← Clinical Trials
Metastatic Castrate-Resistant Prostate Cancer Phase 3 Completed NCT03511664

VISION: Complete Statistical Analysis of 177Lu-PSMA-617 in Metastatic Castrate-Resistant Prostate Cancer

An independent statistical analysis of the randomized phase 3 VISION trial evaluating 177Lu-PSMA-617 plus best supportive/best standard of care versus best supportive/best standard of care alone in metastatic castrate-resistant prostate cancer.

Trial period: 2018-05-29 to 2021-01-27  ·  Enrollment: 861  ·  Randomized parallel design
Scope of this record

This page separates reported trial results from statistical interpretation. Numerical results and trial characteristics are limited to the ClinicalTrials.gov record. The registry provides the official trial record.

Registry note: 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

VISION was a randomized, open-label, parallel phase 3 trial of 177Lu-PSMA-617 plus best supportive/best standard of care versus best supportive/best standard of care alone in prostate cancer. The registry reports 861 enrolled participants, two arms, two primary time-to-event endpoints, and formal analyses for both primary endpoints.

861
Enrollment
Phase 3
2
Arms
Randomized parallel design
0.40
rPFS HR
99.2% CI 0.29–0.57
0.62
OS HR
95% CI 0.52–0.74
FeatureVISION
Trial nameVISION
Brief titleStudy of 177Lu-PSMA-617 In Metastatic Castrate-Resistant Prostate Cancer
PhasePhase 3
StatusCOMPLETED
ConditionProstate Cancer
AllocationRANDOMIZED
Design modelPARALLEL
MaskingNONE
Primary purposeTREATMENT
Enrollment861.0
Interventions177Lu-PSMA-617; Best supportive/best standard of care
Lead sponsorEndocyte
Sponsor typeINDUSTRY
ClinicalTrials.govNCT03511664

2. Clinical Question

The primary statistical question was whether adding 177Lu-PSMA-617 to best supportive/best standard of care improved the two registered time-to-event endpoints compared with best supportive/best standard of care alone.

Population

The trial's brief title identifies patients with metastatic castrate-resistant prostate cancer; the registry condition is Prostate Cancer.

Intervention

177Lu-PSMA-617 plus best supportive/best standard of care.

Comparator

Best supportive/best standard of care alone.

Primary question

Does the intervention improve radiographic progression-free survival and overall survival relative to the comparator?

3. Trial Design

01
Randomize861 enrolled
02
2 armsParallel allocation
03
Treat177Lu-PSMA-617 + BS/BSOC or BS/BSOC
04
AssessrPFS, OS, response and safety
05
AnalyzeSurvival and categorical methods
INTERVENTION

177Lu-PSMA-617 plus BS/BSOC

  • 177Lu-PSMA-617
  • Best supportive/best standard of care
COMPARATOR

BS/BSOC alone

  • Best supportive/best standard of care

The registry describes the allocation as randomized, the design model as parallel, and masking as none. The trial's primary purpose was treatment. These features establish the basic framework for interpreting the efficacy comparisons: treatment assignment is randomized, the two treatment strategies are compared in parallel, and the study is open-label.

4. Trial Timing and Registry Results

2018-05-29

Trial start

The registry lists May 29, 2018 as the trial start date.

2021-01-27

Primary completion

The registry lists January 27, 2021 as the primary completion date.

Registry status

Completed

The ClinicalTrials.gov record is listed as COMPLETED and reports 31 outcome measures, 9 statistical analyses, and results for both registered primary endpoints.

5. Primary Endpoints

EndpointRegistered time frameRegistry definition
Radiographic Progression-free Survival (rPFS) From date of randomization until date of radiographic progression or date of death from any cause, whichever comes first, assessed up to 32 months (Primary Analysis cut-off date = 27-Jan-2021) Time in months from randomization to radiographic disease progression based on central review assessment per PCWG3 criteria or death due to any cause. Patients alive without radiographic progression at the analysis data cut-off were censored.
Overall Survival (OS) From date of randomization until date of death from any cause, assessed up to 32 months Time in months from randomization to death due to any cause. If the patient was not known to have died, OS was censored at the date of the last study visit or contact until the cut-off date.

Both primary endpoints are time-to-event endpoints. That matters statistically because not every participant necessarily experiences the event during observation. Instead of treating follow-up as a simple binary outcome, survival analysis uses the amount of observed follow-up and accommodates right censoring.

