← Clinical Trials
Metastatic Hormone-Sensitive Prostate Cancer Phase 3 Time-to-Event Analysis NCT02799602

ARASENS: Complete Statistical Analysis of Darolutamide in Metastatic Hormone-Sensitive Prostate Cancer

An independent statistical analysis of the randomized phase 3 ARASENS trial evaluating darolutamide in addition to standard androgen deprivation therapy and docetaxel versus placebo plus docetaxel in metastatic hormone-sensitive prostate cancer.

Trial start: 30 NOV 2016  ·  Primary completion: 25 OCT 2021  ·  Enrollment: 1306
Scope of this record

This page separates reported trial results from statistical interpretation. Numerical results are taken from the publicly reported 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

ARASENS was a randomized, parallel-group, quadruple-masked phase 3 trial evaluating darolutamide in addition to standard androgen deprivation therapy and docetaxel in patients with metastatic hormone-sensitive prostate cancer. The registry reports two primary time-to-event outcome measures for overall survival and eight formal statistical analyses, including one primary-endpoint analysis.

1306
Enrolled
Phase 3 trial
2
Treatment arms
Parallel design
0.675
Primary OS HR
95% CI 0.568–0.801
<0.0001
Primary p-value
One-sided log-rank
FeatureARASENS
PhasePhase 3
ConditionMetastatic hormone-sensitive prostate cancer
DesignRandomized, parallel-group
MaskingQuadruple
Primary purposeTreatment
Enrollment1306
Primary endpoint typeTime-to-event
Primary endpoints registered2
Formal statistical analyses posted8
Results postedYes
Lead sponsorBayer
ClinicalTrials.govNCT02799602

2. Clinical Question

The statistical question is whether adding darolutamide to standard androgen deprivation therapy and docetaxel changes time-to-event outcomes compared with placebo plus docetaxel in metastatic hormone-sensitive prostate cancer.

Population

Participants with metastatic hormone-sensitive prostate cancer enrolled in the ARASENS phase 3 trial.

Intervention

BAY1841788 / darolutamide (ODM-201), together with standard androgen deprivation therapy and docetaxel.

Comparator

Placebo plus docetaxel, with standard androgen deprivation therapy forming part of the trial treatment framework.

Primary question

Does the darolutamide-containing treatment strategy produce a statistically superior overall-survival time-to-event outcome compared with the placebo-containing strategy?

3. Trial Design

01
Randomize1306 participants
02
Parallel armsTwo treatment groups
03
Quadruple maskingMasked trial design
04
Time-to-eventOS and secondary outcomes
05
Statistical comparisonStratified log-rank
Allocation
Randomized allocation in a parallel-group phase 3 design.
Masking
Quadruple masking was specified in the registry.
Primary purpose
Treatment.
Endpoint structure
The registered primary endpoints are both time-to-event measures of overall survival.
DAROLUTAMIDE ARM

Darolutamide + Docetaxel

  • BAY1841788 / darolutamide (ODM-201)
  • Standard androgen deprivation therapy
  • Docetaxel
CONTROL ARM

Placebo + Docetaxel

  • Placebo
  • Standard androgen deprivation therapy
  • Docetaxel

The treatment comparison is therefore not a simple comparison of darolutamide versus no therapy. The registered groups compare a darolutamide-containing treatment strategy with a placebo-containing strategy within a randomized trial framework that includes standard androgen deprivation therapy and docetaxel.

4. Primary Endpoints

Registered endpointTime frameEndpoint typeFormal analysis
OS From Date of Randomization Until Death From Any Cause - Number of Events From randomization of the first participant until death from any cause up to 25 OCT 2021 cut-off date; 533 OS events were reached (approximate 59 months) Time-to-event Stratified one-sided log-rank test; HR from stratified Cox regression
OS From Date of Randomization Until Death From Any Cause - Month From randomization of the first participant until death from any cause up to 25 OCT 2021 cut-off date; 533 OS events were reached (approximate 59 months) Time-to-event Primary endpoint measure posted; the formal statistical analysis is reported for the OS event endpoint

Registry definition of overall survival

Overall survival was defined as the time from the date of randomization until death from any cause. The registered treatment period states that treatment was provided twice daily until disease progression, symptomatic progressive disease, change of systemic antineoplastic therapy, unacceptable toxicity, consent withdrawal, withdrawal at the discretion of the investigator, death, or non-compliance. Active follow-up visits from the discontinuation of the darolutamide or placebo treatment period for up to 1 year or until the patient could no longer travel to the clinic, died, was lost to follow-up, or withdrew informed consent and actively objected to collection of further data. Long-term (Survival) follow-up period: After Active follow-up, patients continued to be contacted approximately every 12 weeks by phone. The end of the Survival follow-up period was defined as when the patient died, was lost to follow-up, withdrew consent, or at the end-of-study.

