This page separates information reported in the ClinicalTrials.gov record from statistical interpretation and educational explanation. The ClinicalTrials.gov record contains two posted primary outcome measures but no formal statistical-analysis records.
1. Trial at a Glance
HPTN 084 is a phase 3 randomized, parallel-group trial evaluating the safety and efficacy of long-acting injectable cabotegravir compared with daily oral TDF/FTC for pre-exposure prophylaxis in HIV-uninfected women. The ClinicalTrials.gov record reports an enrollment of 3224 participants, two arms, quadruple masking, and a primary purpose of prevention.
| Feature | HPTN 084 |
|---|---|
| Trial name | HPTN 084 |
| Phase | Phase 3 |
| Condition | HIV Infections |
| Population | HIV-uninfected women |
| Primary purpose | Prevention |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Quadruple |
| Enrollment | 3224 |
| Arms | 2 |
| Trial status | Active, not recruiting |
| Start | 2017-11-07 |
| Primary completion | 2020-11-05 |
| Lead sponsor | National Institute of Allergy and Infectious Diseases (NIAID) |
| Sponsor type | NIH |
| ClinicalTrials.gov | NCT03164564 |
2. Clinical Question
The central clinical question is whether long-acting injectable cabotegravir can provide an effective and safe pre-exposure prophylaxis strategy compared with daily oral TDF/FTC in HIV-uninfected women.
Population
HIV-uninfected women being evaluated in the prevention setting.
Intervention
Long-acting injectable cabotegravir, with the registry also listing oral cabotegravir and placebo for oral cabotegravir among the study interventions.
Comparator
Daily oral TDF/FTC, with placebo components listed in the registry to support the masked design.
Primary question
How do the randomized strategies compare with respect to incident HIV infection and grade 2 or higher clinical and laboratory adverse events?
3. Trial Design
The registry describes HPTN 084 as a randomized, parallel-group phase 3 prevention trial with quadruple masking. The design therefore creates two key statistical features: randomization provides the principal basis for comparing the intervention strategies, while masking is intended to reduce the influence of treatment knowledge on participant behavior, clinical assessment, laboratory assessment, and other trial processes.
CAB-based prevention strategy
- CAB LA
- Oral CAB
- Placebo for oral TDF/FTC
- Placebo for CAB LA
TDF/FTC-based prevention strategy
- Oral TDF/FTC
- Placebo for oral CAB
- Placebo for CAB LA
- Placebo components listed in the registry
The intervention listing in the registry is more detailed than a simple two-treatment comparison because it includes active and placebo components used within the masked study structure. The ClinicalTrials.gov record does not provide dosing schedules, administration intervals, or arm-specific enrollment counts beyond the denominators reported for serious adverse events.
4. Trial Timeline
Trial start
The registry lists November 7, 2017 as the study start date.
Primary completion
The registry lists November 5, 2020 as the primary completion date.
Active, not recruiting
The ClinicalTrials.gov record classifies the study as ACTIVE_NOT_RECRUITING.
5. Endpoints
The ClinicalTrials.gov record lists two primary endpoints. Both are binary outcome measures, and results are posted for both. However, the registry-reported statistical-analysis field contains no formal analysis records, so the ClinicalTrials.gov record does not support reporting an effect estimate, confidence interval, or p-value for either endpoint.
| Primary endpoint | Registry definition / time frame | Type | Results posted |
|---|---|---|---|
| Number of Pariicipants With Documented Incident HIV Infections | HIV tests at enrollment, weeks 2, 4, and 5, then every 4 weeks through week 25, then every 8 weeks. Analyzed through week 185 or the date of DSMB decision to unblind all participants, whichever is earliest. | Binary | Yes |
| Number of Participants Who Experienced Grade 2 or Higher Clinical and Laboratory Adverse Events (AEs) in Steps 1 and 2 | Treatment emergent AE measured with onset date through participant's last study visit, or the date of DSMB decision to unblind all participants, whichever is earliest. Assessed at each visit (injections visits q 8 weeks and safety visits q 8 weeks) | Binary | Yes |
Primary endpoint 1: incident HIV infection
The first primary endpoint is the number of participants with documented incident HIV infections. The registry definition specifies that an incident HIV infection must be confirmed by the HPTN Laboratory Center and Endpoint Adjudication Committee.
