This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record. Numerical results on this page are restricted to the trial data posted on ClinicalTrials.gov for PURPOSE 1.
1. Trial at a Glance
PURPOSE 1 was a randomized, parallel, double-masked phase 3 prevention trial evaluating pre-exposure prophylaxis strategies for HIV infection. The registry reports 5368 participants, 5 arms, 2 registered primary endpoints, 7 posted outcome measures, and 9 posted statistical analyses.
| Feature | PURPOSE 1 |
|---|---|
| Trial name | PURPOSE 1 |
| ClinicalTrials.gov identifier | NCT04994509 |
| Phase | Phase 3 |
| Therapeutic area | Infectious Disease |
| Condition | Pre-Exposure Prophylaxis of HIV Infection |
| Primary purpose | Prevention |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Double |
| Enrollment | 5368 |
| Lead sponsor | Gilead Sciences |
| Sponsor type | Industry |
| Status | Active, not recruiting |
2. Clinical Question
The registry describes PURPOSE 1 as a pre-exposure prophylaxis study of lenacapavir and emtricitabine/tenofovir alafenamide in adolescent girls and young women at risk of HIV infection. The principal statistical question represented in the posted primary analysis was whether HIV-1 incidence in the randomized blinded phase for lenacapavir or F/TAF was lower than the estimated background HIV-1 incidence from the incidence phase.
Population
Adolescent girls and young women at risk of HIV infection, as described in the trial's brief title.
Interventions
The registry lists oral lenacapavir, subcutaneous lenacapavir, F/TAF, F/TDF, placebo SC lenacapavir, PTM oral lenacapavir, PTM F/TAF, and PTM F/TDF.
Comparator framework
The primary posted comparisons evaluate randomized blinded phase HIV-1 incidence against the incidence-phase background HIV-1 incidence estimate. Secondary analyses also compare randomized blinded phase groups.
Primary question
Is HIV-1 incidence lower than the estimated background incidence, and do the randomized blinded phase incidence rates satisfy the prespecified statistical criteria?
3. Trial Design
Intervention structure
Lenacapavir
- Oral lenacapavir is listed among the trial interventions.
- Subcutaneous lenacapavir is also listed.
- Placebo SC lenacapavir is listed among the interventions.
F/TAF and F/TDF
- F/TAF is listed as an intervention.
- F/TDF is listed as an intervention.
- PTM oral LEN, PTM F/TAF, and PTM F/TDF are also listed.
The ClinicalTrials.gov record identifies 5 arms but do not provide a complete arm-by-arm allocation table with sample size for every listed intervention. Accordingly, this page does not assign enrollment counts to individual arms beyond the analysis populations and affected/at-risk safety counts explicitly reported in the ClinicalTrials.gov record.
4. Primary Endpoints
| Endpoint | Registered definition / time frame | Statistical profile |
|---|---|---|
| Incidence Phase: Recent Infection Testing Algorithm (RITA) Estimate of the Background Human Immunodeficiency-1 Virus Infection Incidence Rate (bHIV) Per 100 Person Years (PY) | Time frame: Incidence Phase Screening Visit (Day 1). bHIV per 100 PY in the Incidence Phase was calculated using RITA. The RITA incorporated HIV-1 testing results and recency assay testing results to estimate the bHIV. |
Count / rate |
| Randomized Blinded Phase: HIV-1 Incidence Reported Per 100 PY for LEN and F/TAF Compared to Participants in All Screened Set | Time frame: When at least 50% of the participants completed 52 weeks of follow-up after randomization or permanently discontinued from the study (maximum 143 weeks). | Count / rate; time-to-event |
For the randomized blinded phase endpoint, the registry defines HIV-1 incidence per 100 person-years as the number of participants who acquired HIV-1 divided by the total duration of follow-up while at risk for participants without HIV-1 diagnosis, together with follow-up through confirmed HIV-1 diagnosis for participants diagnosed with HIV-1.
The use of person-years allows participants with different amounts of follow-up to contribute according to the amount of time they were observed while at risk. The registry specifies 365.25 days as one year for this calculation.
