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
ANRS PREVENIR is a single-group clinical study of PrEP with Truvada® for HIV prevention in Île-de-France. The registry identifies prevention as the primary purpose and lists reduction of new HIV diagnosis as the primary endpoint over three years.
| Feature | ANRS PREVENIR |
|---|---|
| Official title | Prevention of HIV in "Île-de-France" |
| Condition | HIV Infections |
| Primary purpose | Prevention |
| Design model | Single-group |
| Masking | None |
| Allocation | NA |
| Enrollment | 3,257 |
| Number of arms | 1 |
| Intervention | PrEP with Truvada® (drug) |
| Lead sponsor | ANRS, Emerging Infectious Diseases |
| Sponsor type | OTHER_GOV |
| Study start | 2017-05-03 |
| Primary completion date | 2024-05-02 |
| ClinicalTrials.gov status | UNKNOWN |
| ClinicalTrials.gov | NCT03113123 |
2. Clinical Question
The registry frames ANRS PREVENIR as a prevention study in HIV infections in Île-de-France. Its primary endpoint is the reduction of new HIV diagnosis, defined as the number of new HIV infections, with a three-year time frame.
Population
The study concerns HIV prevention in Île-de-France. The registry identifies the condition as HIV Infections.
Intervention
PrEP with Truvada® (drug).
Comparator
The registry describes a single-group design with one arm; it does not identify a randomized comparator arm.
Primary question
What is the number of new HIV infections over the registry-defined three-year primary endpoint time frame?
3. Trial Design
The single-group structure is the central statistical feature of ANRS PREVENIR. Because there is one study arm rather than randomized treatment and control groups, the principal analysis question is not a between-arm treatment comparison. Instead, the study measures the occurrence of new HIV infections within the enrolled study population over the prespecified time frame.
4. Endpoints
| Endpoint | Time frame | Registry definition |
|---|---|---|
| Reduction of new HIV diagnosis | three years | Number of new HIV infections |
The registry lists one registered primary endpoint. Its operational description is the number of new HIV infections, measured over a three-year time frame.
Endpoint type
Statistically, this is an event-count endpoint. The fundamental observed quantity is the number of participants who experience a new HIV infection during the defined period. The appropriate analysis depends on how follow-up is actually observed, whether participants contribute equal observation time, and whether the registry's underlying statistical analysis specifies an incidence-based or time-to-event framework.
5. Statistical Structure of a Single-Group Prevention Study
The absence of a comparator changes the interpretation of virtually every conventional clinical-trial statistic. A single-group study can provide a direct estimate of the observed incidence of an event in the study population, but it does not by itself estimate a randomized treatment effect.
Observed event count
The most direct summary is the number of new HIV infections occurring during the defined observation period.
Risk
If the denominator is a defined cohort followed for a common period, cumulative incidence can summarize the proportion experiencing the event.
Incidence rate
If follow-up time differs between participants, an incidence rate based on person-time can be more informative than a simple proportion.
Time-to-event analysis
If individual infection times and censoring times are available, survival methods can account for unequal follow-up and censoring.
These are different estimands. A cumulative risk asks what proportion experiences infection during a defined period. An incidence rate asks how frequently infections occur relative to accumulated person-time. A time-to-event analysis additionally describes the distribution of event times while accounting for censoring.
6. Planned Analysis
The registry identifies the primary outcome as the number of new HIV infections over three years. The ClinicalTrials.gov record does not report posted statistical analyses or numerical outcome results.
How an endpoint of this type is typically analysed
For a fixed three-year observation period with complete follow-up, the primary descriptive quantity could be the number of new HIV infections divided by the number of participants under observation. If participants have different amounts of follow-up, an incidence rate based on person-time would provide a more appropriate summary.
This quantity describes the observed probability of experiencing the event over the specified period when the denominator and follow-up definition support that interpretation.
When event times and censoring information are available, a Kaplan-Meier estimator can be used to estimate the probability of remaining free of the event over time. The corresponding cumulative incidence of the event can then be expressed as one minus the estimated event-free survival function when the event structure supports that interpretation.
Here, S(t) represents the estimated probability of remaining free of the event through time t.