The measurement schedule is unusually important statistically because incident infection is not simply a single end-of-study yes/no question. Participants are repeatedly tested over follow-up. The registry specifies HIV testing at enrollment, weeks 2, 4, and 5, then every 4 weeks through week 25, followed by testing every 8 weeks. The registry-reported time-frame wording ends with "Analyzed through week 185 or the date of DSMB decision to unblind all participants, whichever is earliest.", so no additional completion point is reported here.
Primary endpoint 2: grade 2 or higher clinical and laboratory adverse events
The second primary endpoint is the number of participants who experienced grade 2 or higher clinical and laboratory adverse events in Steps 1 and 2. The registry states that treatment-emergent adverse events are measured from the onset date through the participant's last study visit, or the date of the DSMB decision to unblind all participants, whichever is earliest. Assessed at each visit (injections visits q 8 weeks and safety visits q 8 weeks).
The registry also states that adverse events will be summarized using MedDRA System Organ Class and preferred terms. This creates a different statistical structure from the incident-HIV endpoint: HIV infection is a specific binary clinical outcome, whereas adverse-event analysis involves categorizing potentially numerous events by standardized medical terminology.
6. Planned Analysis
The ClinicalTrials.gov record contains posted outcome measures, but no formal statistical analyses were posted. Therefore, there are no formal treatment-effect estimates, confidence intervals, or p-values in the ClinicalTrials.gov record that can responsibly be attributed to the registry's statistical-analysis records.
How incident HIV infection would typically be analyzed
For a randomized prevention trial in which incident HIV infection is recorded as a binary endpoint, a natural first comparison is the proportion of participants who experience an incident infection in each randomized group. Depending on the prespecified analysis plan, the treatment effect could be expressed as a risk ratio, risk difference, odds ratio, incidence rate ratio, or a time-to-event measure such as a hazard ratio if the timing of infection and censoring are incorporated into the primary analysis.
The correct estimand depends on the protocol and statistical analysis plan. The endpoint description alone does not establish which of these measures was the prespecified primary estimator for HPTN 084.
A risk difference compares absolute event probabilities, while a risk ratio compares them multiplicatively. These are different estimands and should not be treated as interchangeable.
How grade 2 or higher adverse events would typically be analyzed
For a binary safety endpoint defined as experiencing at least one grade 2 or higher clinical or laboratory adverse event, a participant-level analysis could compare the proportion of participants meeting the endpoint in each randomized group. The registry's additional instruction to summarize events using MedDRA System Organ Class and preferred terms means that an overall participant-level comparison and a descriptive breakdown of event categories answer different questions.
For safety, it is especially important to distinguish participants with at least one event from the total number of events. A single participant can experience more than one adverse event, so event counts cannot automatically be substituted for participant counts.
Why the primary analysis cannot be reconstructed from the endpoint name alone
A statistical endpoint definition identifies what is measured, but not necessarily the complete estimand or analysis model. For example, the incident-HIV endpoint could be analyzed using cumulative incidence, incidence rates, or a time-to-event approach. Likewise, an adverse-event endpoint could be summarized as a participant-level binary outcome while individual event categories are presented descriptively.
Without a formal statistical-analysis record in the ClinicalTrials.gov record, choosing one method and presenting its hypothetical result as the trial's official analysis would create information that is not present in the source dataset.
7. Reported Serious Adverse Events
the ClinicalTrials.gov record reports serious adverse events by arm using affected participants over participants at risk. These are the only arm-specific safety counts provided in the trial data.
Serious adverse events
Arm A: CAB + Placebo TDF/FTC + CAB LA · Arm B: TDF/FTC + Placebo CAB + Placebo C
| Safety measure | Arm A | Arm B |
|---|---|---|
| Serious adverse events | 33/1614 | 33/1610 |
The numerators are identical, while the denominators differ by four participants. The ClinicalTrials.gov record does not provide a confidence interval, p-value, formal hypothesis test, or prespecified estimand for this serious-adverse-event comparison. Accordingly, the appropriate interpretation is descriptive rather than a claim of statistical equivalence or superiority.