6. Analysis Populations and Statistical Framework

EndpointAnalysis populationComparisonPrimary methodEffect measure
rPFS PFS Full Analysis Set (PFS-FAS) 177Lu-PSMA-617 + BS/BSOC vs BS/BSOC alone One-sided stratified log-rank test Hazard ratio
OS Full Analysis Set (FAS) 177Lu-PSMA-617 + BS/BSOC vs BS/BSOC alone One-sided stratified log-rank test Hazard ratio
OS secondary model Full Analysis Set (FAS) 177Lu-PSMA-617 + BS/BSOC vs BS/BSOC alone Stratified Cox proportional-hazards model Hazard ratio

The registry analysis text identifies intention-to-treat analysis and stratified analysis concepts for the primary time-to-event comparisons. The rPFS and OS primary tests are described as one-sided stratified log-rank tests. The registry also posts a Cox proportional-hazards analysis for OS.

7. Statistical Methodology

Log-rank testing

The log-rank test compares the observed pattern of event occurrence between randomized groups across follow-up. It is particularly suited to time-to-event endpoints because it uses the ordering of event times while accounting for patients who are censored.

Conceptual comparison
H0: the treatment groups have the same event-time distribution

The registry identifies superiority as the hypothesis type for the posted primary analyses. A one-sided test evaluates evidence in the prespecified direction rather than splitting the type I error between both directions.

Stratified analysis

The primary rPFS and OS analyses are described as stratified. Stratification allows the time-to-event comparison to account for prespecified groups or factors used in the analysis rather than treating all observations as though they necessarily have the same baseline risk structure.

Cox proportional-hazards model

The registry also reports a final OS analysis using a Cox proportional-hazards model. The model estimates a relative hazard while allowing the baseline hazard to remain unspecified. The posted analysis was stratified by LDH (≤260 vs. >260 IU/L), liver metastases (yes/no), ECOG score (0-1 vs. 2), and NAAD inclusion in best supportive care at randomization (yes/no). IRT data were used for stratification.

Conceptual Cox model
h(t|X) = h0(t) exp(βX)

For a treatment indicator, exp(β) is interpreted as a hazard ratio comparing the treatment groups under the model. The hazard ratio is a relative time-to-event measure; it is not an absolute probability and does not directly state how many patients will experience an event.

Chi-squared testing

The registry used a chi-squared method for the binary secondary endpoints Overall Response Rate, Disease Control Rate, Percentage of Participants Achieving Prostate-specific Antigen (PSA) Response, and Prostate-specific Antigen 80 (PSA80) Response. The posted analyses specify two-sided Wald's chi-square tests with stratification.

Intention-to-treat principle

The registry identifies intention-to-treat analysis among the concepts used for the primary survival analyses. The principle is important because randomized treatment assignment remains the basis for comparison rather than selectively retaining only participants who adhered to treatment or completed follow-up.

8. Primary Result: Radiographic Progression-free Survival

The registry reports a formal primary analysis of rPFS in the PFS Full Analysis Set (PFS-FAS). The analysis compared 177Lu-PSMA-617 plus best supportive/best standard of care with best supportive/best standard of care alone using a one-sided stratified log-rank test.

Hazard ratio for radiographic progression or death

0.40

99.2% CI: 0.29–0.57   ·   P < 0.001

Hypothesis type: Superiority  ·  Analysis population: PFS-FAS

The reported hazard ratio of 0.40 corresponds to an estimated instantaneous event rate approximately 60% lower in the 177Lu-PSMA-617 group relative to the comparator under the reported time-to-event analysis.

Clinical Biostats interpretation

What the estimate means: HR 0.40 is a relative comparison of the event hazard over follow-up. Here, the event is radiographic progression or death, whichever comes first.

What it does not mean: it does not mean that 40% of patients avoided progression, that every patient had a 60% reduction in risk, or that the absolute probability of progression or death was reduced by 60 percentage points.

Precision: the reported 99.2% confidence interval is 0.29–0.57. It describes statistical uncertainty around the estimated hazard ratio under the specified analysis framework; it is not a range of individual patient effects.

P-value: P < 0.001 describes the strength of evidence against the null hypothesis under the reported one-sided testing framework. It does not measure the magnitude of the treatment effect. The hazard ratio and confidence interval provide the effect-size information.

Important caution: a hazard ratio is model-based and summarizes relative event hazards rather than giving an absolute survival probability. Interpretation of a Cox-type hazard ratio also depends on the proportional-hazards framework when that model is used.