If person-time is the appropriate denominator, a rate can instead be expressed conceptually as the number of new infections divided by the accumulated time at risk. Confidence intervals would quantify statistical uncertainty around the estimated risk or rate.
7. Statistical Methods Explained
Why is this not a conventional treatment-effect analysis?
The registry describes ANRS PREVENIR as a single-group study with one arm. There is therefore no randomized treatment-versus-control contrast from which to calculate a treatment hazard ratio, risk ratio, risk difference, or odds ratio.
Why does the three-year time frame matter?
A three-year endpoint defines the period over which the primary event is evaluated. A risk calculated over three years answers a different question from a risk calculated over one year or an incidence rate accumulated over an unspecified period. The time horizon is therefore part of the estimand, not merely a descriptive detail.
When is a simple proportion appropriate?
A simple proportion is most straightforward when participants have comparable follow-up and the denominator is clearly defined as the population at risk at the beginning of the observation period. If follow-up differs materially between participants, person-time or time-to-event methods can use the available observation time more efficiently.
Why might Kaplan-Meier methods be useful?
Kaplan-Meier methods account for right censoring and allow the event-free probability to be estimated at different points in time. This is useful when participants enter or leave observation at different times or when the study contains incomplete follow-up.
What would a confidence interval mean here?
A confidence interval around an estimated risk or incidence rate quantifies sampling uncertainty under the specified statistical model or interval procedure. It does not describe the range of outcomes that individual participants will experience.
Why can an external comparison be problematic?
Without randomization or a prespecified concurrent comparator, differences between the study population and an external population can reflect differences in baseline risk, calendar time, ascertainment, follow-up, or other factors rather than the intervention itself. An external benchmark therefore requires explicit statistical and clinical justification.
8. Interpreting New HIV Infection Counts
The registry's endpoint description is deliberately count-based: number of new HIV infections. A count is an essential starting point, but its interpretation depends on the population and observation time behind that count.
| Quantity | Question answered | Key denominator |
|---|---|---|
| Number of new HIV infections | How many new infections occurred? | None directly |
| Cumulative incidence | What proportion experienced a new infection over the defined period? | Individuals at risk |
| Incidence rate | How frequently did new infections occur relative to observation time? | Person-time at risk |
| Time-to-event estimate | How does the probability of remaining infection-free change over time? | Individuals at risk over time |
These measures should not be treated as interchangeable. A count of infections can remain identical while the incidence rate changes if total follow-up time differs. Similarly, a cumulative incidence estimate depends on the number of people contributing to the denominator and the definition of the observation period.
9. What a Single-Group Design Can Establish
A single-group prevention study can characterize the frequency and timing of new HIV infections in the enrolled population under the study intervention. Its strength for that purpose comes from systematic observation of a defined cohort.
The design does not create the same randomized causal contrast as a parallel-group trial. In particular, an observed infection frequency cannot by itself establish how many infections would have occurred in the same individuals without the intervention. That counterfactual quantity requires either a randomized comparator or a separate design and analytical framework capable of addressing confounding.
The primary endpoint is naturally interpreted as an event-frequency outcome within the study cohort. The central statistical questions concern the number of events, the population at risk, the amount of follow-up, and the uncertainty around the resulting estimate.
The number of new HIV infections alone is not a relative treatment effect. It does not by itself provide a risk ratio, hazard ratio, attributable risk, or causal comparison with an untreated population.
10. Population and Follow-Up Considerations
The study enrolled 3,257 participants. The primary endpoint has a three-year time frame, so interpretation depends on how those participants contribute observation time across that period.
If all participants were observed for the same complete period, a three-year cumulative incidence calculation would be conceptually straightforward. If participants had different follow-up durations, a person-time or survival-analysis framework would better reflect the information actually observed. The registry information presented here does not report the participant-level follow-up structure.
11. Design Features That Are Not Reported
The ClinicalTrials.gov record identifies the single-group structure, lack of masking, and prevention purpose, but does not provide additional statistical design specifications in the information reported here.
| Design topic | Registry information |
|---|---|
| Randomized allocation | Allocation: NA |
| Comparator arm | One arm; single-group design |
| Masking | None |
| Phase | NA |
| Primary endpoint | Reduction of new HIV diagnosis |
| Primary endpoint definition | Number of new HIV infections |
| Primary endpoint time frame | three years |
| Posted statistical analyses | None |
The registry information does not report a non-inferiority margin, factorial structure, crossover, multiplicity strategy, interim-analysis plan, Bayesian methods, missing-data or imputation strategy, or stratification factors. Those methods therefore cannot be incorporated into the statistical description of this record.