The reported serious-adverse-event counts show 33 affected participants in each arm, with denominators of 1614 and 1610 respectively. The identical event counts do not by themselves establish that the safety profiles are statistically equivalent, because equivalence or non-inferiority requires a prespecified margin and an appropriate inferential procedure.
The denominators also show why raw counts alone are insufficient for comparing safety: the populations at risk are not exactly the same size. The ClinicalTrials.gov record does not provide the confidence interval or hypothesis test needed to quantify uncertainty around the difference.
8. Statistical Methodology
Randomization
HPTN 084 is explicitly described as randomized. Randomization is the central design mechanism for creating comparable treatment groups with respect to both measured and unmeasured characteristics, in expectation.
The validity of this chain depends on appropriate trial conduct, follow-up, outcome ascertainment, and analysis according to the prespecified design.
Parallel-group design
The registry identifies the design model as parallel. In a parallel-group trial, participants are assigned to one randomized strategy and followed within that strategy rather than being deliberately exposed to multiple randomized treatment sequences.
Quadruple masking
The registry identifies HPTN 084 as quadruple-masked. Masking is particularly relevant in a prevention study because knowledge of assignment can potentially affect participant behavior, reporting of symptoms, clinical management, and assessment processes.
The ClinicalTrials.gov record does not specify exactly which four groups of individuals are included in the registry's quadruple-masking designation. Therefore, this page does not assign specific personnel categories to the four masked groups.
Binary endpoints
Both registered primary endpoints are classified as binary. At the participant level, a binary endpoint ultimately distinguishes whether the participant meets the endpoint definition or does not.
The risk difference expresses the absolute difference between the event probabilities in two groups. A risk ratio instead uses p1/p0. The ClinicalTrials.gov record does not identify which measure was used for the formal primary analysis.
MedDRA classification for adverse events
The registry states that adverse events will be summarized using MedDRA System Organ Class and preferred terms. This provides a standardized vocabulary for organizing clinical and laboratory safety findings and is distinct from the statistical test used to compare randomized groups.
9. Statistical Methods Explained
Why does randomization matter?
Randomization is what allows a treatment comparison to be interpreted primarily as a comparison of assigned strategies rather than as a comparison of self-selected groups. In expectation, randomization balances prognostic factors across groups. Any particular randomized sample can still show numerical differences in baseline characteristics, but those differences are not created by systematic treatment selection.
Why is a binary endpoint different from a continuous endpoint?
A binary endpoint records whether a participant meets a specified condition. Incident HIV infection is an example: the participant either has a documented incident infection according to the endpoint definition or does not. A continuous endpoint, by contrast, records a numerical measurement such as a laboratory value or change from baseline.
What does 33/1614 mean?
The notation 33/1614 means that 33 affected participants were reported among 1614 participants at risk for the stated serious-adverse-event measure in Arm A. It is a count over a denominator, not by itself a treatment-effect estimate.
The corresponding Arm B value is 33/1610. The two numerators are identical, but the denominators are slightly different. A formal comparison would need to use the appropriate participant-level estimand and account for statistical uncertainty.
Why can't identical event counts establish equivalence?
Two groups can have the same number of observed events while differing in their denominators, follow-up, or underlying event probabilities. More importantly, statistical equivalence is a prespecified hypothesis about whether the true treatment difference lies within an acceptable margin. Observing equal counts does not establish that condition.
Why does the HIV testing schedule matter?
Incident HIV infection is detected through repeated testing rather than by a single assessment. The registry specifies testing at enrollment, weeks 2, 4, and 5, then every 4 weeks through week 25, and every 8 weeks thereafter. Repeated assessment improves the opportunity to identify incident infections during follow-up and also creates timing information that can be relevant to the choice of statistical estimand.
Why is quadruple masking statistically useful?
Masking does not replace randomization, but it can reduce opportunities for knowledge of assignment to influence trial behavior or assessment. In a prevention trial, this can matter for both subjective safety reporting and processes surrounding clinical assessment. The ClinicalTrials.gov record identifies quadruple masking but do not specify the four masked groups.