5. Statistical Methodology
Likelihood ratio testing for the primary lenacapavir comparison
The primary analysis comparing randomized blinded phase lenacapavir with the incidence-phase background HIV-1 incidence used a likelihood ratio test. The effect measure was a rate ratio. The registry describes the confidence interval as based on a likelihood-based method.
Wald testing for F/TAF
The F/TAF primary comparisons used a Wald test, with the rate ratio estimated relative to the incidence-phase background rate. The confidence interval was based on the delta method.
Poisson regression for the randomized secondary comparisons
For the secondary endpoint comparing HIV-1 incidence among lenacapavir, F/TAF, and F/TDF, the registry reports Poisson models. The effect measure was a rate ratio. The lenacapavir-versus-F/TDF confidence interval was obtained from an exact conditional Poisson model, while the F/TAF-versus-F/TDF analysis used a Poisson model.
Hybrid approach for rate differences
The registry also reports a hybrid approach for rate differences between lenacapavir and F/TDF and between F/TAF and F/TDF. The corresponding confidence intervals were described as exact confidence intervals based on the hybrid approach.
Exact conditional logistic regression for adherence
The secondary adherence analysis used an exact conditional logistic regression model. The effect measure was an exact odds ratio comparing participants with HIV-1 diagnosis with matched controls among the F/TAF case-control analysis set.
Intention-to-treat concept
The registry analysis text identifies intention-to-treat analysis as an analysis concept for the incidence comparisons. The randomized blinded phase primary analysis used the Full Analysis Set, which included randomized participants who received at least 1 dose of any study drug and had not been diagnosed with HIV-1, according to the registry description.
6. Primary Results
Lenacapavir versus background HIV-1 incidence
Rate ratio versus background incidence
95% CI: 0.000–0.042 · P < 0.0001
Likelihood ratio test · two-sided 95% CI · superiority hypothesis
| Primary comparison | Rate ratio | 95% CI | P-value | Null hypothesis |
|---|---|---|---|---|
| LEN / bHIV | 0.000 | 0.000–0.042 | <0.0001 | LEN / bHIV ≥ 1 |
| LEN / bHIV | 0.000 | 0.000–0.042 | <0.0001 | LEN / bHIV ≥ 0.8 |
The reported rate ratio of 0.000 means that the registry's estimated HIV-1 incidence rate in the analyzed lenacapavir group, relative to the background incidence estimate, was reported as 0.000. It does not mean that the treatment makes HIV infection biologically impossible, nor does it mean that every participant has zero future risk.
The 95% confidence interval extends from 0.000 to 0.042. The upper limit is therefore the most important part of the interval for understanding how large the rate ratio could be under the reported statistical framework. The interval also lies below both of the tested null thresholds described in the registry analysis notes.
The P < 0.0001 value addresses evidence against the specified null hypothesis under the stated testing procedure. A p-value does not measure the magnitude of an effect, the probability that the treatment works, or the probability that the null hypothesis is true.
The analysis is based on person-time and therefore depends on how follow-up time, HIV-1 diagnosis, and censoring are handled. The registry also defines the analysis around the Full Analysis Set and a comparison with an incidence-phase background rate rather than a conventional randomized treatment-versus-treatment comparison.
F/TAF versus background HIV-1 incidence
Rate ratio versus background incidence
95% CI: 0.550–1.279 · P = 0.20697
Wald test · two-sided 95% CI · superiority hypothesis
| Primary comparison | Rate ratio | 95% CI | P-value | Null hypothesis |
|---|---|---|---|---|
| F/TAF / bHIV | 0.839 | 0.550–1.279 | 0.20697 | F/TAF / bHIV ≥ 1 |
| F/TAF / bHIV | 0.839 | 0.550–1.279 | 0.58674 | F/TAF / bHIV ≥ 0.8 |
A rate ratio of 0.839 means the estimated HIV-1 incidence rate in the F/TAF group was 0.839 times the background incidence estimate under the registry's analysis. As a simple relative description, this corresponds to an estimated rate about 16.1% lower than the background rate, because 1 − 0.839 = 0.161.
The 95% confidence interval of 0.550–1.279 includes 1.000. It therefore spans values compatible with a lower incidence rate as well as values compatible with a higher incidence rate relative to the background estimate.