The primary endpoint therefore has a clear time origin—randomization—and a clearly defined event—death from any cause. This structure is important statistically because participants who have not experienced death by the analysis cutoff are not simply treated as if they had survived indefinitely; their follow-up is handled through time-to-event methods and censoring.

5. Statistical Methodology

Primary comparison: stratified log-rank test

The registry reports a log-rank test as the formal method for the primary overall-survival comparison. The analysis notes specify a one-sided p-value and stratification by extent of disease (EOD) and alkaline phosphatase (ALP).

Primary testing framework
One-sided log-rank test + stratification by EOD and ALP

The formal hypothesis type was superiority. The registry therefore describes a directional test in which the analysis evaluates whether the darolutamide-containing arm has a lower hazard than the placebo-containing arm.

Cox regression for the hazard ratio

The reported hazard ratio and its 95% confidence interval were based on a Cox regression model stratified by EOD and ALP. The EOD categories were specified as non-regional lymph-node metastases only; bone metastases with or without lymph-node metastases; and visceral metastases with or without lymph-node metastases or with or without bone metastases. ALP was categorized as <ULN versus ≥ULN.

Hazard-ratio framework
HR < 1  →  lower estimated hazard in the darolutamide + docetaxel arm

The registry explicitly states that an HR below 1 indicates superiority of the darolutamide + docetaxel arm over the placebo + docetaxel arm.

Why both methods are reported

The log-rank test and Cox model answer related but distinct statistical questions. The log-rank test supplies the formal comparison of the time-to-event distributions, while the Cox model provides a relative effect estimate—the hazard ratio—and its confidence interval.

This distinction matters because a p-value alone does not communicate the magnitude or precision of an effect. Conversely, a hazard ratio without its uncertainty interval does not show how precisely the treatment effect has been estimated.

6. Primary Overall Survival Result

The primary analysis compared darolutamide (BAY1841788) + docetaxel with placebo + docetaxel. The registry reports 533 OS events at the 25 OCT 2021 cutoff.

Hazard ratio for overall survival

0.675

95% CI: 0.568–0.801   ·   One-sided P < 0.0001

Method: stratified log-rank test; HR and 95% CI from stratified Cox regression

Primary endpointDarolutamide + Docetaxel vs Placebo + Docetaxel
OS events533
Hazard ratio0.675
95% confidence interval0.568–0.801
P-value<0.0001
HypothesisSuperiority
TestOne-sided stratified log-rank
Cox modelStratified by EOD and ALP
Clinical Biostats interpretation

An HR of 0.675 means that, under the fitted Cox model, the estimated instantaneous hazard of death in the darolutamide + docetaxel group was 0.675 times that in the placebo + docetaxel group. Expressed as a relative complement, this corresponds to an estimated 32.5% lower hazard because 1 − 0.675 = 0.325.

The HR does not mean that 32.5% of participants avoided death, that 32.5% of participants benefited, or that every participant experienced the same reduction in individual risk. It is a relative time-to-event measure derived from the Cox model.

The 95% CI of 0.568–0.801 describes statistical uncertainty around the estimated hazard ratio under the analysis framework. It does not describe the range of treatment effects across individual patients.

The p-value <0.0001 addresses the evidence against the null hypothesis under the specified one-sided log-rank testing framework. It does not measure the size of the treatment effect. The effect size is described by the HR, while its precision is described by the confidence interval.

Because the effect estimate comes from a Cox model, interpretation also depends on the appropriateness of the proportional-hazards framework. The registry does not provide enough information in the ClinicalTrials.gov record to evaluate the proportional-hazards assumption directly.

What the primary result establishes statistically

The posted analysis is explicitly a superiority analysis. The HR is below 1, the 95% confidence interval is 0.568–0.801, and the one-sided log-rank p-value is <0.0001. Taken together, these are the reported statistical components of the primary comparison.

