This page separates reported trial results from statistical interpretation. The numerical results on this page are restricted to the statistical analyses and trial information from ClinicalTrials.gov. ClinicalTrials.gov provides the official trial registry record.
1. Trial at a Glance
ARIEL3 was a randomized, double-blind, parallel phase 3 study of rucaparib versus placebo as switch maintenance after platinum in relapsed high-grade serous or endometrioid ovarian cancer. The registry lists ovarian cancer, fallopian tube cancer, and peritoneal cancer as the study conditions.
| Feature | ARIEL3 |
|---|---|
| Phase | Phase 3 |
| Brief title | Phase 3 Study of Rucaparib as Switch Maintenance After Platinum in Relapsed High Grade Serous or Endometrioid Ovarian Cancer (ARIEL3) |
| Conditions | Ovarian Cancer; Fallopian Tube Cancer; Peritoneal Cancer |
| Design | Randomized, parallel, double-blind |
| Primary purpose | Treatment |
| Enrollment | 564 |
| Interventions | Rucaparib 600 mg Tablets; Placebo Tablets |
| Primary endpoint type | Time-to-event |
| Statistical analyses posted | 2 |
| ClinicalTrials.gov | NCT01968213 |
2. Clinical Question
The central statistical question was whether rucaparib used as switch maintenance after platinum improved investigator-assessed progression-free survival compared with placebo in the trial population.
Population
Patients enrolled in the phase 3 study of switch maintenance after platinum for relapsed high-grade serous or endometrioid ovarian cancer. The registry also lists fallopian tube cancer and peritoneal cancer among the study conditions.
Intervention
Rucaparib 600 mg Tablets.
Comparator
Placebo Tablets.
Primary question
Does rucaparib produce a different time-to-progression-or-death experience from placebo, with superiority evaluated using a Cox proportional-hazards model?
3. Trial Design
Rucaparib
- Rucaparib 600 mg Tablets
- Switch-maintenance intervention after platinum
- Compared with placebo in a randomized, double-blind design
Placebo
- Placebo Tablets
- Control condition for the switch-maintenance comparison
- Compared with rucaparib in the same randomized trial
4. Trial Timeline
Study start
The registry lists April 7, 2014 as the study start date.
Primary completion
The registry lists April 1, 2017 as the primary completion date.
Completed
The trial is listed as completed, with results posted on ClinicalTrials.gov.
5. Primary Endpoint
| Endpoint | Registry definition / assessment | Analysis |
|---|---|---|
| Disease Progression According to RECIST Version 1.1, as Assessed by the Investigator, or Death From Any Cause (Investigator Progression Free Survival as Per invPFS) | Investigator progression-free survival is defined as the time from randomization to disease progression according to RECIST v1.1 criteria as assessed by the investigator, or death due to any cause, whichever occurs first. Assessments were performed every 12 calendar weeks, with within 7 days prior permitted, after start of treatment until treatment discontinuation due to the registry-specified reason. | Intent-to-treat population; Cox proportional-hazards model; superiority hypothesis |
This endpoint combines two possible first events: investigator-assessed disease progression or death from any cause. Statistically, it is therefore a time-to-event endpoint, not simply a binary response measure at a fixed date.
6. Secondary Endpoint
| Endpoint | Registry definition / assessment | Analysis |
|---|---|---|
| Disease Progression According to RECIST v1.1, as Assessed by Independent Radiology Review (IRR), or Death From Any Cause (irrPFS) | Independent-radiology-review progression-free survival using disease progression according to RECIST v1.1 or death from any cause. Assessments were performed every 12 calendar weeks, with within 7 days prior permitted, after start of treatment until treatment discontinuation due to the registry-specified reason. | Intent-to-treat population; Cox proportional-hazards model; superiority hypothesis |
The two reported efficacy analyses therefore address related versions of the progression-free-survival question: one based on investigator assessment and one based on independent radiology review.
7. Statistical Methodology
Intention-to-treat analysis
The registry specifies Intent-to-treat: All patients randomized as the analysis population for both posted statistical analyses. This means the efficacy comparison is anchored to the treatment assignment produced by randomization rather than being restricted to patients who remained on treatment.
The ITT principle preserves the treatment comparison created by randomization. For a time-to-event endpoint, it also means that discontinuation of treatment does not automatically remove a patient from the primary efficacy population.
Cox proportional-hazards model
The registry reports Regression, Cox as the statistical method, normalized here as a Cox proportional-hazards model. This is a standard survival-analysis model for estimating a relative treatment effect when follow-up times and event times differ across patients.