9. Primary Result: Overall Survival

Overall survival was the second registered primary endpoint. The registry reports a primary OS analysis in the Full Analysis Set (FAS) using a one-sided stratified log-rank test.

Hazard ratio for death

0.62

95% CI: 0.52–0.74   ·   P < 0.001

Hypothesis type: Superiority  ·  Analysis population: FAS

The reported OS hazard ratio of 0.62 corresponds to an estimated instantaneous rate of death approximately 38% lower in the 177Lu-PSMA-617 group relative to best supportive/best standard of care alone, under the reported analysis.

Clinical Biostats interpretation

What the estimate means: HR 0.62 is a relative comparison of the estimated hazard of death between the randomized groups. A value below 1 indicates a lower estimated hazard in the 177Lu-PSMA-617 group.

What it does not mean: it does not mean that 38% of participants survived, that 38% of participants were cured, or that every patient experienced exactly a 38% reduction in their personal probability of death.

Precision: the 95% CI of 0.52–0.74 gives the statistical uncertainty around the estimated hazard ratio. It is narrower than the rPFS interval partly because the two endpoints use different confidence levels; the interval itself should not be interpreted as a distribution of individual treatment effects.

P-value: P < 0.001 indicates strong evidence against the null under the reported one-sided stratified log-rank framework. It does not quantify the size or clinical importance of the observed effect.

Analysis population and censoring: OS is defined from randomization to death, with censoring for participants not known to have died. Consequently, the analysis depends on the recorded follow-up and censoring information.

OS Cox proportional-hazards analysis

The registry separately reports a Cox proportional-hazards analysis for OS in the same Full Analysis Set.

Final OS Cox proportional-hazards estimate

0.68

95% CI: 0.58–0.81

Stratified by LDH, liver metastases, ECOG score, and NAAD inclusion in best supportive care at randomization.

The two OS estimates should not be treated as duplicate measurements of a single number. The registry identifies the 0.62 estimate with the one-sided stratified log-rank primary analysis and the 0.68 estimate with the Cox proportional-hazards model. They arise from different statistical components of the analysis framework and therefore answer closely related, but not identical, statistical questions.

Clinical Biostats interpretation

The Cox estimate of 0.68 remains below 1, indicating a lower estimated hazard of death for the intervention group under that model. Its 95% CI of 0.58–0.81 quantifies uncertainty around that model-based estimate.

The difference between 0.62 and 0.68 is not evidence by itself that one analysis is "correct" and the other is "incorrect." The registry identifies different methods: a one-sided stratified log-rank test for the primary OS comparison and a stratified Cox model for the additional OS analysis.

10. Secondary Time-to-Event Results

Time to First Symptomatic Skeletal Event

The registry reports Time to First Symptomatic Skeletal Event (SSE) as a secondary time-to-event endpoint in the PFS Full Analysis Set. The analysis used a two-sided stratified log-rank test.

Hazard ratio for Time to First Symptomatic Skeletal Event

0.50

95% CI: 0.40–0.62   ·   P < 0.001

Under the reported analysis, HR 0.50 represents an estimated instantaneous event rate approximately 50% lower in the intervention group relative to the comparator. As with the primary survival endpoints, this is a relative hazard measure rather than an absolute event probability.

Progression-free Survival

The registry separately reports Progression-free Survival (PFS) as a secondary time-to-event endpoint in the PFS Full Analysis Set. The analysis used a two-sided stratified log-rank test and reports a hazard-ratio effect measure.

Hazard ratio for progression or death

0.30

95% CI: 0.24–0.38   ·   P < 0.001

The reported HR of 0.30 corresponds to an estimated instantaneous rate approximately 70% lower for the defined progression-or-death event in the intervention group under the reported analysis.

Secondary endpointPopulationMethodEffect95% CIP-value
Time to First Symptomatic Skeletal Event (SSE)PFS-FASTwo-sided stratified log-rankHR 0.500.40–0.62< 0.001
Progression-free Survival (PFS)PFS-FASTwo-sided stratified log-rankHR 0.300.24–0.38< 0.001

11. Secondary Binary Endpoint Results

Overall Response Rate

Overall Response Rate (ORR) was analyzed in the Response Evaluable Analysis Set. The registry reports a two-sided Wald's chi-square test with stratification and an odds-ratio effect measure.