10. Clinical Biostats Interpretation of the Primary Endpoints
The primary HIV endpoint is defined around documented incident infection, confirmed by the HPTN Laboratory Center and Endpoint Adjudication Committee. That definition is stronger than simply recording a preliminary positive test: the endpoint requires confirmation under the specified adjudication process.
Because the registry-reported statistical-analysis field is empty, there is no basis here for reporting a hazard ratio, risk ratio, risk difference, confidence interval, or p-value as the formal HPTN 084 result. Any such number would need to come from a source explicitly reporting the analysis.
The safety endpoint counts participants experiencing treatment-emergent grade 2 or higher clinical and laboratory adverse events in Steps 1 and 2. The endpoint therefore has a participant-level threshold: the relevant question is whether a participant experienced at least one qualifying event during the defined period.
The ClinicalTrials.gov record does not provide the formal comparative analysis. The separate serious-adverse-event counts of 33/1614 and 33/1610 should therefore remain a descriptive safety result rather than being converted into an unsupported statistical conclusion.
11. What the Registry Data Do and Do Not Establish
| Question | What the ClinicalTrials.gov record establishes |
|---|---|
| Was the trial randomized? | Yes. Allocation is listed as RANDOMIZED. |
| Was it parallel-group? | Yes. The design model is PARALLEL. |
| Was it masked? | Yes. Masking is listed as QUADRUPLE. |
| How many participants were enrolled? | 3224. |
| How many arms were there? | 2. |
| Were primary outcomes posted? | Yes. Two outcome measures are posted. |
| Are formal statistical-analysis records reported? | No. No formal statistical analyses were posted. |
| Are primary endpoint effect estimates reported? | No formal estimates are in the posted results. |
| Are confidence intervals posted on ClinicalTrials.gov for the primary endpoints? | No. |
| Are p-values posted on ClinicalTrials.gov for the primary endpoints? | No. |
| Are serious adverse-event counts by arm reported? | Yes: 33/1614 in Arm A and 33/1610 in Arm B. |
This distinction is important for statistical reporting. A trial record can contain outcome data while still not exposing the full analysis specification or inferential results in the particular data extract being reviewed. The safest approach is to report exactly what is available and explain the statistical framework that would ordinarily be applied.
12. Non-Inferiority, Multiplicity, Interim Analysis, and Missing Data
The ClinicalTrials.gov record does not provide a non-inferiority margin, a superiority or non-inferiority hypothesis statement, an alpha-allocation scheme, an interim-analysis schedule, a multiplicity procedure, a missing-data or imputation strategy, or a Bayesian analysis specification.
Non-inferiority margin
No non-inferiority margin is reported in the ClinicalTrials.gov record. A margin should not be inferred from the observed event counts or from the intervention's clinical context.
Multiplicity
No multiplicity-adjustment strategy is reported. The presence of two primary endpoints does not, by itself, reveal how type I error was controlled.
Interim analysis
No interim-analysis schedule or stopping boundary is reported.
Missing data
No imputation method or missing-data strategy is reported in the ClinicalTrials.gov record.
Stratification
No randomization stratification factors are reported in the ClinicalTrials.gov record.
Bayesian methods
No Bayesian analysis specification is reported.
These omissions are not evidence that the trial lacked such procedures. They mean only that the ClinicalTrials.gov record does not document them. A statistical analysis page should distinguish between "not reported in the ClinicalTrials.gov record" and "not used in the trial."
13. Safety Analysis Framework
The primary safety endpoint is defined as the number of participants experiencing grade 2 or higher clinical and laboratory adverse events in Steps 1 and 2. The registry specifies that treatment-emergent adverse events are measured from onset through the participant's last study visit, or the date of the DSMB decision to unblind all participants, whichever is earliest. Assessed at each visit (injections visits q 8 weeks and safety visits q 8 weeks).
There are several statistical layers to such an endpoint:
- Participant-level incidence: whether each participant experienced at least one qualifying adverse event.
- Severity threshold: only grade 2 or higher events meet the primary endpoint definition.
- Time window: the endpoint is restricted to treatment-emergent events within the registry-defined period.
- Clinical classification: events are summarized by MedDRA System Organ Class and preferred terms.
- Seriousness: serious adverse events are a separate safety concept from the grade threshold and should not automatically be treated as synonymous with grade 2 or higher events.