For a two-group treatment comparison, the hazard ratio is represented by the exponential of the treatment coefficient. The model compares the estimated instantaneous event rates between treatment groups over time.
Stratified analysis
For both the primary and secondary analyses, the registry states that the analysis was performed by randomization strata of HRD classification by CTA, best response, and penultimate platinum progression-free interval. The statistical comparison therefore was not described as an unstratified Cox analysis.
Hazard ratio as the effect measure
The reported effect measure is a hazard ratio. A hazard ratio below 1 indicates a lower estimated instantaneous event rate in the rucaparib group relative to the placebo group under the fitted model.
Primary effect estimate
95% two-sided CI: 0.295–0.451
P < 0.0001
Time-to-event structure
Progression-free survival differs from an ordinary continuous outcome because some patients may not experience progression or death during observed follow-up. Those patients contribute information until their censoring time. The analysis therefore needs methods that account for both event times and censoring rather than treating every patient as if an event had occurred.
8. Primary Result: Investigator Progression-Free Survival
The posted primary analysis evaluated investigator-assessed progression-free survival in the intent-to-treat population, comparing rucaparib 600 mg Tablets with placebo Tablets using a Cox proportional-hazards model.
Investigator progression-free survival
95% two-sided CI: 0.295–0.451 · P < 0.0001
Analysis population: Intent-to-treat, all patients randomized
| Element | Reported value |
|---|---|
| Endpoint | Investigator progression-free survival (invPFS) |
| Groups | Rucaparib 600 mg Tablets vs Placebo Tablets |
| Endpoint type | Time-to-event |
| Analysis population | Intent-to-treat: All patients randomized |
| Method | Cox proportional-hazards model |
| Effect measure | Hazard ratio |
| Estimate | 0.365 |
| 95% CI | 0.295–0.451 |
| P-value | <0.0001 |
| Hypothesis | Superiority |
The hazard ratio of 0.365 means that, under the fitted Cox model, the estimated instantaneous rate of investigator-assessed progression or death in the rucaparib group was approximately 36.5% of that in the placebo group. Equivalently, the point estimate corresponds to an estimated 63.5% lower hazard relative to placebo.
The hazard ratio does not mean that 63.5% of patients avoided progression, that every patient experienced the same reduction, or that the probability of progression was reduced by exactly 63.5%. A hazard ratio is a relative time-to-event measure, not an absolute risk difference.
The 95% two-sided confidence interval of 0.295–0.451 describes statistical uncertainty around the estimated hazard ratio under the model and sampling framework. Its width provides information about precision; it does not describe the range of effects that individual patients could experience.
The p-value of <0.0001 addresses evidence against the null hypothesis under the specified testing framework. It does not measure the size of the treatment effect, the probability that the treatment works, or the clinical importance of the effect.
Because the analysis uses a Cox proportional-hazards model, the usual proportional-hazards interpretation should be kept in mind. A single hazard ratio is most straightforward when the relative hazards are reasonably represented by the model over the relevant follow-up. The ClinicalTrials.gov record does not provide additional diagnostics for that assumption.
The analysis is based on all randomized patients, which is important for preserving the randomized comparison. The reported analysis was stratified by HRD classification by CTA, best response, and penultimate platinum progression-free interval, so the estimate should be interpreted as the result of that specified stratified analysis rather than as an unadjusted comparison.
9. What Does HR 0.365 Mean?
The most important statistical distinction is between a hazard ratio and a probability or percentage of patients experiencing an event.
What it does mean
Under the fitted Cox model, the estimated instantaneous rate of investigator-assessed progression or death was lower in the rucaparib group, with a hazard ratio of 0.365.
What it does not mean
It does not mean that 36.5% of patients progressed, that 63.5% were protected from progression, or that every patient's individual risk was reduced by the same percentage.
Relative versus absolute effect
The hazard ratio is relative. Absolute survival probabilities or median times would answer different questions and are not reported in the ClinicalTrials.gov record used for this page.
Model-based estimate
The value is produced by a Cox proportional-hazards model. It is therefore a model-based summary of the time-to-event comparison rather than a simple ratio of two raw event percentages.
10. Confidence Interval and Precision
The primary hazard-ratio estimate was 0.365, with a 95% two-sided confidence interval of 0.295–0.451.
The interval lies below 1.0, which is the conventional null value for a hazard ratio. Within the stated statistical framework, the interval is therefore consistent with a lower hazard in the rucaparib group across the values represented by the interval.