12. Results Status
ClinicalTrials.gov results status
The registry record contains no posted statistical analyses for the primary endpoint.
Because no numerical endpoint results or formal statistical analyses are posted in the ClinicalTrials.gov record, there is no reported estimate, confidence interval, or p-value to interpret for the primary endpoint.
This distinction is important statistically. A registered endpoint describes what the study intends to measure; it does not establish the observed value of that endpoint. For ANRS PREVENIR, the registry identifies the endpoint as the number of new HIV infections over three years, while the reported record does not contain the numerical outcome or formal analysis.
13. Limitations
- No randomized comparator: the single-group design does not provide a concurrent randomized control group for estimating a between-group treatment effect.
- Event count requires context: the number of new HIV infections must be interpreted alongside the number of participants at risk and the amount of follow-up.
- Follow-up structure: unequal observation time can make a simple proportion less informative than a person-time or time-to-event analysis.
- External comparisons: comparisons with historical or external populations can be affected by differences in baseline risk, calendar time, ascertainment, and other characteristics.
- Unreported statistical specifications: the registry information does not report the detailed statistical analysis plan, including missing-data, interim-analysis, multiplicity, stratification, or Bayesian specifications.
- No posted numerical results: the ClinicalTrials.gov record does not report a numerical primary endpoint result or formal statistical analysis.
- Generalizability: interpretation applies to the population represented by the study and its enrollment setting rather than automatically to every population at risk for HIV infection.
14. Why This Trial Matters Statistically
ANRS PREVENIR is a useful statistical teaching case because it illustrates a different problem from the conventional randomized clinical trial. The primary outcome is an event count in a prevention cohort, and the study uses a single-group design rather than a randomized treatment-control comparison.
| Concept | How it appears in ANRS PREVENIR |
|---|---|
| Single-group design | One study arm |
| Prevention endpoint | Reduction of new HIV diagnosis |
| Event-count outcome | Number of new HIV infections |
| Time horizon | Three years |
| Cumulative incidence | Potential summary when the denominator and follow-up support a fixed-period risk estimate |
| Incidence rate | Potential summary when participants contribute differing amounts of person-time |
| Time-to-event analysis | Potential framework when event and censoring times are available |
| Uncertainty estimation | Confidence intervals can quantify uncertainty around an estimated risk or rate |
| Causal interpretation | Requires care because there is no randomized comparator |
The trial therefore provides a compact example of an important principle in biostatistics: the correct analysis begins with the estimand and design. The same event can be analysed as a count, a proportion, a rate, or a time-to-event outcome, but each representation answers a somewhat different question.
15. Statistical Methods Explained in More Depth
Count data versus incidence data
A count of new infections is an absolute number. It becomes an incidence measure only after the population at risk and observation time are specified. This distinction prevents a common analytical error: comparing raw event counts from populations with different sizes or different amounts of follow-up.
Person-time and incidence rates
If one participant contributes a longer period of observation than another, person-time methods preserve that information. An incidence rate can be conceptualized as the number of incident events divided by the total time contributed while participants are at risk.
The units depend on the time unit used to accumulate person-time. The interpretation is a rate of event occurrence rather than a probability of experiencing the event.
Risk versus rate
Cumulative incidence is bounded between 0 and 1 because it represents a probability over a defined period. An incidence rate is not a probability and does not have the same upper bound. These measures should therefore not be substituted for one another simply because both describe how frequently events occur.
Kaplan-Meier estimation and censoring
When exact infection times and censoring times are available, Kaplan-Meier estimation can incorporate participants who are observed to remain event-free until their last observation. A censored participant contributes information up to the censoring time rather than being automatically treated as having experienced the event.
Why the hazard ratio is not the natural primary summary here
A hazard ratio is fundamentally a relative comparison between hazard functions. With one study arm and no randomized comparator, there is no treatment-group hazard to divide by another randomized treatment-group hazard. A hazard ratio would therefore require a separately specified comparison framework rather than arising directly from the registered single-group design.