14. Primary Endpoint Interpretation: Statistical Questions to Ask
| Statistical question | Why it matters |
|---|---|
| What is the estimand? | It determines whether the treatment effect is expressed as a risk difference, risk ratio, incidence-rate measure, hazard ratio, or another quantity. |
| What population is analyzed? | Randomized, treated, per-protocol, and other populations can answer different questions. |
| How is follow-up handled? | Repeated HIV testing and variable observation time can affect whether a cumulative-risk or time-to-event approach is appropriate. |
| How are missing assessments handled? | The absence of a test does not automatically mean that an infection did or did not occur. |
| How is multiplicity handled? | Two primary endpoints can create a multiple-testing problem if both are formally tested. |
| What is the prespecified hypothesis? | A superiority, non-inferiority, or other hypothesis requires different interpretation of confidence intervals and decision thresholds. |
15. Serious Adverse Events: A Quantitative Reading
| Arm | Affected | At risk | Reported form |
|---|---|---|---|
| Arm A | 33 | 1614 | 33/1614 |
| Arm B | 33 | 1610 | 33/1610 |
The reported counts permit a straightforward descriptive observation: the number of affected participants is 33 in both arms. The denominators are 1614 and 1610, so the underlying observed proportions are not mathematically identical. Because the task requires the trial's reported numbers to be preserved exactly, this page does not replace the reported fractions with newly calculated decimal percentages.
The ClinicalTrials.gov record stop before the formal inferential-analysis stage. The absence of an analysis record means the page should not fill in the missing statistical output.
16. Why the HIV Endpoint Is Statistically Interesting
The incident-HIV endpoint combines a clinically definitive outcome with repeated surveillance. That combination makes it a useful example for understanding the difference between a simple binary endpoint and a longitudinally observed event.
If every participant had identical follow-up and infection status were known with certainty at a common fixed endpoint, a simple two-group risk comparison could be sufficient. In a prevention trial with repeated testing, however, participants can have different amounts of observation and infection can occur between scheduled assessments. The exact statistical approach therefore matters.
The registry's use of an Endpoint Adjudication Committee also emphasizes the distinction between measurement and analysis. Laboratory testing supplies evidence about infection status, adjudication confirms that the endpoint meets the trial definition, and the statistical analysis then compares those endpoint determinations between randomized strategies.
17. Why the Safety Endpoint Is Statistically Interesting
Safety analysis creates a different set of issues. A participant may experience multiple adverse events, the same participant may have events in different organ systems, and events can vary in severity and seriousness. A single participant-level primary endpoint can therefore coexist with a much richer event-level descriptive safety dataset.
For HPTN 084, the registry wording specifies grade 2 or higher clinical and laboratory adverse events and states that events will be summarized using MedDRA System Organ Class and preferred terms. This means the statistical interpretation should distinguish the primary participant-level endpoint from the detailed taxonomy of individual events.
Participant-level question
How many participants experienced at least one qualifying grade 2 or higher treatment-emergent adverse event?
Event-level question
Which adverse-event categories and preferred terms occurred, and how were they distributed across treatment groups?
18. Statistical Interpretation of Confidence Intervals and P-values
No confidence intervals or p-values are reported in the ClinicalTrials.gov record for either primary endpoint. That absence is important because a point estimate without uncertainty can give a false impression of precision.
A confidence interval would quantify statistical uncertainty around the chosen effect estimator under the specified model or sampling framework. It would not describe the range of outcomes that individual participants could experience.
A p-value, if part of the prespecified analysis, would quantify how compatible the observed data are with a specified null hypothesis under the assumptions of the statistical test. It would not measure the size or clinical importance of the treatment effect.
A statistically significant result can correspond to a small effect, while a clinically meaningful effect may fail to reach a conventional significance threshold in a small sample. Effect size, uncertainty, and the prespecified decision rule therefore need to be considered together.
19. Analysis Populations and Causal Interpretation
The ClinicalTrials.gov record does not specify the formal analysis populations. This matters because randomized clinical trials can define several analysis sets, including all randomized participants, treated participants, and protocol-adherent or other restricted populations.