The point estimate alone does not communicate how precisely the treatment effect was estimated. The interval from 0.295 to 0.451 gives a range of values compatible with the statistical uncertainty represented by the 95% confidence procedure. It should not be interpreted as saying that there is a 95% probability that the true hazard ratio lies inside this particular interval.
11. Secondary Result: Independent Radiology Review PFS
The registry also reports a secondary time-to-event analysis based on progression assessed by independent radiology review or death from any cause.
Independent radiology review PFS
95% two-sided CI: 0.278–0.450 · P < 0.0001
Analysis population: Intent-to-treat, all patients randomized
| Element | Reported value |
|---|---|
| Endpoint | Independent radiology review progression-free survival (irrPFS) |
| Groups | Rucaparib 600 mg Tablets vs Placebo Tablets |
| Endpoint type | Time-to-event |
| Analysis population | Intent-to-treat: All patients randomized |
| Method | Cox proportional-hazards model |
| Effect measure | Hazard ratio |
| Estimate | 0.354 |
| 95% CI | 0.278–0.450 |
| P-value | <0.0001 |
| Hypothesis | Superiority |
The independent-review estimate is close to the investigator-assessed estimate: 0.354 versus 0.365. The two analyses are not identical because the source of progression assessment differs, but both use the same general time-to-event framework and both were analyzed in the ITT population.
The secondary hazard ratio of 0.354 corresponds to an estimated instantaneous rate of progression or death approximately 35.4% of the placebo-group rate under the fitted model, or approximately a 64.6% lower estimated hazard relative to placebo.
The 95% two-sided confidence interval of 0.278–0.450 quantifies uncertainty around that estimate. As with the primary result, the confidence interval does not describe individual patient outcomes, and the p-value of <0.0001 does not measure the magnitude or clinical importance of the effect.
The similarity between the investigator and independent-review estimates is descriptively informative because the endpoint definitions use different assessment sources. It should not, however, be converted into an additional claim about agreement beyond what the reported estimates themselves show.
12. Comparing the Two PFS Analyses
| Feature | Investigator PFS | Independent Review PFS |
|---|---|---|
| Endpoint | invPFS | irrPFS |
| Progression assessment | Investigator | Independent radiology review |
| Event definition | Progression according to RECIST v1.1 or death | Progression according to RECIST v1.1 or death |
| Analysis population | Intent-to-treat | Intent-to-treat |
| Model | Cox proportional-hazards | Cox proportional-hazards |
| Hazard ratio | 0.365 | 0.354 |
| 95% two-sided CI | 0.295–0.451 | 0.278–0.450 |
| P-value | <0.0001 | <0.0001 |
The estimates are close in magnitude, with both point estimates substantially below 1. The important methodological distinction is not simply the numerical difference between 0.365 and 0.354; it is that the progression event was assessed by different sources. Independent review can provide a distinct assessment framework from investigator determination.
13. Statistical Methods Explained
Why was a Cox proportional-hazards model used?
Progression-free survival is a time-to-event endpoint. Patients can have different follow-up durations, and some can be censored before progression or death. A Cox model is designed to compare event rates over time while incorporating the timing of events rather than reducing the outcome to a single fixed-time binary result.
What does a hazard ratio of 0.365 mean?
A hazard ratio of 0.365 means that the fitted model estimates the instantaneous rate of progression or death in the rucaparib group at about 36.5% of the corresponding rate in the placebo group. The complementary interpretation is an estimated 63.5% lower hazard. It is not a 63.5-percentage-point change in the probability of progression.
Why is the confidence interval important?
The estimate 0.365 is a single point on a range of statistically compatible values. The 95% two-sided confidence interval of 0.295–0.451 communicates the uncertainty around the estimate. A narrower interval generally provides greater precision than a wider interval, although precision and clinical importance are separate concepts.
Why does the p-value not measure effect size?
The p-value reflects how compatible the observed data are with the null hypothesis under the specified testing framework. It does not tell us how large the hazard ratio is. The magnitude is conveyed by the hazard ratio, while the confidence interval provides information about uncertainty around that magnitude.
Why does the ITT population matter?
The registry defines the analysis population as all randomized patients. An ITT analysis retains patients according to their randomized assignment, preserving the comparison created by randomization. This is especially important when treatment exposure or discontinuation differs after randomization.
Why are there both investigator and independent-review PFS analyses?
The two endpoints use different sources for progression assessment. Comparing their estimates can provide a useful descriptive check of how the treatment effect looks under different assessment frameworks. They remain distinct analyses and should not be treated as duplicate measurements of exactly the same statistical endpoint.
