convention determines how that information enters a time-to-event analysis.
Unequal follow-up
If participants contribute different amounts of observation time, an incidence-rate analysis can incorporate person-time rather than treating every participant as if they were observed for exactly the same duration.
Statistical precision
The precision of an incidence comparison depends heavily on the number of observed HIV infections and the amount of follow-up. Enrollment alone does not determine the width of the eventual confidence interval.
10. Important Limitations and Interpretation Issues
- No posted statistical results: the ClinicalTrials.gov record contains no posted statistical analyses for the registered primary endpoint, so the record does not provide a numerical effect estimate, confidence interval, or p-value.
- Endpoint timing: HIV status was assessed at baseline and monthly, so the statistical treatment of the timing of infection would need to follow the prespecified analysis approach.
- Follow-up duration: the registry specifies an expected average of 18 months rather than a single identical follow-up duration for every participant.
- Incidence versus prevalence: the primary endpoint concerns new HIV infection during follow-up. It should not be confused with the prevalence of HIV infection at a single point in time.
- Design-model terminology: the registry records SINGLE_GROUP as the design model while separately recording randomized allocation and two arms. These fields should be distinguished rather than silently reconciled.
- Model assumptions: if a Cox proportional-hazards model were used, its interpretation would depend on the proportional-hazards assumption. A single hazard ratio can be less descriptive when relative hazards vary materially over time.
- Analysis population: the ClinicalTrials.gov record does not specify an intention-to-treat, per-protocol, or other formal analysis population for the primary endpoint.
- Multiplicity: the ClinicalTrials.gov record does not describe a multiplicity strategy for multiple endpoints or multiple statistical looks.
11. Why This Trial Matters Statistically
CAPRISA 004 provides a useful framework for understanding how randomized prevention trials translate repeated clinical testing into an incidence comparison. The central statistical challenge is straightforward to state but important to handle carefully: the outcome is whether HIV infection occurs, while the observation process unfolds over time.
| Concept | How it appears in CAPRISA 004 |
|---|---|
| Randomization | The registry identifies randomized allocation between two study arms. |
| Double masking | The registry identifies the study as double masked. |
| Incidence | The primary endpoint compares HIV incidence between tenofovir and placebo. |
| Repeated assessment | HIV testing occurs at baseline and monthly during the study. |
| Time-to-event analysis | When event and censoring times are available, survival-analysis methods can use the timing of HIV infection and follow-up. |
| Censoring | Participants can contribute information up to their last observed HIV-status assessment. |
| Absolute effect | An incidence difference describes the absolute separation between randomized groups. |
| Relative effect | An incidence rate ratio or hazard ratio can describe the treatment effect on a relative scale when appropriate. |
| Statistical uncertainty | Confidence intervals quantify uncertainty around the estimated treatment effect. |
| Prevention endpoint | The outcome directly addresses occurrence of new HIV infection rather than treatment response after infection. |
12. Statistical Interpretation vs Clinical Interpretation
Statistical interpretation
The primary endpoint is a between-arm comparison of HIV incidence assessed from baseline through monthly testing. A suitable formal analysis would quantify the difference or ratio in incidence and its uncertainty, using time-to-event methods when the event-time structure supports them.
Clinical interpretation
The clinical question is whether assignment to tenofovir gel is associated with a different occurrence of HIV infection than assignment to placebo during the study period. The registry identifies this as the study's primary effectiveness question.
13. Sources
- ClinicalTrials.gov: CAPRISA 004 — NCT00441298
Continue through the Clinical Biostats trial library
Explore additional clinical trial analyses and statistical methods used to interpret randomized evidence.
14. Record Summary
CAPRISA 004 is registered as a completed phase 2 randomized, double-masked prevention study with 889 participants and two intervention arms: tenofovir gel and Universal HEC placebo. Its primary endpoint is change in HIV status compared between arms, with baseline and monthly HIV testing for the duration of the study and an expected average of 18 months.
Statistically, the key feature is the distinction between a simple binary outcome and an outcome observed through repeated follow-up. The primary question concerns HIV incidence, while the monthly testing schedule creates a time dimension that can support time-to-event methods when the underlying event and censoring information are available. The ClinicalTrials.gov record does not post formal statistical analyses or numerical results for this endpoint, so the appropriate interpretation of the registry record is centered on the prespecified clinical question, endpoint structure, and statistical framework rather than on an estimated treatment effect.