The appropriate interpretation is narrower than saying that darolutamide "reduces mortality by 32.5%" in an absolute sense. The result is a model-based relative hazard comparison between the randomized treatment strategies over the analysis period.

7. Secondary Time-to-Event Results

The registry contains seven additional formal statistical analyses. Each uses the same broad framework: a log-rank test, one-sided testing, stratification by EOD and ALP, and a hazard ratio based on a Cox regression model stratified by those factors.

Secondary endpointHR95% CIP-value
Time to Castration-Resistant Prostate Cancer 0.357 0.302–0.421 <0.0001
Time to Pain Progression 0.792 0.660–0.950 0.0058
Symptomatic Skeletal Event Free Survival (SSE-FS) 0.609 0.516–0.718 <0.0001
Time to First Symptomatic Skeletal Event 0.712 0.539–0.940 0.0081
Time to Initiation of Subsequent Antineoplastic Therapy 0.388 0.328–0.458 <0.0001
Time to Worsening of Disease-Related Physical Symptoms 1.043 0.894–1.217 0.7073
Time to Initiation of Opioid Use for ≥7 Consecutive Days 0.688 0.523–0.906 0.0037

Time to Castration-Resistant Prostate Cancer

Hazard ratio

0.357

95% CI: 0.302–0.421   ·   P < 0.0001

The HR of 0.357 corresponds to a lower estimated instantaneous hazard of the first castration-resistant prostate cancer event in the darolutamide-containing group under the fitted model. The confidence interval is entirely below 1, and the registry reports a one-sided p-value of <0.0001.

Time to Pain Progression

Hazard ratio

0.792

95% CI: 0.660–0.950   ·   P = 0.0058

The reported HR is below 1, indicating a lower estimated hazard of pain progression in the darolutamide-containing group under the analysis model. The confidence interval provides the uncertainty around that relative effect estimate.

Symptomatic Skeletal Event Free Survival

Hazard ratio

0.609

95% CI: 0.516–0.718   ·   P < 0.0001

SSE-FS was defined as a time-to-event outcome involving the first occurrence of a symptomatic skeletal event or death from any cause, whichever occurred first. An HR below 1 therefore represents a lower estimated hazard of reaching this composite endpoint in the darolutamide-containing group.

Time to First Symptomatic Skeletal Event

Hazard ratio

0.712

95% CI: 0.539–0.940   ·   P = 0.0081

The HR of 0.712 indicates a lower estimated hazard of the first symptomatic skeletal event in the darolutamide-containing group under the reported model. This endpoint differs conceptually from SSE-FS because SSE-FS includes death from any cause as part of the event definition.

Time to Initiation of Subsequent Antineoplastic Therapy

Hazard ratio

0.388

95% CI: 0.328–0.458   ·   P < 0.0001

The HR of 0.388 indicates a substantially lower estimated instantaneous hazard of initiation of the first subsequent systemic antineoplastic therapy in the darolutamide-containing group under the reported Cox model. This is a treatment-pathway endpoint rather than a direct measure of survival.

Time to Worsening of Disease-Related Physical Symptoms

Hazard ratio

1.043

95% CI: 0.894–1.217   ·   P = 0.7073

This estimate is above 1 rather than below 1. The 95% CI spans 1, and the reported p-value is 0.7073. Thus, within this posted analysis, the estimate does not provide evidence of a statistically detectable lower hazard for this endpoint under the specified superiority testing framework.

Time to Initiation of Opioid Use for ≥7 Consecutive Days

Hazard ratio

0.688

95% CI: 0.523–0.906   ·   P = 0.0037

The HR of 0.688 indicates a lower estimated hazard of the first opioid use for at least 7 consecutive days in the darolutamide-containing group under the reported model.

Multiplicity caution: The registry identifies these analyses as superiority tests and reports one-sided p-values, but the ClinicalTrials.gov record does not provide a complete multiplicity hierarchy or alpha-allocation scheme across the secondary endpoints. Individual p-values should therefore be interpreted in the context of the trial's overall testing strategy rather than treated as though each secondary endpoint were an isolated confirmatory experiment.

8. Endpoints and Time Frames

The registry lists 17 posted outcome measures, with two registered primary endpoints and the formal analyses summarized above. The registry text contains shortened time-frame descriptions for several secondary measures; those descriptions are retained here without adding dates or follow-up values not present in the ClinicalTrials.gov record.

