This page separates trial information reported in the ClinicalTrials.gov record from statistical interpretation. The registry reports that results have been posted, but the ClinicalTrials.gov record contains no formal statistical analyses.
1. Trial at a Glance
PERSEUS is a randomized phase 3 trial evaluating daratumumab added to bortezomib, lenalidomide and dexamethasone in subjects with previously untreated multiple myeloma. The ClinicalTrials.gov record identifies progression-free survival (PFS) as the single registered primary endpoint and classifies it as a time-to-event outcome.
| Feature | PERSEUS |
|---|---|
| Trial name | PERSEUS |
| ClinicalTrials.gov identifier | NCT03710603 |
| Phase | Phase 3 |
| Condition | Multiple Myeloma |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | None |
| Primary purpose | Treatment |
| Enrollment | 709 |
| Primary endpoint | Progression Free Survival (PFS) |
| Primary endpoint type | Time-to-event |
| Trial status | Active, not recruiting |
| Start date | 2018-12-14 |
| Primary completion | 2023-08-01 |
| Lead sponsor | Stichting European Myeloma Network |
| Sponsor type | Network |
2. Clinical Question
The central statistical question is whether adding daratumumab to bortezomib, lenalidomide and dexamethasone changes progression-free survival compared with bortezomib, lenalidomide and dexamethasone alone in subjects with previously untreated multiple myeloma.
Population
Subjects with multiple myeloma who had not previously been treated, according to the registry-reported trial title and registry record.
Intervention
Daratumumab combined with Velcade (bortezomib), lenalidomide and dexamethasone, abbreviated D-VRd.
Comparator
Velcade (bortezomib), lenalidomide and dexamethasone, abbreviated VRd.
Primary question
Does the randomized treatment assignment affect the time from randomization to progressive disease or death?
The ClinicalTrials.gov record does not provide the randomized sample size within each arm. They report total enrollment of 709 and identify two arms.
3. Trial Design
Daratumumab + VRd
- Daratumumab
- Velcade (bortezomib)
- Lenalidomide
- Dexamethasone
VRd
- Velcade (bortezomib)
- Lenalidomide
- Dexamethasone
The ClinicalTrials.gov record identifies the interventions and the two-arm randomized structure but does not provide an arm-specific enrollment table, randomization ratio, treatment schedules, stratification factors, or crossover rules. Those features therefore are not inferred here.
4. Primary Endpoint
| Endpoint | Registry definition and time frame | Endpoint type |
|---|---|---|
| Progression Free Survival (PFS) | From the date of randomization to either progressive disease or death whichever occurred first, up to a maximum follow-up time of 54.41 months. | Time-to-event |
The registry definition further states that PFS was defined as duration from the date of randomization to either progressive disease (PD) or death, whichever occurred first. The registry definition describes progressive disease using disease-assessment criteria, including an increase of ≥ 25% in the level of serum M-protein from the lowest response value with an absolute increase of ≥ 0.5 g/dL, and an increase of ≥ 25% in 24-hour urinary light-chain excretion from the lowest response value.
5. Results Status and Planned Analysis
For a time-to-event primary endpoint such as PFS, the statistical analysis would ordinarily compare the time-to-progression-or-death distributions between the randomized treatment groups. A conventional analysis would estimate the survivor function with Kaplan-Meier methods, compare treatment groups with a log-rank test when appropriate, and estimate a treatment hazard ratio with a Cox proportional-hazards model. Those methods describe how this endpoint is commonly analysed; they are not reported statistical results from PERSEUS in the ClinicalTrials.gov record.
Patients who have not experienced either event at their last evaluable follow-up are ordinarily right-censored according to the prespecified analysis rules. The ClinicalTrials.gov record does not provide the censoring rules or a formal PERSEUS statistical-analysis specification.
What is not reported in the ClinicalTrials.gov record
- A PFS hazard ratio.
- A confidence interval for a treatment effect.
- A P-value for the primary PFS comparison.
- Median PFS in either treatment group.
- The number of PFS events in either treatment group.
- An arm-specific efficacy analysis population.
- Randomization stratification factors.
- An interim-analysis plan or alpha-spending procedure.
- A multiplicity-adjustment strategy.
- A missing-data or imputation procedure.
- A Bayesian analysis specification.
These omissions are important statistically. They prevent a numerical reconstruction of the primary efficacy result from the ClinicalTrials.gov record.
6. Statistical Methodology
Why PFS is a time-to-event endpoint
PFS is not simply a binary response measure. Each participant contributes a time from randomization until the first qualifying event: progressive disease or death. Participants without an observed event by the relevant follow-up point contribute censored observations under the applicable analysis rules.
Kaplan-Meier estimation
For a randomized trial with a time-to-event endpoint, Kaplan-Meier estimation is a standard way to describe the distribution of time to the event while accounting for right censoring.
Here, di is the number of events at event time ti, and ni is the number of participants at risk immediately before that time.
A Kaplan-Meier curve would therefore show how the estimated proportion remaining free of progression or death changes over time. It would also make the timing of events and censoring visible in a way that a single summary number cannot.