The reported p-values are 0.20697 for the null hypothesis F/TAF/bHIV ≥ 1 and 0.58674 for the more stringent prespecified criterion F/TAF/bHIV ≥ 0.8. Neither p-value is a measure of effect size. The different null hypotheses explain why the same point estimate has different p-values.
These comparisons should not be interpreted as if the background incidence estimate were simply another randomized treatment arm. It was generated from the incidence phase using the RITA methodology described in the registry.
7. Secondary Incidence Results
Lenacapavir versus F/TDF
Rate ratio
95% CI: 0.000–0.101 · P < 0.0001
Poisson model · exact conditional Poisson confidence interval
The reported rate ratio of 0.000 indicates that the estimated HIV-1 incidence rate for lenacapavir relative to F/TDF was reported as 0.000. The 95% confidence interval extends from 0.000 to 0.101.
The p-value of < 0.0001 provides the registry's reported evidence under its specified superiority analysis. It should not be read as a probability that the observed rate ratio is exactly zero, nor as a measure of the clinical magnitude of the rate difference.
F/TAF versus F/TDF
Rate ratio
95% CI: 0.669–2.143 · P = 0.72820
Poisson model · two-sided 95% CI
The rate ratio of 1.198 indicates an estimated HIV-1 incidence rate 1.198 times that in the F/TDF group in this analysis. The 95% confidence interval, 0.669–2.143, includes 1 and is relatively broad.
The reported p-value of 0.72820 does not provide evidence against the null comparison under the reported statistical procedure. Importantly, this does not establish equality of the two incidence rates. The confidence interval allows for both a lower and a higher incidence rate relative to F/TDF.
Rate differences versus F/TDF
| Comparison | Rate difference | 95% CI | P-value | Registry analysis |
|---|---|---|---|---|
| LEN vs F/TDF | -1.685 | -2.737 to -0.939 | <0.0001 | Hybrid approach; exact CI |
| F/TAF vs F/TDF | 0.333 | -0.869 to 1.367 | 0.20900 | Hybrid approach; exact CI |
The registry specifies a comparability criterion of at most 0.8/100 PY higher for the HIV-1 incidence in the LEN group relative to F/TDF, and the same 0.8/100 PY criterion for F/TAF relative to F/TDF. These are stated as comparability criteria in the registry analysis notes rather than being treated here as an independently inferred equivalence conclusion.
A rate difference is an absolute measure on the incidence-rate scale, unlike a rate ratio, which is relative. The LEN-versus-F/TDF estimate of -1.685 means the estimated incidence rate in LEN was 1.685 HIV-1 events per 100 person-years lower than the F/TDF rate on the reported scale.
The F/TAF-versus-F/TDF estimate of 0.333 means the estimated incidence rate was 0.333 HIV-1 events per 100 person-years higher than the F/TDF rate. Its 95% confidence interval, -0.869 to 1.367, spans both negative and positive rate differences.
8. Adherence Analysis
F/TAF adherence odds ratio
95% CI: 2.06–83.36 · P = 0.0006
Exact conditional logistic regression · exact two-sided 95% CI
| Endpoint | Comparison | Effect | 95% CI | P-value |
|---|---|---|---|---|
| Percentage of participants adherent to F/TAF in HIV-1 diagnosed participants and matched controls | F/TAF participants with HIV-1 diagnosis vs matched controls without HIV-1 diagnosis | Exact OR 9.00 | 2.06–83.36 | 0.0006 |
An odds ratio of 9.00 means that the odds of the adherence outcome were estimated to be 9.00 times as high in the HIV-1 diagnosed participants as in their matched controls under the exact conditional logistic regression analysis.
This is an association within the case-control adherence analysis; it is not a randomized treatment effect and should not be interpreted as saying that poor adherence causes HIV-1 infection on the basis of this estimate alone.
The confidence interval of 2.06–83.36 is wide. It indicates substantial uncertainty about the magnitude of the association even though the interval does not include 1. The p-value of 0.0006 addresses the specified statistical test and does not describe the magnitude or practical importance of the association.