For the efficacy endpoint, the randomized assignment is generally the foundation for causal comparison. For safety, exposure to treatment can become an important consideration. The exact definitions used by HPTN 084 are not reported here and therefore are not assigned to the trial.
| Potential analysis concept | Statistical role |
|---|---|
| Randomized population | Preserves the treatment assignment created by randomization. |
| Treated population | Can be useful for describing exposure-related safety. |
| Protocol-adherent population | Can address adherence or protocol-defined efficacy questions, but may lose some protection reported by randomization. |
| Registry-specified population | The actual analysis set should be taken from the trial's prespecified statistical documentation. |
20. Missing Data and Censoring
The ClinicalTrials.gov record does not specify a missing-data or imputation strategy. Nevertheless, missingness is an important statistical consideration for both primary endpoints.
For HIV prevention, missed testing visits can create uncertainty about the timing or detection of incident infection. For adverse events, incomplete follow-up can affect the opportunity to observe treatment-emergent events. The appropriate handling depends on the estimand and the prespecified analysis plan.
Importantly, missingness should not automatically be interpreted as treatment failure or treatment success. The statistical analysis needs to define how incomplete observation contributes to the estimator.
21. Trial Design Topics Not Specified in the Supplied Data
| Design topic | Status in the ClinicalTrials.gov record |
|---|---|
| Non-inferiority margin | Not reported |
| Superiority margin | Not reported |
| Multiplicity procedure | Not reported |
| Interim analysis schedule | Not reported |
| Stopping boundaries | Not reported |
| Missing-data / imputation method | Not reported |
| Randomization stratification factors | Not reported |
| Bayesian methods | Not reported |
| Formal statistical-analysis records | None in the posted results |
This section deliberately distinguishes unavailable information from negative findings. For example, the ClinicalTrials.gov record does not establish that no interim analysis occurred; they establish only that no interim-analysis details were provided in the dataset used for this page.
22. Limitations
- No formal statistical-analysis records: no formal statistical analyses were posted, so formal treatment-effect estimates, confidence intervals, and p-values cannot be reproduced from the ClinicalTrials.gov record.
- No arm-level enrollment counts: the overall enrollment is 3224, while the registry-reported serious-adverse-event denominators are 1614 and 1610. The page does not infer that these denominators represent the complete randomized arm sizes.
- No formal non-inferiority framework reported: no margin or decision rule is provided, so the serious-adverse-event data cannot be interpreted as evidence of non-inferiority or equivalence.
- No multiplicity specification: two primary endpoints are listed, but the ClinicalTrials.gov record does not state how type I error was allocated between them.
- No analysis-population definitions: the ClinicalTrials.gov record does not identify the formal efficacy and safety populations.
- No missing-data strategy: the ClinicalTrials.gov record does not specify how missed HIV tests, incomplete follow-up, or other missing observations were handled.
- No subgroup results: the ClinicalTrials.gov record contains no subgroup estimates or interaction analyses.
- No baseline table: baseline characteristics are not reported and are not reported here.
- Serious adverse events are not the same as the primary safety endpoint: the registry-reported 33/1614 and 33/1610 values are reported as serious adverse events, whereas the registered primary safety endpoint concerns grade 2 or higher clinical and laboratory adverse events in Steps 1 and 2.
23. Why This Trial Matters Statistically
HPTN 084 is a useful statistical teaching case because it combines randomized prevention research, repeated ascertainment of a clinically definitive binary endpoint, masked treatment assignment, and participant-level safety outcomes.
| Concept | How it appears in HPTN 084 |
|---|---|
| Randomization | The registry identifies the allocation as randomized. |
| Parallel-group design | Participants are assigned within a parallel design with two arms. |
| Masking | The registry identifies quadruple masking. |
| Binary endpoint | Both registered primary endpoints are classified as binary. |
| Repeated outcome assessment | HIV testing is scheduled repeatedly during follow-up. |
| Endpoint adjudication | Incident HIV infection requires confirmation by the HPTN Laboratory Center and Endpoint Adjudication Committee. |
| Safety analysis | Grade 2 or higher clinical and laboratory adverse events are a registered primary endpoint. |
| MedDRA classification | Adverse events are to be summarized by System Organ Class and preferred terms. |
| Statistical estimands | The endpoint definitions do not by themselves identify the formal effect measure. |
| Uncertainty | No confidence intervals are reported in the statistical-analysis field. |
| Hypothesis testing | No p-values or formal hypothesis tests are reported in the statistical-analysis field. |
| Serious adverse events | Arm-specific affected/at-risk counts are reported as 33/1614 and 33/1610. |
24. A Worked Statistical Thought Process
Suppose an analyst is given the HPTN 084 registry record without a formal statistical-analysis table. A disciplined analysis would proceed in the following order.