EndpointRegistry time frame
OS From Date of Randomization Until Death From Any Cause - Number of Events From randomization of the first participant until death from any cause up to 25 OCT 2021 cut-off date; 533 OS events were reached (approximate 59 months)
OS From Date of Randomization Until Death From Any Cause - Month From randomization of the first participant until death from any cause up to 25 OCT 2021 cut-off date; 533 OS events were reached (approximate 59 months)
Time to Castration-Resistant Prostate Cancer - Month From randomization of the first participant to the first occurrence of an CRPC event up to 25 OCT 2021 cut-off date approximately 59 months
Time to Pain Progression - Month From randomization of the first participant to the first occurrence of a pain progression event up to 25 OCT 2021 cut-off date approximately 59 months
Symptomatic Skeletal Event Free Survival (SSE-FS) - Month From randomization of the first participant to the first occurrence of an SSE event or death from any cause, whichever occurred first up to 25 OCT 2021 cut-off date approximately 59 months
Time to First Symptomatic Skeletal Event (SSE) - Month From randomization of the first participant to the first occurrence of an SSE event up to 25 OCT 2021 cut-off date approximately 59 months
Time to Initiation of Subsequent Antineoplastic Therapy - Month From randomization of the first participant to the initiation of first subsequent systemic antineoplastic therapy up to 25 OCT 2021 cut-off date approximately 59 months
Time to Worsening of Disease-Related Physical Symptoms - Month From randomization of the first participant to the first increase in disease-related physical symptoms based on the NCCN-FACT-FPSI-17 questionnaire up to 25 OCT 2021 cut-off date approximately 59 months
Time to Initiation of Opioid Use for ≥7 Consecutive Days - Month From randomization of the first participant to the first opioid use for ≥7 consecutive days up to 25 OCT 2021 cut-off date approximately 59 months

9. Statistical Methods Explained

Why was a log-rank test used?

The registered primary endpoint is time-to-event overall survival. A log-rank test is designed to compare the event-time distributions of randomized groups while accounting for different follow-up times and censoring. The ARASENS registry specifically reports the log-rank test as its formal comparison method.

What does an HR of 0.675 mean?

It means the estimated instantaneous hazard of death in the darolutamide + docetaxel group was 0.675 times the hazard in the placebo + docetaxel group under the fitted Cox model. It is a relative hazard measure, not an absolute probability and not a statement that every participant experiences the same treatment effect.

Why is the confidence interval important?

The 95% CI of 0.568–0.801 shows the statistical uncertainty surrounding the primary HR estimate. A point estimate alone cannot show how precisely the treatment effect has been estimated. The interval also remains below 1, which is consistent with the reported superiority result.

Why does the p-value not measure effect size?

The p-value evaluates evidence against the null hypothesis under the specified testing framework. It depends on both the observed data and the amount of information available. The HR communicates the estimated relative effect, while the confidence interval communicates precision.

Why was the analysis stratified?

The registry specifies stratification by extent of disease and alkaline phosphatase. Stratified analysis allows the time-to-event comparison to account for these prespecified categories rather than treating all participants as though the distribution of these factors were identical across the analysis.

What does one-sided testing change?

The ARASENS registry explicitly reports a one-sided p-value for the log-rank test. A one-sided test allocates the specified type I error to a directional alternative. Consequently, its p-value should not be interpreted as though it came from an unspecified two-sided test. The reported confidence interval, however, is explicitly two-sided and uses 95% coverage.

Where does Kaplan-Meier estimation fit?

Kaplan-Meier estimation is the standard descriptive framework for displaying and estimating time-to-event survival functions in randomized trials. The registry-reported ARASENS registry data identify the endpoint type and report log-rank and Cox methods, but do not separately identify Kaplan-Meier estimation as a posted statistical-method field. It is therefore useful as a methodological teaching concept without claiming that a specific Kaplan-Meier estimate was reported here.

10. Stratification by EOD and ALP

The primary and secondary analysis notes consistently specify stratification by extent of disease (EOD) and alkaline phosphatase (ALP). This is an important feature of the statistical design because the same stratification structure is carried through the reported log-rank comparisons and Cox regression models.