14. Stratification in the Cox Analysis
The registry states that the primary and secondary analyses were performed by randomization strata of HRD classification by CTA, best response, and penultimate platinum progression-free interval.
Included as a randomization-stratification factor in the reported analysis.
Included as a randomization-stratification factor in the reported analysis.
Included as a randomization-stratification factor in the reported analysis.
The treatment effect is estimated while accounting for the specified randomization strata.
Stratification is particularly relevant in a randomized trial because the factors used to stratify randomization are incorporated into the analysis structure. The reported hazard ratio should therefore be understood as coming from the specified stratified model.
15. Safety: Serious Adverse Events
The ClinicalTrials.gov record reports serious adverse events by treatment arm using affected patients over patients at risk.
| Treatment arm | Serious adverse events | Interpretation of registry-reported count |
|---|---|---|
| Rucaparib 600 mg Tablets | 91 / 372 | 91 affected patients among 372 at risk |
| Placebo Tablets | 20 / 189 | 20 affected patients among 189 at risk |
The ClinicalTrials.gov record supports reporting these serious-adverse-event counts, but they do not provide a formal statistical comparison, confidence interval, p-value, event definition beyond the serious-adverse-event classification, or a complete safety-event table. Accordingly, this page does not calculate or infer an arm-level comparative safety effect from the counts.
16. What the Registry Results Do and Do Not Establish
Supported by the posted analysis
The primary investigator-assessed PFS analysis produced a hazard ratio of 0.365 with a 95% two-sided confidence interval of 0.295–0.451 and P < 0.0001 in the ITT population.
Also reported
The independent-review PFS analysis produced a hazard ratio of 0.354 with a 95% two-sided confidence interval of 0.278–0.450 and P < 0.0001.
Not represented by the hazard ratio
The hazard ratio is not an absolute probability, median survival time, response rate, or percentage of patients who benefit.
Safety is separate
Serious adverse events are reported as 91/372 for rucaparib and 20/189 for placebo, but the ClinicalTrials.gov record does not include a formal comparative safety analysis.
17. Multiplicity, Interim Analysis, and Other Design Topics
The ClinicalTrials.gov record identifies a superiority hypothesis and report two statistical analyses, one primary and one secondary. They do not provide information about an alpha-spending strategy, interim efficacy boundaries, multiplicity adjustment, a non-inferiority margin, crossover, factorial design, or Bayesian methods.
| Design topic | What is supported by the ClinicalTrials.gov record |
|---|---|
| Superiority | Yes. Both posted statistical analyses are identified as superiority analyses. |
| Non-inferiority margin | Not part of the ClinicalTrials.gov record. |
| Crossover | Not reported in the ClinicalTrials.gov record. |
| Factorial design | No. The registry describes a parallel design with 2 arms. |
| Interim analysis | Not reported in the ClinicalTrials.gov record. |
| Missing-data / imputation method | Not reported in the ClinicalTrials.gov record. |
| Multiplicity adjustment | Not reported in the ClinicalTrials.gov record. |
| Bayesian methods | Not reported; the posted method is a Cox regression analysis. |
This distinction is important because the absence of a reported design feature in the ClinicalTrials.gov record is not evidence that the feature was absent from the complete protocol or statistical analysis plan. It simply means that the feature cannot be characterized from the information used for this page.
18. Understanding the P-Values
The investigator-assessed PFS analysis reports P < 0.0001. This indicates strong statistical evidence against the null hypothesis under the reported superiority testing framework. It does not quantify the size of the treatment effect; that is the role of the hazard ratio and its confidence interval.
The independent-review PFS analysis also reports P < 0.0001. Again, the p-value should be read together with the hazard ratio of 0.354 and the 95% confidence interval of 0.278–0.450.
Because the ClinicalTrials.gov record does not describe the complete multiplicity strategy, the most defensible interpretation is to preserve the registry's stated endpoint roles and hypothesis types rather than infer a broader familywise-error framework.
19. A Worked Statistical Reading of the Primary Result
Suppose a reader encounters the registry entry as follows: HR 0.365, 95% CI 0.295–0.451, P < 0.0001. A disciplined interpretation proceeds in several steps.
- Identify the endpoint. The outcome is investigator-assessed progression-free survival, defined as time from randomization to RECIST v1.1 progression or death from any cause.
- Identify the population. The analysis is ITT: all patients randomized.
- Identify the comparison. Rucaparib 600 mg Tablets are compared with placebo Tablets.
- Identify the model. The analysis uses a Cox proportional-hazards model, with analysis performed by the specified randomization strata.