9. Safety
The ClinicalTrials.gov record reports serious adverse events by arm for the randomized blinded phase. These counts are presented as affected participants over participants at risk.
| Randomized blinded phase group | Serious adverse events | At risk |
|---|---|---|
| Lenacapavir | 81 | 2140 |
| F/TAF | 105 | 2135 |
| F/TDF | 44 | 1070 |
The ClinicalTrials.gov record does not provide a formal statistical comparison or confidence interval for these serious adverse-event counts. Accordingly, the counts are described without calculating additional rates or p-values.
10. Statistical Methods Explained
Why use person-years rather than simply count infections?
Participants can contribute different amounts of follow-up. A person-time incidence rate accounts for this by using the amount of time each participant was observed while at risk. PURPOSE 1 reports HIV-1 incidence per 100 person-years, making the denominator a measure of observed risk time rather than simply the number enrolled.
What does a rate ratio of 0.839 mean?
A rate ratio of 0.839 means the estimated incidence rate in the numerator group was 0.839 times the comparison rate. Expressed as a simple relative difference, that is approximately 16.1% lower because 1 − 0.839 = 0.161. It does not mean that exactly 16.1% fewer participants became infected.
Why were likelihood ratio and Wald tests both used?
The registry reports a likelihood ratio test for the lenacapavir primary comparison and a Wald test for the F/TAF primary comparison. Both are hypothesis-testing frameworks, but they obtain their test statistics differently. The choice is part of the prespecified analysis methodology and should not be interpreted as a difference in the importance of the two endpoints simply because the test names differ.
Why use Poisson regression?
Poisson regression is a natural modeling framework for event counts observed over person-time. It can estimate relative incidence rates while incorporating the amount of exposure time. PURPOSE 1 used Poisson models for the randomized blinded phase comparisons involving F/TDF.
Why is the confidence interval important for a rate ratio?
The confidence interval describes statistical uncertainty around the estimated rate ratio under the analysis model. For F/TAF versus background incidence, the interval of 0.550–1.279 spans 1, so the data are compatible with both lower and higher incidence rates relative to the background estimate.
Why is the odds ratio of 9.00 not the same as a risk ratio?
An odds ratio compares odds, not probabilities or incidence rates. An odds ratio of 9.00 therefore cannot be translated directly into "nine times the risk" without additional information about the underlying probabilities and study design. Here, the estimate comes from an exact conditional logistic regression analysis of matched controls.
Why does the p-value not measure effect size?
A p-value measures the statistical evidence against a specified null hypothesis under the analysis framework. It depends on the effect estimate, its variability, the amount of information, and the null value. Effect magnitude is communicated more directly by the rate ratio, rate difference, odds ratio, and their confidence intervals.
11. Non-Inferiority and Comparability Logic
The secondary rate-difference analyses contain an explicit comparability criterion. For LEN versus F/TDF, the registry states that comparability is defined as the HIV-1 incidence in the LEN group being at most 0.8/100 PY higher than in the F/TDF group. The same criterion is stated for F/TAF versus F/TDF.
The registry's criterion is expressed on the absolute incidence-rate-difference scale. It is distinct from the superiority tests against background incidence and should not be collapsed into a single p-value interpretation.
The ClinicalTrials.gov record does not label these analyses as a conventional non-inferiority hypothesis with a conventional non-inferiority p-value. The page therefore describes the registry's stated comparability criterion rather than assigning a separate non-inferiority conclusion.
12. Multiple Hypotheses and the Primary Testing Structure
The primary randomized blinded phase endpoint generated four posted primary analyses: two for lenacapavir versus the background incidence and two for F/TAF versus the background incidence. The paired tests correspond to different null thresholds.
| Analysis | Null threshold | Test | Purpose described in registry |
|---|---|---|---|
| LEN / bHIV | ≥ 1 | Likelihood ratio test | Reject if LEN incidence was significantly lower than bHIV |
| LEN / bHIV | ≥ 0.8 | Likelihood ratio test | Reject if LEN incidence was significantly and at least 20% lower than bHIV |
| F/TAF / bHIV | ≥ 1 | Wald test | Reject if F/TAF incidence was significantly lower than bHIV |
| F/TAF / bHIV | ≥ 0.8 | Wald test | Reject if F/TAF incidence was significantly and at least 20% lower than bHIV |
The distinction matters because the same estimated rate ratio can be tested against different null values. For F/TAF, the point estimate is 0.839, which is below 1 but above 0.8; the two reported p-values are therefore testing different propositions.