Stratification factorRegistered categories
EOD Non-regional lymph nodes metastases only; bone metastases with or without lymph node metastases; visceral metastases with or without lymph node metastases or with or without bone metastases
ALP <ULN vs ≥ULN

Stratification does not change the basic interpretation of the HR. Rather, it specifies how the randomized treatment comparison is adjusted for the prespecified strata in the survival analysis. The same EOD and ALP structure is reported for the primary and secondary formal analyses reported in the ClinicalTrials.gov record.

11. Understanding the Time-to-Event Framework

All eight posted formal analyses reported in the ClinicalTrials.gov record are time-to-event analyses. This means that the statistical outcome is not merely whether an event occurred, but when the event occurred relative to randomization.

Time origin

The registered definitions use randomization as the starting point for the time-to-event analysis.

Event

Each endpoint has a specific event definition, ranging from death to disease-related, skeletal, symptom, therapy-initiation, and opioid-use events.

Censoring

Time-to-event methods allow participants without an observed event at the analysis cutoff to contribute their available follow-up without being treated as having experienced the event.

Relative effect

The hazard ratio summarizes the relative instantaneous event hazard between the randomized treatment groups under the Cox model.

This framework is especially useful when participants have different amounts of follow-up. A conventional binary analysis at a single time point would discard much of the timing information. Survival analysis instead uses the observed event times and available follow-up.

Conceptual survival function
S(t) = P(T > t)

The survival function is the probability that the event time T exceeds time t. Kaplan-Meier estimation is commonly used to estimate this function, while the log-rank test compares survival distributions and the Cox model estimates a relative hazard.

12. Primary Endpoint Interpretation: Relative vs Absolute Effects

The primary result is reported as a hazard ratio rather than as a median survival time, survival probability at a particular time, or absolute risk difference. These measures should not be treated as interchangeable.

MeasureWhat it communicatesARASENS result reported
Hazard ratio Relative instantaneous event hazard under the Cox model 0.675
95% CI Statistical uncertainty around the HR 0.568–0.801
Log-rank p-value Evidence against the null under the one-sided testing framework <0.0001
OS events Number of observed deaths contributing to the posted primary analysis 533

The ClinicalTrials.gov record does not provide median overall survival, survival probabilities at specified time points, or absolute risk differences. Those measures are therefore not substituted for the reported HR.

Clinical Biostats interpretation

A useful way to read the primary result is to keep three questions separate: How large is the estimated relative effect? The HR is 0.675. How precisely is it estimated? The 95% CI is 0.568–0.801. How strong is the statistical evidence under the prespecified test? The one-sided log-rank p-value is <0.0001.

None of those quantities tells us the absolute probability that an individual participant will survive, the probability of cure, or the magnitude of benefit for every patient. Those questions require different estimands and additional data.

13. Secondary Endpoint Interpretation

The secondary results illustrate why a trial should not be reduced to a single hazard ratio. The estimated effects range from HR 0.357 for time to castration-resistant prostate cancer to HR 1.043 for time to worsening of disease-related physical symptoms.

Reported hazard ratios for formal analyses
Time to CRPC
0.357
Subsequent therapy
0.388
SSE-free survival
0.609
Opioid initiation
0.688
First SSE
0.712
Pain progression
0.792
Physical symptoms
1.043

The visual comparison is descriptive rather than a ranking of endpoints. Different endpoints represent different clinical events and therefore should not be interpreted as though their HRs were measurements of the same underlying outcome.

14. Safety Results

The ClinicalTrials.gov record reports serious adverse events by randomized treatment arm as affected participants divided by participants at risk.

Serious adverse eventsAffected / at risk
Darolutamide (BAY1841788) + Docetaxel 306 / 652
Placebo + Docetaxel 276 / 650

Darolutamide arm

306 participants were reported as affected among 652 at risk for the registry-reported serious-adverse-event measure.

Placebo arm

276 participants were reported as affected among 650 at risk for the registry-reported serious-adverse-event measure.

The ClinicalTrials.gov record does not provide a formal statistical comparison, confidence interval, or p-value for these serious adverse-event counts. They should therefore be presented as descriptive safety information rather than converted into an inferential claim that is not contained in the registry data.

Safety interpretation: The serious-adverse-event figures are separate from the time-to-event efficacy analyses. A treatment effect on overall survival cannot be combined mechanically with serious-adverse-event counts to produce a single benefit-risk statistic without a prespecified framework for doing so.

15. Analysis Populations and What the Registry Supports

The ClinicalTrials.gov record identifies the randomized design, enrollment, treatment groups, statistical methods, and reported analyses, but they do not provide a complete analysis-population table or detailed definitions of every analysis set.