Odds ratio for Overall Response Rate

24.99

95% CI: 6.05–103.24   ·   P < 0.001

An odds ratio of 24.99 means that the estimated odds of the defined response were 24.99 times as large in the intervention group as in the comparator group under the reported analysis.

Clinical Biostats interpretation

Odds are not probabilities. An odds ratio of 24.99 should not be read as "24.99 times as many patients responded." Odds are calculated as probability divided by one minus probability, so the relationship between an odds ratio and an absolute response-rate difference depends on the underlying response probabilities.

The 95% CI of 6.05–103.24 is wide, indicating substantial statistical uncertainty around the estimated odds ratio despite the reported P-value of < 0.001. The confidence interval is therefore important when interpreting the magnitude of the effect.

The P-value addresses evidence against the null hypothesis under the reported test. It does not measure the size of the response effect or the width of the confidence interval.

Disease Control Rate

Disease Control Rate (DCR) was also analyzed in the Response Evaluable Analysis Set using a two-sided Wald's chi-square test with stratification.

Odds ratio for Disease Control Rate

5.79

95% CI: 3.18–10.55   ·   P < 0.001

The reported odds ratio of 5.79 indicates that the estimated odds of disease control were 5.79 times as large in the intervention group as in the comparator group under the registry analysis.

PSA Response

The registry reports Percentage of Participants Achieving Prostate-specific Antigen (PSA) Response as a secondary binary endpoint in the PFS Full Analysis Set. The time frame extends from randomization until 30 days of safety follow-up and was assessed up to approximately 32 months.

Odds ratio for PSA Response

11.19

95% CI: 6.3–20.0   ·   P < 0.001

The estimated odds of PSA response were 11.19 times as large in the intervention group as in the comparator group under the reported analysis.

PSA80 Response

Prostate-specific Antigen 80 (PSA80) Response was another secondary binary endpoint analyzed in the PFS Full Analysis Set using a two-sided Wald's chi-square test with stratification.

Odds ratio for PSA80 Response

23.6

95% CI: 8.6–65.1   ·   P < 0.001

The estimated odds of PSA80 response were 23.6 times as large in the intervention group as in the comparator group under the reported analysis.

Secondary binary endpointAnalysis populationMethodOdds ratio95% CIP-value
Overall Response Rate (ORR)Response Evaluable Analysis SetTwo-sided Wald's chi-square, stratified24.996.05–103.24< 0.001
Disease Control Rate (DCR)Response Evaluable Analysis SetTwo-sided Wald's chi-square, stratified5.793.18–10.55< 0.001
Percentage achieving PSA responsePFS-FASTwo-sided Wald's chi-square, stratified11.196.3–20.0< 0.001
PSA80 ResponsePFS-FASTwo-sided Wald's chi-square, stratified23.68.6–65.1< 0.001

12. Endpoint Definitions and Censoring

The registry's endpoint definitions illustrate why the analysis of VISION cannot be reduced to a simple comparison of percentages. Both primary endpoints are defined from the date of randomization and incorporate censoring.

rPFS

Events are radiographic disease progression based on central review according to PCWG3 criteria or death due to any cause, whichever comes first. Participants alive without radiographic progression at the analysis cut-off are censored.

OS

The event is death from any cause. Participants not known to have died are censored at their last study visit or contact according to the registry definition.

Why censoring matters

A censored participant contributes information up to the time of last known follow-up rather than being treated as though an event had or had not occurred after that point.

Why randomization matters

Starting follow-up at randomization preserves a common time origin for the randomized comparison of treatment strategies.

13. One-Sided and Two-Sided Testing

The VISION registry record provides an instructive contrast between the testing frameworks used for different analyses. The primary rPFS and OS analyses are described as one-sided stratified log-rank tests. The secondary binary endpoints use two-sided Wald's chi-square tests, and the SSE analysis uses a two-sided stratified log-rank test.

AnalysisTesting framework reportedWhy the distinction matters
rPFS primary analysisOne-sided stratified log-rankTests the prespecified superiority direction rather than allocating the testing error to both directions.
OS primary analysisOne-sided stratified log-rankSame directional framework for the primary survival hypothesis.
ORRTwo-sided Wald's chi-squareAllows evidence in either direction relative to the null.
DCRTwo-sided Wald's chi-squareUses a two-sided categorical comparison.
Time to First SSETwo-sided stratified log-rankUses a two-sided survival comparison.