Cox proportional-hazards model
A Cox model is commonly used to estimate a relative treatment effect for a time-to-event endpoint. The resulting hazard ratio compares the estimated instantaneous event rates between treatment groups under the model.
A hazard ratio below 1 would indicate a lower estimated instantaneous event rate in the D-VRd group under the fitted model. It would not mean that the same percentage of every participant's individual risk was reduced.
Log-rank comparison
A log-rank test is commonly used to compare survival distributions between randomized groups. It uses information across the follow-up period rather than comparing only one fixed time point.
Randomization and causal comparison
Randomization is especially important for interpreting a treatment comparison. If the randomization procedure is successfully implemented, treatment assignment is independent of measured and unmeasured baseline prognostic characteristics in expectation. This provides the foundation for an intention-to-treat comparison of outcomes according to randomized assignment.
Analysis population
The ClinicalTrials.gov record does not specify whether the primary PFS analysis uses an intention-to-treat population, a modified intention-to-treat population, or another prespecified population. Consequently, no particular efficacy analysis population is attributed to PERSEUS here.
7. Statistical Methods Explained
Why is randomization important for PFS?
Randomization determines treatment assignment independently of the investigator's choice. In a sufficiently large randomized comparison, prognostic characteristics tend to be distributed between groups by chance rather than systematically selected. This supports an interpretation of between-group differences as effects associated with the randomized treatment strategies, subject to protocol adherence, censoring, follow-up, and other design considerations.
Why is PFS analysed as time-to-event data?
The endpoint contains more information than whether progression occurred. It also contains when progression or death occurred. A participant who remains progression-free for a longer period contributes a different amount of follow-up information from one who experiences an event early. Time-to-event methods preserve this timing information and can appropriately handle participants whose event has not been observed by the analysis cutoff.
What does a PFS hazard ratio measure?
A hazard ratio is a relative measure of the event rate at a given point in time under a fitted survival model. For PFS, the event is the first qualifying occurrence of progressive disease or death. An HR below 1 would correspond to a lower estimated instantaneous event rate in the numerator group. It is not equivalent to a probability of progression, a relative risk, or a difference in median PFS.
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 equal. Without the actual PERSEUS hazard ratio and confidence interval in the ClinicalTrials.gov record, there is no numerical precision estimate to interpret.
Why does the P-value not measure effect size?
A P-value addresses compatibility of the observed data with a specified null hypothesis under the statistical model. It depends on both the magnitude of the observed difference and the amount of information available. A small P-value does not itself tell the reader whether an effect is clinically large, and a larger P-value does not prove that the treatment effects are identical.
Why does censoring matter?
Participants can enter the analysis without experiencing progression or death during the available observation period. Their information is therefore incomplete with respect to the event time. Time-to-event methods treat these observations as censored under prespecified rules rather than assigning them an arbitrary event time. If censoring is related to prognosis in ways not handled by the analysis, interpretation can become more complicated.
8. Safety
The ClinicalTrials.gov record provides serious adverse events by randomized arm as affected participants over participants at risk:
| Arm | Serious adverse events affected / at risk |
|---|---|
| Velcade Lenalidomide Dexamethasone (VRd) | 171 / 347 |
| Daratumumab + VRd (D-VRd) | 200 / 351 |
The denominators reported for these safety data are 347 for VRd and 351 for D-VRd. They should not be replaced with the total trial enrollment of 709, because the registry-reported safety measure explicitly reports its own affected/at-risk denominators.
9. What the PFS Endpoint Does — and Does Not — Mean
PFS measures the duration from randomization until the first qualifying occurrence of progressive disease or death. It therefore combines two possible clinical events into a single time-to-event endpoint.
PFS is not the same as overall survival. A participant can remain alive after experiencing progression, so an effect on PFS does not automatically specify the magnitude or direction of an effect on overall survival.
Because progressive disease is defined according to prespecified disease-assessment criteria, the statistical endpoint depends on both the timing of assessment and the rules used to classify progression. The registry definition includes quantitative criteria involving serum M-protein and 24-hour urinary light-chain excretion.
Even if a PFS hazard ratio were available, it would not fully describe the trial. A complete statistical interpretation would ideally consider the estimated survival curves, event counts, censoring, confidence interval, analysis population, and the clinical definition of progression.
10. Design Features That Are Not Reported in the ClinicalTrials.gov record
| Design topic | What the ClinicalTrials.gov record establishes | What is not established |
|---|---|---|
| Randomization | The study is randomized. | Randomization ratio and stratification factors are not reported. |
| Factorial design | The design model is parallel with two arms. | No factorial structure is reported. |
| Crossover | No crossover rule is reported. | Whether crossover was permitted or prohibited is not established here. |
| Non-inferiority | The registry-reported primary endpoint is PFS. | No non-inferiority objective or margin is reported. |
| Interim analysis | No interim-analysis information is reported. | No alpha-spending or stopping boundary can be attributed to the trial from these data. |
| Multiplicity | One registered primary endpoint is reported. | No multiplicity procedure is reported. |
| Missing data / imputation | No imputation strategy is reported. | No specific missing-data method can be attributed to the primary analysis. |
| Bayesian methods | No Bayesian method is reported. | No posterior probability, prior distribution, or Bayesian decision rule is reported. |
This distinction is important because statistical methodology should be tied to the actual prespecified trial analysis rather than inferred from what would be conventional in another study.