FeatureWhat is supported by the ClinicalTrials.gov record
Randomized designYes; allocation is listed as randomized.
Parallel groupsYes.
Quadruple maskingYes.
Primary endpoint typeTime-to-event.
Formal primary analysisYes; one primary-endpoint analysis is posted.
Hazard-ratio estimate with CIYes.
Detailed ITT definitionNot provided in the ClinicalTrials.gov record.
Detailed per-protocol definitionNot provided in the ClinicalTrials.gov record.
Detailed missing-data/imputation strategyNot provided in the ClinicalTrials.gov record.

This distinction is important. It is reasonable to explain what an ITT or per-protocol analysis means generally, but it would not be appropriate to assign a specific analysis population to ARASENS without supporting information in the ClinicalTrials.gov record.

16. Missing Data, Censoring, and Follow-Up

Time-to-event analysis naturally involves censoring because participants can reach the analysis cutoff without experiencing the event. The ClinicalTrials.gov record establishes that the endpoints are time-to-event measures and provide the analysis cutoff for the primary OS analysis, but they do not provide a detailed censoring algorithm or missing-data/imputation plan.

That limitation matters when interpreting survival analyses. Censoring is not equivalent to an observed absence of the event for all future time. A censored participant contributes information up to the point at which their event status is no longer observed under the analysis framework.

What is not assumed here: No specific missing-data imputation method, censoring rule beyond the registered endpoint structure, or sensitivity analysis is attributed to ARASENS because those details are not contained in the ClinicalTrials.gov record.

17. Non-Inferiority, Equivalence, and Superiority

ARASENS is explicitly identified in the ClinicalTrials.gov record as a superiority trial. There is no non-inferiority or equivalence margin reported in the ClinicalTrials.gov record.

Superiority

The statistical objective is to test whether the treatment effect favors the darolutamide-containing strategy rather than merely demonstrating that it is not unacceptably worse.

No NI margin reported

The ClinicalTrials.gov record does not provide a non-inferiority margin, so none is introduced into the interpretation.

This distinction is important because an HR below 1 in a superiority analysis has a different inferential context from a non-inferiority analysis. In a non-inferiority trial, the key question is whether the confidence interval excludes the prespecified unacceptable-loss margin. That logic does not apply to the ARASENS result as reported in the registry.

18. Interim Analysis, Factorial Design, and Crossover

The ClinicalTrials.gov record identifies a randomized parallel-group design and do not report a factorial design, crossover component, or interim-analysis procedure.

Design topicInformation in the ClinicalTrials.gov record
Factorial designNot reported; design model is parallel.
CrossoverNot reported in the ClinicalTrials.gov record.
Interim analysisNot reported in the ClinicalTrials.gov record.
Alpha-spending methodNot reported in the ClinicalTrials.gov record.
Multiplicity hierarchyNot fully reported in the ClinicalTrials.gov record.
Bayesian methodsNot reported; the posted analyses use frequentist log-rank and Cox methods.

Absence of these details from the ClinicalTrials.gov record should not be interpreted as proof that a protocol did not contain additional statistical provisions. It means only that they are not available in the trial-data extract used for this page.

19. Interpreting the Seven Secondary Hazard Ratios

The secondary analyses demonstrate several important principles of survival-analysis interpretation.

HR relationship to 1Statistical meaningARASENS examples
HR < 1 Lower estimated event hazard in the darolutamide-containing group 0.357, 0.388, 0.609, 0.688, 0.712, 0.792
HR approximately 1 Little estimated relative difference in instantaneous hazard 1.043 for worsening of disease-related physical symptoms
CI includes 1 The interval includes no relative hazard difference 0.894–1.217 for worsening of disease-related physical symptoms

It is tempting to interpret every HR below 1 as interchangeable evidence of "benefit." Statistically, that is too broad. Each endpoint has a different event definition, and the meaning of the HR depends on the clinical event being modeled.

For example, an HR of 0.388 for time to initiation of subsequent antineoplastic therapy describes the relative hazard of a treatment-pathway event. It is not an HR for death. Similarly, an HR of 0.609 for SSE-FS concerns a composite endpoint whose event definition includes either a symptomatic skeletal event or death from any cause.