A one-sided P-value and a two-sided P-value should therefore not be compared without considering the prespecified hypothesis and testing framework. The directionality of the test is a design feature, not an effect-size measure.

14. Stratification in the OS Cox Model

The registry provides unusually specific information about the stratification factors used in the final OS Cox proportional-hazards analysis.

Stratification factorCategories reported
LDH≤260 vs. >260 IU/L
Liver metastasesYes/no
ECOG score0-1 vs. 2
NAAD inclusion in best supportive care at randomizationYes/no
Stratification data sourceIRT data

In a stratified Cox model, the baseline hazard can differ across strata while the treatment effect is estimated across the specified strata. This can be useful when important baseline factors are expected to influence the underlying event hazard.

Interpretive distinction: stratification does not mean that the trial was separately randomized within each factor in the sense of creating independent trials. It means the analysis accounts for the specified strata when estimating or testing the treatment comparison.

15. Safety Results

The ClinicalTrials.gov record reports serious adverse events by arm. The affected and at-risk counts are presented exactly as provided.

Study componentTreatment groupSerious adverse events
Main Study177Lu-PSMA-617 Plus Best Sup195/529
Main StudyBest Supportive/Best Standar58/205
Sub Study177Lu-PSMA-617 Plus Best Supp9/30

These figures should be interpreted as affected participants / participants at risk, not as hazard ratios, odds ratios, or percentages. The ClinicalTrials.gov record distinguishes a Main Study from a Sub Study, so the counts should not be pooled into a single safety estimate.

Safety interpretation: the ClinicalTrials.gov record provides serious-adverse-event counts by study component and arm, but do not provide a corresponding formal between-arm statistical analysis in the statistical-analyses ClinicalTrials.gov record. It would therefore be inappropriate to infer a comparative P-value, confidence interval, or effect measure from the counts alone.

16. Statistical Methods Explained

Why was a log-rank test used for rPFS and OS?

rPFS and OS are time-to-event endpoints. A log-rank test is designed to compare the event-time experience of two groups while incorporating censoring. It is therefore more appropriate than a simple chi-squared comparison at one arbitrary follow-up time when the endpoint is defined by time from randomization to an event.

What does an rPFS hazard ratio of 0.40 mean?

It means the estimated instantaneous rate of the defined progression-or-death event was approximately 40% of the comparator rate under the reported analysis. Equivalently, it represents an approximately 60% lower estimated hazard. It does not say that 60% of participants were protected from progression or death.

Why is the rPFS confidence interval 99.2% rather than 95%?

The registry reports a 99.2% two-sided confidence interval for the rPFS hazard ratio. Confidence-level choice is part of the prespecified statistical framework. A higher confidence level generally produces a wider interval than a lower confidence level for the same underlying estimate and data.

What does an odds ratio of 24.99 mean for ORR?

An OR of 24.99 means that the estimated odds of the defined response were 24.99 times as large in the intervention group as in the comparator group. Odds are not the same as probabilities, so the odds ratio cannot be interpreted as a 24.99-fold increase in the response percentage.

Why use a Cox proportional-hazards model in addition to the log-rank test?

The log-rank test provides a hypothesis test for the survival distributions, while the Cox model provides an estimated hazard ratio and allows the analysis to incorporate stratification. In VISION, the registry reports a stratified Cox model for OS with four specified stratification factors.

Why does the analysis population matter?

The rPFS analysis uses the PFS Full Analysis Set, OS uses the Full Analysis Set, and response endpoints use the Response Evaluable Analysis Set. These populations are not interchangeable. An effect estimate is meaningful only in the population to which its analysis applies.

Why does a small P-value not prove a large treatment effect?

A P-value measures evidence against a null hypothesis under the specified statistical model and testing framework. It does not measure the magnitude of an effect. That is why the VISION results should be read using the hazard ratio or odds ratio together with its confidence interval, rather than relying on the P-value alone.

17. Confidence Intervals: Reading the VISION Results

EndpointEstimateConfidence levelConfidence intervalWhat it communicates
rPFSHR 0.4099.2%0.29–0.57Uncertainty around the primary rPFS hazard-ratio estimate.
OS, log-rank analysisHR 0.6295%0.52–0.74Uncertainty around the primary OS hazard-ratio estimate.
OS, Cox modelHR 0.6895%0.58–0.81Uncertainty around the stratified Cox model estimate.
ORROR 24.9995%6.05–103.24Uncertainty around the odds-ratio estimate; notably wide.
DCROR 5.7995%3.18–10.55Uncertainty around the estimated odds ratio.
Time to First SSEHR 0.5095%0.40–0.62Uncertainty around the time-to-event effect.
PFSHR 0.3095%0.24–0.38Uncertainty around the secondary PFS hazard ratio.
PSA ResponseOR 11.1995%6.3–20.0Uncertainty around the estimated response odds.
PSA80 ResponseOR 23.695%8.6–65.1Uncertainty around the estimated PSA80 response odds.