13. Intention-to-Treat and the Full Analysis Set
The registry identifies intention-to-treat analysis as an analysis concept for the incidence analyses. The primary analysis population is more specifically described as the Full Analysis Set, including participants in the LEN and F/TAF groups who were randomized, received at least 1 dose of any study drug, and had not been diagnosed with HIV-1.
Why randomization matters
Randomization establishes the intended basis for comparing groups because treatment assignment is determined before outcome information is observed.
Why the analysis population matters
An effect estimate is defined by the population actually included in the analysis. Exclusions based on treatment exposure or HIV-1 diagnosis can affect which participants contribute to the primary estimate.
The ClinicalTrials.gov record does not provide a complete participant flow diagram or the exact number analyzed in every analysis population, so no additional denominators are inferred.
14. Censoring and Person-Time
HIV-1 incidence in this trial is fundamentally a person-time outcome. Participants without HIV-1 diagnosis contribute follow-up while at risk, whereas participants diagnosed with HIV-1 contribute time through the confirmed diagnosis as specified by the registry definition.
Two groups with the same number of HIV-1 events can have different incidence rates if their total time at risk differs. This is why a simple comparison of event counts would not reproduce the registry's incidence-rate analysis.
The registry also states that the primary endpoint was evaluated when at least 50% of participants had completed 52 weeks of follow-up after randomization or permanently discontinued from the study (maximum 143 weeks). The early-stop caveat means that the timing of the analysis is an important part of interpretation.
15. Why a Rate Ratio Is Different From a Risk Ratio
A rate ratio compares incidence rates measured using person-time. A risk ratio would compare cumulative probabilities over a defined period. These quantities are related but are not interchangeable.
| Measure | What it compares | PURPOSE 1 example |
|---|---|---|
| Rate ratio | Incidence rates using person-time | LEN / bHIV = 0.000 |
| Rate difference | Absolute difference in incidence rates | LEN − F/TDF = -1.685 |
| Odds ratio | Odds of an outcome between groups | Adherence analysis OR = 9.00 |
Keeping these effect measures separate prevents a common statistical error: describing a rate ratio as though it were a risk ratio or describing an odds ratio as though it were a relative risk.
16. Results Summary
| Endpoint / comparison | Effect estimate | 95% CI | P-value | Method |
|---|---|---|---|---|
| LEN vs bHIV, primary | Rate ratio 0.000 | 0.000–0.042 | <0.0001 | Likelihood ratio test |
| LEN vs bHIV, 20% lower criterion | Rate ratio 0.000 | 0.000–0.042 | <0.0001 | Likelihood ratio test |
| F/TAF vs bHIV, primary | Rate ratio 0.839 | 0.550–1.279 | 0.20697 | Wald test |
| F/TAF vs bHIV, 20% lower criterion | Rate ratio 0.839 | 0.550–1.279 | 0.58674 | Wald test |
| LEN vs F/TDF | Rate ratio 0.000 | 0.000–0.101 | <0.0001 | Poisson model |
| F/TAF vs F/TDF | Rate ratio 1.198 | 0.669–2.143 | 0.72820 | Poisson model |
| LEN vs F/TDF | Rate difference -1.685 | -2.737 to -0.939 | <0.0001 | Hybrid approach |
| F/TAF vs F/TDF | Rate difference 0.333 | -0.869 to 1.367 | 0.20900 | Hybrid approach |
| F/TAF adherence analysis | Exact OR 9.00 | 2.06–83.36 | 0.0006 | Exact conditional logistic regression |
17. What the Primary Estimates Do — and Do Not — Mean
The reported LEN rate ratio of 0.000 is a model-based estimate of the incidence-rate ratio relative to the background HIV-1 incidence. It should not be interpreted as proof that future HIV-1 infections are impossible or that the biological risk is literally zero.
The F/TAF rate ratio of 0.839 is below 1, corresponding to an estimated rate approximately 16.1% lower than the background estimate on the relative scale. The 95% CI of 0.550–1.279 shows that the data are compatible with a substantially lower rate as well as a higher rate.