- Interpret the point estimate. HR 0.365 corresponds to a modeled instantaneous event rate approximately 63.5% lower in relative terms for rucaparib versus placebo.
- Interpret precision. The 95% CI is 0.295–0.451, showing the uncertainty represented by the confidence procedure.
- Interpret the p-value. P < 0.0001 indicates strong evidence against the null hypothesis under the specified testing framework, but does not measure effect size.
20. Limitations
- Registry-level information: This page is constrained to the ClinicalTrials.gov record and does not reconstruct analyses that are not represented in that data.
- Limited endpoint results: The statistical analyses posted on ClinicalTrials.gov contain one primary endpoint and one secondary endpoint. Other potential clinical outcomes are not characterized here unless explicitly reported in the ClinicalTrials.gov record.
- No median time estimates: The ClinicalTrials.gov record provides hazard ratios and confidence intervals but do not provide median progression-free survival values.
- No survival probabilities: The ClinicalTrials.gov record does not provide Kaplan-Meier survival probabilities at specific time points.
- No subgroup estimates: Although stratification factors are identified, the ClinicalTrials.gov record does not provide subgroup-specific hazard ratios or interaction tests.
- No formal safety comparison: Serious adverse-event counts are provided, but no formal statistical comparison or confidence interval is posted on ClinicalTrials.gov for the safety endpoint.
- Proportional-hazards assumption: Cox-model interpretation relies on the model's assumptions. The ClinicalTrials.gov record does not provide diagnostics assessing proportional hazards.
- Multiplicity: The ClinicalTrials.gov record does not describe the complete multiplicity-control strategy, so additional conclusions about familywise error should not be inferred.
- Missing-data methods: No imputation or censoring rules beyond the endpoint description reported in the registry are characterized here.
21. Why This Trial Matters Statistically
ARIEL3 is a useful teaching example because its posted results illustrate how a modern randomized clinical trial can combine an ITT population, a time-to-event endpoint, stratified Cox regression, a hazard-ratio effect measure, and both investigator and independent-review assessments of progression.
| Concept | How it appears in ARIEL3 |
|---|---|
| Randomization | Randomized parallel-group phase 3 design with 2 arms. |
| Double blinding | The registry identifies the study as double-blind. |
| ITT analysis | All randomized patients are included in the reported efficacy analysis population. |
| Time-to-event endpoint | Investigator PFS and independent-review PFS are defined as time from randomization to progression or death. |
| Cox regression | Both posted statistical analyses use Cox regression. |
| Hazard ratio | The primary estimate is 0.365; the secondary estimate is 0.354. |
| Confidence interval | Both analyses provide two-sided 95% confidence intervals around the hazard ratio. |
| Superiority testing | Both posted analyses are identified as superiority hypotheses. |
| Stratification | Analyses are performed by HRD classification by CTA, best response, and penultimate platinum progression-free interval. |
| Independent review | A secondary endpoint uses independent radiology review for progression assessment. |
22. Related Statistical Learning Pathway
Learn more about the methods used in this trial:
23. Related Statistical Calculators
24. Sources
- ClinicalTrials.gov: ARIEL3, NCT01968213.
- PubMed: PubMed record — PMID 37262961.
- PubMed: PubMed record — PMID 35397664.
- PubMed: PubMed record — PMID 35170751.
- PubMed: PubMed record — PMID 32861537.
- PubMed: PubMed record — PMID 32359490.
Continue learning from the statistical methods in ARIEL3
Explore survival-analysis tutorials, statistical calculators, and other clinical-trial results to deepen the concepts used in this analysis.
25. Record Summary
ARIEL3 provides a clear example of randomized time-to-event analysis. The primary endpoint was investigator-assessed progression-free survival, analyzed in all randomized patients with a Cox proportional-hazards model according to the trial's randomization strata. The reported hazard ratio was 0.365, with a 95% two-sided confidence interval of 0.295–0.451 and P < 0.0001. A secondary independent-radiology-review PFS analysis produced a hazard ratio of 0.354, with a 95% two-sided confidence interval of 0.278–0.450 and P < 0.0001.
The statistical lesson is broader than either individual number. The endpoint is time-to-event, so censoring and event timing matter. The Cox model expresses the treatment comparison as a hazard ratio rather than an absolute risk difference. The confidence interval communicates precision, while the p-value addresses compatibility with the null hypothesis under the stated testing framework. Finally, the distinction between investigator and independent-review progression assessment illustrates how the definition and measurement of an endpoint are part of the statistical analysis itself.