11. Trial Timeline
PERSEUS begins
The ClinicalTrials.gov record gives 2018-12-14 as the study start date.
Primary completion date
The ClinicalTrials.gov record gives 2023-08-01 as the primary completion date.
Active, not recruiting
the ClinicalTrials.gov record classifies PERSEUS as ACTIVE_NOT_RECRUITING.
12. Results Reporting: What Can Be Said Responsibly
The ClinicalTrials.gov record creates an important distinction between results availability and statistical results availability in the extracted analysis data. the ClinicalTrials.gov record says that results are posted and that 15 outcome measures are posted. However, the statistical-analysis section in the ClinicalTrials.gov record contains no formal analyses.
What is established
PERSEUS has a registered PFS primary endpoint, results are reported as posted in the ClinicalTrials.gov record, and serious-adverse-event counts by arm are available in the ClinicalTrials.gov record.
What is not established numerically
No PFS effect estimate, confidence interval, P-value, median PFS, or event count is reported in the ClinicalTrials.gov record.
Why this distinction matters
A posted outcome measure is not equivalent to a complete statistical-analysis record. The endpoint name and reported value must be distinguished from a formal inferential comparison.
Educational approach
The appropriate teaching point is how PFS would ordinarily be analysed without presenting a conventional analysis as though it were a reported PERSEUS result.
13. Limitations
- No formal statistical analyses reported: no formal statistical analyses were posted, preventing numerical interpretation of the primary efficacy comparison.
- No treatment-effect estimate: the ClinicalTrials.gov record does not contain a PFS hazard ratio or another formal comparative effect measure.
- No confidence interval or P-value: statistical uncertainty and formal hypothesis-test results cannot be evaluated numerically from the ClinicalTrials.gov record.
- No median PFS: the ClinicalTrials.gov record does not provide median progression-free survival for either arm.
- No event counts: the number of PFS events by treatment group is not reported in the ClinicalTrials.gov record.
- Unknown analysis population: the ClinicalTrials.gov record does not identify the exact population used for the primary PFS analysis.
- Unknown stratification: randomization is documented, but no stratification factors are reported.
- Unknown censoring rules: the ClinicalTrials.gov record does not provide the detailed rules used to censor PFS observations.
- Safety denominator differences: the reported serious-adverse-event denominators of 347 and 351 are not the same as total enrollment of 709 and should be interpreted according to the registry's reported at-risk populations.
- Unmasked design: the registry classifies masking as none, which can be relevant to treatment administration and some forms of outcome assessment, although the ClinicalTrials.gov record does not provide further assessment details.
14. Why This Trial Matters Statistically
PERSEUS is a useful teaching case because its design combines randomization with a primary time-to-event endpoint. The statistical challenge is not simply to compare the number of patients who progressed. It is to compare the timing of progression or death while accounting appropriately for participants whose event has not yet been observed.
| Concept | How it appears in PERSEUS |
|---|---|
| Randomization | The trial is classified as randomized. |
| Parallel-group design | The design model contains two parallel treatment arms. |
| Time-to-event analysis | PFS is the registered primary endpoint and is classified as time-to-event. |
| Composite event definition | PFS ends at progressive disease or death, whichever occurs first. |
| Kaplan-Meier estimation | A standard framework for describing PFS over time when censoring is present. |
| Cox regression | A standard framework for estimating a relative treatment effect for a time-to-event endpoint. |
| Log-rank comparison | A standard framework for comparing time-to-event distributions between randomized groups. |
| Safety analysis | Serious-adverse-event counts are reported separately by arm. |
| Uncertainty | Confidence intervals would normally accompany the primary effect estimate, but none are reported in the extracted analysis data. |
15. Related Tutorials
Learn more about the methods and statistical concepts relevant to this trial:
16. Related Statistical Calculators
17. Sources
- ClinicalTrials.gov: PERSEUS, NCT03710603.
- Linked publication: PubMed record for PMID 41060326.
- Linked publication: PubMed record for PMID 40171013.
- Linked publication: PubMed record for PMID 38084760.
- Linked publication: PubMed record for PMID 35985959.
Continue through Clinical Biostats
Explore statistical tutorials and calculators that build the methods needed to understand randomized clinical trials and their endpoints.
18. Record Summary
PERSEUS is a randomized, parallel-group phase 3 treatment trial in previously untreated multiple myeloma comparing daratumumab plus bortezomib, lenalidomide and dexamethasone with bortezomib, lenalidomide and dexamethasone. The registry identifies progression-free survival as the single registered primary endpoint and classifies it as a time-to-event outcome. The ClinicalTrials.gov record also report serious adverse events by arm. The principal statistical teaching point is therefore the correct analysis of a randomized time-to-event endpoint: defining the event precisely, preserving follow-up time, handling censoring appropriately, and distinguishing descriptive survival estimates from formal treatment-effect inference.