A 95% confidence interval describes uncertainty around the estimated parameter under the relevant statistical framework. It is not the range within which individual participant outcomes must fall, and it is not a probability statement that the true value lies inside the particular reported interval.
The p-values reported by the registry quantify evidence against specified null hypotheses. They do not quantify the size of the treatment effect, the probability that a hypothesis is true, or the probability that a future participant will acquire HIV-1.
18. Important Limitations and Interpretation Issues
- Background comparison: the primary lenacapavir and F/TAF analyses compare randomized blinded phase incidence with an incidence-phase background HIV-1 incidence estimate rather than simply comparing two randomized treatment groups.
- Early stopping: the registry states that the study could stop early if prespecified evaluation criteria were met, and the reported completion date represents the point by which at least 50% had been followed for at least 52 weeks.
- Follow-up: incidence-rate estimates depend on person-time and the rules used to determine when participants remain at risk.
- Rate versus risk: rate ratios and rate differences should not be translated directly into cumulative infection risks without additional information.
- Different null hypotheses: the F/TAF analyses illustrate why a p-value depends on the null value being tested. The hypotheses of F/TAF/bHIV ≥ 1 and F/TAF/bHIV ≥ 0.8 are not the same question.
- Adherence analysis: the odds ratio of 9.00 comes from a matched case-control analysis and is not a randomized treatment-effect estimate.
- Safety: the registry-reported serious-adverse-event data are affected/at-risk counts without a corresponding formal comparison in the statistical analyses posted on ClinicalTrials.gov.
19. Why This Trial Matters Statistically
PURPOSE 1 is a useful statistical teaching case because its analysis combines randomized prevention-trial methodology with incidence rates, person-time, a background-incidence comparator, superiority testing, a stated comparability criterion, Poisson regression, exact methods, and a matched adherence analysis.
| Concept | How it appears in PURPOSE 1 |
|---|---|
| Randomization | Randomized allocation in a parallel phase 3 design. |
| Blinding | The registry reports double masking. |
| Person-time | HIV-1 incidence is reported per 100 person-years. |
| Background incidence | The primary LEN and F/TAF analyses compare randomized-phase incidence with bHIV from the incidence phase. |
| Rate ratio | Primary and secondary incidence comparisons use rate ratios. |
| Rate difference | Secondary comparisons with F/TDF include absolute rate differences. |
| Likelihood ratio test | Used for the LEN primary comparisons with bHIV. |
| Wald test | Used for the F/TAF primary comparisons with bHIV. |
| Poisson regression | Used for randomized blinded phase incidence comparisons involving F/TDF. |
| Exact methods | Exact conditional Poisson and exact conditional logistic regression are reported. |
| Odds ratio | The F/TAF adherence analysis reports an exact odds ratio of 9.00. |
| Confidence intervals | Two-sided 95% intervals are reported for the posted effect estimates. |
| Intention-to-treat concept | Identified in the registry analysis text for the incidence analyses. |
20. Statistical Methods Explained in More Depth
Likelihood ratio testing for an incidence rate ratio
A likelihood ratio test compares how well statistical models fit under a constrained null hypothesis versus a less constrained alternative. In PURPOSE 1, the registry reports this approach for the LEN-to-background incidence comparisons. The corresponding rate ratio and confidence interval provide the effect-size and uncertainty perspective alongside the hypothesis test.
Wald testing and the delta method
The Wald test uses an estimate relative to its estimated standard error to assess a specified null hypothesis. For the F/TAF primary analysis, the registry reports a Wald test and states that the confidence interval for the rate ratio versus bHIV is based on the delta method. The delta method approximates the variance of a transformed estimator using a local derivative-based approximation.
Poisson models and exact conditional analysis
Poisson regression is suited to count data observed over person-time. The model can therefore express an incidence-rate relationship between groups. PURPOSE 1 also reports an exact conditional Poisson model for the LEN-versus-F/TDF confidence interval, illustrating that the model used for an estimate and the method used to construct its confidence interval can be specified separately.
Rate differences and rate ratios answer different questions
A rate ratio describes a multiplicative relationship. A rate difference describes an absolute separation on the incidence-rate scale. For example, the LEN-versus-F/TDF results include both a rate ratio of 0.000 and a rate difference of -1.685. These are not redundant statistics: one is relative and the other is absolute.