Why confidence intervals matter even without a comparator
Statistical uncertainty remains relevant in a single-group study. If the observed infection frequency is estimated from a finite cohort, the estimate is subject to sampling variability. A confidence interval communicates that uncertainty without turning the single-group estimate into a treatment-effect comparison.
16. Trial Timeline
Study start
The registry lists May 3, 2017 as the study start date.
Single-group prevention study
The registry describes a one-arm, unmasked study evaluating PrEP with Truvada® for HIV prevention.
Three-year assessment
The registered primary endpoint is reduction of new HIV diagnosis, described as the number of new HIV infections, with a three-year time frame.
Primary completion date
The registry lists May 2, 2024 as the primary completion date.
17. Statistical Interpretation of the Study Design
The design naturally supports description of the occurrence of new HIV infections among the enrolled study population over the defined observation period. Depending on the available follow-up information, this can be expressed as a count, cumulative incidence, incidence rate, or time-to-event estimate.
A claim that PrEP caused a particular reduction relative to what would have happened without PrEP requires a counterfactual comparison. In a single-group study, that comparison is not reported by randomization and therefore requires a separately justified design or analytical framework.
The same number of infections can imply very different risks if the number of participants differs. Likewise, the same number of infections can imply different incidence rates if the accumulated person-time differs. Statistical interpretation therefore depends on both events and exposure to risk.
18. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The registered primary endpoint is an event-count measure: the number of new HIV infections over three years. Its analysis should account for the study's single-group design and the structure of participant follow-up.
Clinical interpretation
The study addresses HIV prevention in Île-de-France using PrEP with Truvada®. The registry identifies prevention as the study's primary purpose, while the reported record does not contain numerical primary endpoint results.
19. What the Three-Year Endpoint Does — and Does Not — Mean
The registered endpoint is reduction of new HIV diagnosis, with the description number of new HIV infections and a time frame of three years. The endpoint therefore focuses on incident HIV infection during the specified period.
The endpoint definition alone does not provide a treatment-effect estimate, a comparator risk, a relative risk, or a hazard ratio. Those quantities require additional data and an explicitly defined comparison or statistical model.
If infections occur at different times and participants have different follow-up durations, reducing the analysis to a single final count can discard information about when events occurred. Time-to-event methods can preserve that timing information when the necessary data are available.
20. Key Statistical Takeaways
- Design determines interpretation: ANRS PREVENIR is a single-group study with one arm, so the primary statistical question is not a randomized comparison between treatments.
- The primary endpoint is an event count: the registry defines reduction of new HIV diagnosis as the number of new HIV infections over three years.
- The denominator matters: event counts should be interpreted with the number of participants at risk and the amount of follow-up.
- Risk and rate are different: cumulative incidence describes probability over a defined period, whereas an incidence rate incorporates person-time.
- Time-to-event methods may be informative: when infection and censoring times are available, survival methods can account for unequal follow-up.
- A single-group estimate is not automatically a causal effect: a counterfactual comparison requires additional design or analytical assumptions.
- No formal analyses are posted in the registry record: the ClinicalTrials.gov record does not report a numerical primary endpoint result, confidence interval, or p-value.
21. Sources
- ClinicalTrials.gov: ANRS PREVENIR (NCT03113123).
Continue through the Clinical Biostats knowledge graph
Explore statistical concepts across clinical trial design, endpoint analysis, survival methods, and prevention research.
22. Record Summary
ANRS PREVENIR is a single-group prevention study enrolling 3,257 participants and evaluating PrEP with Truvada® for HIV prevention in Île-de-France. The registered primary endpoint is reduction of new HIV diagnosis, defined as the number of new HIV infections over three years. Statistically, the study illustrates why the distinction between an event count, a cumulative incidence, an incidence rate, and a time-to-event estimate matters, particularly when there is no randomized comparator.
The ClinicalTrials.gov record identifies the study design and primary endpoint but contains no posted statistical analyses or numerical primary endpoint results. The appropriate interpretation is therefore centered on the structure of the estimand and the statistical consequences of a single-group prevention design rather than on an observed treatment-effect estimate.