The confidence intervals also show why effect estimates should not be read in isolation. For example, the ORR odds ratio is 24.99, but its 95% confidence interval extends from 6.05 to 103.24. That interval indicates much more uncertainty about the exact magnitude of the odds ratio than the point estimate alone would suggest.

18. Primary vs Secondary Analyses

The registry identifies two primary endpoints and seven posted secondary analyses in the ClinicalTrials.gov record. The distinction is important because primary endpoints form the central confirmatory questions of the trial, while secondary endpoints provide additional information about disease control, response, and related time-to-event outcomes.

EndpointRoleEndpoint typeEffect measure
Radiographic Progression-free Survival (rPFS)PrimaryTime-to-eventHazard ratio
Overall Survival (OS)PrimaryTime-to-eventHazard ratio
Overall Response Rate (ORR)SecondaryBinaryOdds ratio
Disease Control Rate (DCR)SecondaryBinaryOdds ratio
Time to First Symptomatic Skeletal Event (SSE)SecondaryTime-to-eventHazard ratio
Progression-free Survival (PFS)SecondaryTime-to-eventHazard ratio
PSA ResponseSecondaryBinaryOdds ratio
PSA80 ResponseSecondaryBinaryOdds ratio

The registry reports 9 statistical analyses in total, including 3 primary-endpoint analyses and 6 additional statistical analyses beyond those primary analyses. The ClinicalTrials.gov record includes the two primary endpoints plus the secondary analyses summarized above.

19. Multiplicity and Interpretation of Multiple Endpoints

VISION evaluates multiple endpoints, including two primary time-to-event endpoints and several secondary outcomes. Multiple statistical tests create a broader inferential framework than a single isolated hypothesis test.

Two primary endpoints

rPFS and OS are both designated primary endpoints in the registry. Their results therefore form the central efficacy analysis.

Different endpoint types

rPFS, OS, SSE and PFS are time-to-event endpoints, while ORR, DCR, PSA response and PSA80 response are binary endpoints.

Different tests

Primary survival analyses use one-sided stratified log-rank tests, while the listed secondary binary endpoints use two-sided Wald's chi-square tests.

Do not pool P-values

Each P-value belongs to its own endpoint, analysis population, effect measure, and testing framework. They should not be treated as interchangeable evidence units.

The ClinicalTrials.gov record does not specify an alpha-allocation hierarchy or multiplicity-adjustment procedure beyond identifying the primary analyses and their testing directions. Accordingly, this page does not infer an unreported multiplicity procedure.

20. What the Hazard Ratio Does — and Does Not — Mean

rPFS: HR 0.40

The rPFS hazard ratio of 0.40 indicates an estimated instantaneous rate of radiographic progression or death approximately 60% lower in the intervention group than in the comparator under the reported analysis.

It does not mean that 60% of patients were progression-free, nor does it establish a 60-percentage-point absolute improvement in progression-free survival.

OS: HR 0.62

The primary OS hazard ratio of 0.62 indicates an estimated instantaneous rate of death approximately 38% lower in the intervention group under the reported one-sided stratified log-rank analysis.

It does not specify the absolute survival probability at any particular time and does not mean that every participant experienced the same proportional reduction in individual risk.

Cox OS: HR 0.68

The stratified Cox estimate of 0.68 is a separate model-based estimate of the relative hazard of death. Its 95% CI of 0.58–0.81 describes uncertainty around that estimate.

Why the confidence interval matters

Point estimates are only one part of statistical interpretation. Confidence intervals indicate how precisely the study data estimate the underlying relative effect under the stated analysis framework. The wide ORR interval of 6.05–103.24, for example, communicates much greater uncertainty about the exact odds-ratio magnitude than the ORR point estimate of 24.99 alone would show.