Exact conditional logistic regression
For the adherence endpoint, the registry reports an exact conditional logistic regression model. The conditional framework is appropriate to the matched case-control structure described in the analysis population. The resulting odds ratio of 9.00 therefore belongs to that matched analysis and should not be interpreted as though it were a randomized incidence-rate ratio.
21. Results in Statistical Context
The strongest statistical distinction in PURPOSE 1 is between the primary background-incidence comparisons and the secondary randomized comparisons with F/TDF.
Primary background comparison
LEN and F/TAF were compared with the incidence-phase bHIV estimate. LEN had a reported rate ratio of 0.000, while F/TAF had a reported rate ratio of 0.839.
Randomized F/TDF comparison
LEN versus F/TDF produced a reported rate ratio of 0.000 and rate difference of -1.685. F/TAF versus F/TDF produced a rate ratio of 1.198 and rate difference of 0.333.
This separation is important because the statistical meaning of a result depends on its comparator. A rate ratio of 0.000 versus bHIV and a rate ratio of 0.000 versus F/TDF are numerically identical point estimates but answer different scientific questions.
22. Trial Timeline
PURPOSE 1 begins
The registry reports the trial start date as 2021-08-30.
Incidence assessment
The randomized blinded phase evaluated HIV-1 incidence per 100 person-years using follow-up after randomization.
Primary analysis timing
The primary randomized blinded phase endpoint was evaluated when at least 50% of participants completed 52 weeks of follow-up after randomization or permanently discontinued from the study (maximum 143 weeks).
Registry primary completion date
The registry reports 2024-05-27 as the primary completion date and explains the early-stop/follow-up convention in its limitation note.
23. What the Incidence Rate Means
An incidence rate of 1 event per 100 person-years describes events relative to accumulated person-time. It should not automatically be read as a one-year cumulative risk of exactly 1% without assumptions about the underlying event process and follow-up.
This distinction is particularly important in prevention trials. Participants may contribute different amounts of follow-up, and the incidence-rate analysis incorporates that unequal observation time. The statistical result is therefore fundamentally a rate-based result rather than a simple percentage of enrolled participants who experienced HIV-1 infection.
24. Primary Statistical Interpretation
The registry reports a LEN-to-background rate ratio of 0.000, with a two-sided 95% CI of 0.000–0.042 and P < 0.0001 for both specified primary null thresholds. Under the reported statistical framework, the estimate is far below the corresponding null values.
The registry reports an F/TAF-to-background rate ratio of 0.839, with a two-sided 95% CI of 0.550–1.279. The p-values are 0.20697 and 0.58674 for the two different null thresholds. The interval spans 1, and the more stringent 0.8 threshold is above the point estimate.
The randomized secondary analyses report a LEN-versus-F/TDF rate ratio of 0.000 and rate difference of -1.685, while the F/TAF-versus-F/TDF analyses report a rate ratio of 1.198 and rate difference of 0.333. These provide a different reference framework from the primary background-incidence analyses.
25. Related Tutorials
Learn more about the methods used in this trial:
26. Related Statistical Calculators
27. Sources
- ClinicalTrials.gov: PURPOSE 1, NCT04994509.
Continue through Clinical Biostats
Explore statistical tutorials and calculators that explain the methods used in randomized clinical trials and incidence-rate analyses.
28. Record Summary
PURPOSE 1 provides a detailed example of prevention-trial statistics built around HIV-1 incidence per 100 person-years. The primary analyses use rate ratios to compare randomized blinded phase incidence with an incidence-phase background HIV-1 incidence estimate, with likelihood ratio testing for LEN and Wald testing for F/TAF. Secondary analyses extend the statistical framework to Poisson regression, rate differences, exact conditional methods, and an adherence analysis using exact conditional logistic regression.
The central statistical lesson is that the comparator and effect measure define what a result means. A rate ratio against background incidence answers a different question from a rate ratio against F/TDF, while an odds ratio from a matched adherence analysis answers a different question again. Confidence intervals provide information about precision, p-values address specified null hypotheses, and person-time determines the scale on which the incidence-rate analyses operate.