20. Limitations

21. Why This Trial Matters Statistically

ARASENS is a useful teaching case because it concentrates several core concepts in modern randomized survival analysis within one trial. The same analytical framework is applied across overall survival and multiple secondary time-to-event outcomes, allowing the statistical interpretation to be compared across endpoints without changing the fundamental model structure.

Statistical conceptHow it appears in ARASENS
RandomizationRandomized phase 3 parallel-group design
MaskingQuadruple masking
Time-to-event endpointsPrimary OS and multiple secondary outcomes
Log-rank testingFormal statistical method for the posted analyses
Hazard ratioPrimary and secondary effect measure
Confidence interval95% two-sided intervals around reported HRs
One-sided testingOne-sided p-values for the reported log-rank analyses
Stratified analysisEOD and ALP used for stratification
Cox regressionSource of reported HRs and confidence intervals
Superiority testingHypothesis type for the formal analyses
Composite time-to-event endpointSSE-FS combines SSE and death from any cause
Safety reportingSerious adverse events reported as affected / at risk by arm

The most instructive feature is the alignment between design and analysis. Randomization creates the treatment comparison; time-to-event methodology preserves information about event timing and censoring; stratification accounts for the specified EOD and ALP categories; the log-rank test supplies the formal comparison; and the Cox model expresses the relative effect as a hazard ratio with an uncertainty interval.

22. A Practical Reading Sequence for the ARASENS Results

A statistically disciplined reading of the primary result can proceed in five steps.

Step 1

Identify the estimand

The primary outcome is time from randomization until death from any cause.

Step 2

Identify the effect measure

The treatment effect is expressed as a hazard ratio from a stratified Cox regression model.

Step 3

Read the point estimate

The reported HR is 0.675, indicating a lower estimated hazard in the darolutamide-containing group under the model.

Step 4

Read the uncertainty

The 95% CI is 0.568–0.801, describing uncertainty around the HR estimate.

Step 5

Read the hypothesis test

The one-sided stratified log-rank p-value is <0.0001 under a superiority framework.

This sequence prevents a common statistical error: starting with the p-value and treating it as the main result. The effect estimate and confidence interval should be understood first, with the p-value providing the inferential test specified by the trial.

23. Clinical Interpretation vs Statistical Interpretation

Statistical interpretation

The primary randomized comparison produced an OS HR of 0.675 with a 95% CI of 0.568–0.801 and a one-sided log-rank p-value of <0.0001. The registered hypothesis type was superiority.

Clinical interpretation

The ClinicalTrials.gov record supports describing a difference between the randomized treatment strategies in the reported time-to-event outcomes. The magnitude and meaning of that difference depend on the specific endpoint, while safety is represented separately by the reported serious-adverse-event counts.

These two levels of interpretation should remain distinct. Statistical significance describes evidence under a specified model and testing framework; clinical interpretation requires consideration of the endpoint, treatment context, magnitude of effect, precision, and safety evidence.

24. Statistical Concepts in This Trial

Learn more about the methods used in this trial:

25. Related Statistical Calculators

26. Sources

Continue through the Clinical Biostats statistical pathway

Explore the underlying survival-analysis methods, statistical calculators, and other clinical-trial results.

27. Record Summary

ARASENS provides a clear example of randomized survival analysis in a phase 3 oncology trial. The ClinicalTrials.gov record describes a randomized, parallel, quadruple-masked design with 1306 participants and two registered primary overall-survival measures. The primary formal analysis used a one-sided stratified log-rank test, with the hazard ratio and 95% confidence interval obtained from a Cox regression model stratified by extent of disease and alkaline phosphatase.

The primary OS result was HR 0.675, with a 95% CI of 0.568–0.801 and a one-sided p-value of <0.0001, based on 533 reported OS events at the 25 OCT 2021 cutoff. Seven additional formal secondary time-to-event analyses were also posted, with hazard ratios ranging from 0.357 to 1.043.

The statistically important lesson is that these numbers answer different questions. The hazard ratio describes relative event hazard, the confidence interval describes uncertainty around that estimate, and the p-value describes evidence under the specified hypothesis-testing framework. None alone represents an absolute treatment benefit for an individual participant.

Clinical Biostats methodology: A trial-results page should not merely repeat a headline result. The goal is to reconstruct the statistical structure of the trial—design, endpoint, estimand, analysis method, effect measure, uncertainty, testing framework, and limitations—while clearly separating reported evidence from educational interpretation.