21. Limitations and Interpretation Issues

22. Why This Trial Matters Statistically

VISION is a useful teaching case because the ClinicalTrials.gov record bring together several core clinical-trial methods in a single randomized phase 3 study: time-to-event endpoints, stratified log-rank testing, Cox proportional-hazards modeling, binary response analysis, odds ratios, one-sided and two-sided testing, multiple analysis populations, and censoring.

ConceptHow it appears in VISION
RandomizationThe registry classifies allocation as RANDOMIZED.
Parallel designThe registry identifies the design model as PARALLEL with 2 arms.
Time-to-event analysisBoth registered primary endpoints are time-to-event outcomes.
Log-rank testUsed for the primary rPFS and OS analyses and for secondary SSE and PFS analyses.
One-sided testingThe primary rPFS and OS analyses are described as one-sided stratified log-rank tests.
Stratified analysisStratification is identified in the primary analyses and explicitly specified for the OS Cox model.
Cox modelA stratified Cox proportional-hazards model is reported for OS.
Hazard ratioUsed for rPFS, OS, SSE, and PFS.
Chi-squared testUsed for ORR, DCR, PSA response, and PSA80 response.
Odds ratioUsed for the four listed binary secondary endpoints.
Confidence intervalsReported alongside the posted hazard-ratio and odds-ratio estimates.
CensoringExplicitly incorporated into the registry definitions of rPFS and OS.
Different analysis populationsPFS-FAS, FAS, and Response Evaluable Analysis Set are used across endpoints.

23. A Statistical Reading of the Complete Results

The results form several related statistical layers rather than one single summary number. The two primary endpoints address time-to-event outcomes. The rPFS analysis reports HR 0.40 with a 99.2% CI of 0.29–0.57 and P < 0.001. The OS primary analysis reports HR 0.62 with a 95% CI of 0.52–0.74 and P < 0.001. A separate stratified Cox analysis of OS reports HR 0.68 with a 95% CI of 0.58–0.81.

The secondary analyses extend that picture. Time to First Symptomatic Skeletal Event has HR 0.50 (95% CI 0.40–0.62; P < 0.001), while PFS has HR 0.30 (95% CI 0.24–0.38; P < 0.001). The binary endpoints also produce odds ratios above 1: ORR 24.99, DCR 5.79, PSA response 11.19, and PSA80 response 23.6, all with P < 0.001 under the reported analyses.

Statistically, the important point is that these estimates are not interchangeable. A hazard ratio summarizes a relative time-to-event effect, while an odds ratio compares odds for a binary outcome. Their confidence intervals quantify uncertainty on different scales. The analysis population and testing framework also differ across endpoints.

This is why a rigorous trial interpretation should preserve the structure of the statistical analysis rather than compressing all results into a single claim. The primary endpoints, secondary endpoints, effect measures, confidence intervals, P-values, censoring rules, and analysis populations each contribute different information.

24. Related Tutorials

Learn more about the methods used in this trial:

25. Related Statistical Calculators

26. Sources

Continue through the Clinical Biostats statistical pathway

Explore tutorials and statistical calculators covering survival analysis, categorical data, confidence intervals, hazard ratios, odds ratios, and clinical-trial methodology.

27. Record Summary

VISION provides a detailed example of how randomized clinical-trial evidence can be expressed through several complementary statistical methods. The two registered primary endpoints are time-to-event outcomes analyzed with one-sided stratified log-rank testing. The reported rPFS hazard ratio is 0.40 with a 99.2% CI of 0.29–0.57, while the primary OS hazard ratio is 0.62 with a 95% CI of 0.52–0.74. A separate stratified Cox analysis reports an OS hazard ratio of 0.68 with a 95% CI of 0.58–0.81.

The secondary analyses add categorical and time-to-event perspectives: ORR, DCR, PSA response, and PSA80 response are evaluated with stratified chi-squared methods and odds ratios, while Time to First Symptomatic Skeletal Event and PFS are evaluated with stratified log-rank methods and hazard ratios. Across these analyses, the appropriate interpretation depends on the endpoint definition, analysis population, effect measure, confidence interval, and testing framework.

The central statistical lesson is that a clinical-trial result is not adequately represented by a P-value alone. A complete analysis connects the randomized comparison, endpoint definition, analysis population, statistical method, effect estimate, confidence interval, and censoring framework. VISION provides examples of each of these components within a single phase 3 trial.

Clinical Biostats methodology: This page distinguishes registry-reported results from statistical interpretation. No median survival, subgroup estimates, baseline characteristics, or other numerical results not contained in the ClinicalTrials.gov record has been added or inferred.