For the incident-HIV endpoint, the analyst would first identify exactly what constitutes an event and what follow-up is included. The analyst would then determine whether the prespecified estimand is a cumulative risk, risk ratio, risk difference, incidence rate, hazard ratio, or another measure. Only after that choice could the appropriate confidence interval and hypothesis test be selected.
For the safety endpoint, the analyst would establish whether the primary unit is the participant, the event, or both. Because the registry says "Number of Participants Who Experienced" qualifying adverse events, a participant-level binary analysis is a natural interpretation of the endpoint definition, while MedDRA summaries provide additional event-category information.
25. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The trial uses randomized allocation, a parallel-group design, quadruple masking, and two binary primary endpoints. The ClinicalTrials.gov record does not contain the formal inferential analysis needed to state a numerical primary treatment effect.
Clinical interpretation
The trial addresses HIV prevention in HIV-uninfected women by comparing a long-acting injectable cabotegravir strategy with daily oral TDF/FTC and evaluates both incident HIV infection and treatment-emergent adverse events.
Keeping these two levels separate is important. The clinical question concerns prevention and safety; the statistical question concerns how the randomized comparison should be quantified and how uncertainty should be represented.
26. Results Reporting: Why the Missing Statistical Analysis Matters
The ClinicalTrials.gov record is unusual in an instructive way: it identifies two primary endpoints and states that results are posted, but it supplies no formal entries in the statistical-analysis array. That means a careful statistical page cannot simply infer the official effect estimate from the endpoint definitions.
For example, the incident-HIV endpoint could plausibly support several effect measures. A risk ratio, risk difference, incidence rate ratio, and hazard ratio are not numerically interchangeable, even when calculated from the same underlying participants. They answer related but distinct questions.
The same principle applies to safety. The serious-adverse-event counts are useful descriptive information, but a formal claim about equality, non-inferiority, or a treatment difference requires a prespecified estimand and inferential framework.
27. Related Tutorials
Learn more about the methods used in this trial:
28. Related Statistical Calculators
29. Sources
- ClinicalTrials.gov: HPTN 084, NCT03164564.
- PubMed record: PMID 39311597.
- PubMed record: PMID 35378077.
- PubMed record: PMID 32497490.
The sources above are provided as the permitted primary registry and publication records for this trial. Numerical claims on this page are restricted to the ClinicalTrials.gov record.
Continue through Clinical Biostats
Use the related tutorials and statistical calculators to explore the design and analysis concepts that arise in randomized prevention trials.
30. Record Summary
HPTN 084 is a phase 3 randomized, parallel-group, quadruple-masked prevention trial enrolling 3224 participants and comparing a long-acting injectable cabotegravir strategy with daily oral TDF/FTC in HIV-uninfected women. Its two registered primary endpoints are binary: documented incident HIV infection and grade 2 or higher clinical and laboratory adverse events in Steps 1 and 2. The registry defines incident HIV infection through confirmation by the HPTN Laboratory Center and Endpoint Adjudication Committee and specifies repeated HIV testing throughout follow-up. The safety endpoint is summarized using MedDRA System Organ Class and preferred terms.
The ClinicalTrials.gov record also report serious adverse events in 33 of 1614 participants in Arm A and 33 of 1610 in Arm B. Those counts are useful descriptive safety information, but they do not establish statistical equivalence or non-inferiority because no corresponding margin or formal inferential analysis is reported.
Most importantly, the registry-reported statistical-analysis field is empty. This distinction illustrates a fundamental principle of clinical-trial statistics: the endpoint definition tells us what was measured, while the prespecified estimand and statistical analysis determine how the treatment comparison should be quantified.