← Clinical Trials
Ovarian Cancer Phase 3 Time-to-Event Analysis NCT04209855

MIRASOL: Complete Statistical Analysis of Mirvetuximab Soravtansine in Platinum-Resistant Ovarian Cancer

An independent statistical review of the randomized phase 3 MIRASOL trial comparing mirvetuximab soravtansine with investigator's choice chemotherapy in platinum-resistant, advanced high-grade epithelial ovarian, primary peritoneal, or fallopian tube cancers with high folate receptor-alpha (FRα) expression.

Trial status: COMPLETED  ·  Enrollment: 453  ·  Primary completion: March 6, 2023
Scope of this record

This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record. View NCT04209855 on ClinicalTrials.gov.

1. Trial at a Glance

MIRASOL was a randomized, open-label, parallel phase 3 trial evaluating mirvetuximab soravtansine against investigator's choice chemotherapy in patients with platinum-resistant, advanced high-grade epithelial ovarian, primary peritoneal, or fallopian tube cancers with high FRα expression. The registered primary endpoint was progression-free survival (PFS) assessed by RECIST Version 1.1.

453
Enrolled
Randomized trial
2
Arms
Parallel design
0.63
PFS HR
95% CI 0.513–0.785
<0.0001
P-value
Two-sided
FeatureMIRASOL
PhasePhase 3
PopulationPlatinum-resistant, advanced high-grade epithelial ovarian, primary peritoneal, or fallopian tube cancers with high FRα expression
DesignRandomized, parallel, open-label
AllocationRandomized
Primary endpointProgression-free Survival (PFS) as Assessed by Response Evaluation Criteria in Solid Tumors (RECIST) Version 1.1
Primary endpoint typeTime-to-event
Primary hypothesisSuperiority
Primary analysisLog-rank test
Effect measureHazard ratio from a Cox proportional-hazards analysis
Enrollment453
Lead sponsorAbbVie
Sponsor typeIndustry
Trial startDecember 31, 2019
Primary completionMarch 6, 2023
ClinicalTrials.govNCT04209855

2. Clinical Question

The central statistical question was whether mirvetuximab soravtansine would produce longer progression-free survival than investigator's choice chemotherapy in the specified platinum-resistant cancer population with high FRα expression.

Population

Patients with platinum-resistant, advanced high-grade epithelial ovarian, primary peritoneal, or fallopian tube cancers with high FRα expression.

Intervention

Mirvetuximab soravtansine.

Comparator

Investigator's Choice (IC) chemotherapy.

Primary question

Does mirvetuximab soravtansine improve progression-free survival relative to investigator's choice chemotherapy?

3. Trial Design

01
Randomize 453 participants
02
Assign 2 treatment strategies
03
Follow Time to PD or death
04
Analyze ITT population
05
Compare Log-rank + hazard ratio
Allocation
Randomized allocation was used to create the treatment comparison.
Design model
Parallel-group design with two arms.
Masking
None. The trial was open-label.
Primary purpose
Treatment.
ARM 1

Mirvetuximab Soravtansine

  • Mirvetuximab soravtansine
  • Randomized treatment arm
ARM 2

Investigator's Choice Chemotherapy

  • Investigator's Choice chemotherapy
  • Permitted chemotherapy interventions included paclitaxel, topotecan, and pegylated liposomal doxorubicin

The design is statistically important because randomization establishes the treatment groups as the principal basis for causal comparison. The absence of masking does not invalidate a randomized time-to-event comparison, but it is relevant when considering endpoints or clinical decisions that can be influenced by knowledge of treatment assignment. For the primary PFS analysis, the registry identifies an objective radiologic progression definition under RECIST Version 1.1 together with death as an event.

4. Endpoints

EndpointRegistry definition / time frameStatistical role
Progression-free Survival (PFS) as Assessed by Response Evaluation Criteria in Solid Tumors (RECIST) Version 1.1 From randomization until PD or death, whichever occurred first (up to approximately 36 months) Primary endpoint; time-to-event; superiority hypothesis

Registered PFS definition

PFS was defined as the time from randomization until progressive disease (PD) or death whichever occurred first. PD: At least a 20% increase in the sum of the longest diameters (SoD) of target lesion, taken as reference the smallest (nadir) SoD since and including baseline. In addition to the relative increase of 20%, the SoD must also demonstrate an absolute increase of at least 5 millimeters (mm). Unequivocal progression of non-target lesions and appearance of new lesions. Unequivocal progression should not normally trump target lesion status. It must be representative of overall disease status change, not a single lesion increase.

Why the endpoint definition matters: PFS is not simply a measure of whether a tumor eventually grew. It is a time-to-event endpoint in which the event is the first occurrence of either progression under the registered RECIST definition or death. The time origin is randomization, so the treatment comparison is anchored to the randomized allocation rather than to the date treatment was actually received.

5. Primary Results: Progression-Free Survival

The registry reports one formal statistical analysis for the primary endpoint. The analysis population was the intention-to-treat population, defined as all participants randomized to the study regardless of whether or not they received study treatment.

Hazard ratio for progression or death

0.63

95% CI: 0.513–0.785   ·   P < 0.0001   ·   Two-sided

Analysis: log-rank test; effect measure: Cox proportional hazard.

Primary endpointAnalysis populationMethodEffect estimate95% CIP-value
Progression-free Survival (PFS) ITT Log-rank Hazard ratio 0.63 0.513–0.785 <0.0001
Clinical Biostats interpretation

A hazard ratio of 0.63 means that, under the Cox proportional-hazards interpretation, the estimated instantaneous rate of progression or death in the mirvetuximab soravtansine group was approximately 63% of that in the investigator's choice chemotherapy group. Expressed as a relative complement, this corresponds to a 37% lower estimated hazard of progression or death.

The hazard ratio does not mean that 37% of patients avoided progression, that every patient experienced exactly a 37% reduction in risk, or that median PFS differed by a particular number of months. Those would require different statistics.

The 95% confidence interval of 0.513–0.785 describes statistical uncertainty around the estimated hazard ratio. It does not describe the range of effects experienced by individual patients. Because the entire interval is below 1, every value in this reported interval corresponds to a lower estimated hazard in the mirvetuximab soravtansine group relative to the comparator.

The P-value < 0.0001 addresses evidence against the null hypothesis used for the statistical comparison. It does not measure the size or clinical importance of the treatment effect. The effect size is described by the hazard ratio and its confidence interval; the p-value answers a different question about statistical evidence under the null framework.

Because the effect is expressed through a Cox proportional-hazards model, the usual interpretation assumes that a single hazard ratio is an appropriate summary of the relative hazards over the analyzed period. The ClinicalTrials.gov record does not provide sufficient information to independently assess the proportional-hazards assumption.

6. How to Read the Primary Result

The most informative way to read the MIRASOL primary result is to keep three statistical quantities separate: the hazard ratio, its confidence interval, and its p-value.

Hazard ratio: magnitude

The estimate of 0.63 is the principal relative-effect measure reported for PFS. Values below 1 favor the first-listed treatment in this comparison with respect to the event of progression or death.

Confidence interval: precision

The interval 0.513–0.785 shows the range of hazard-ratio values compatible with the reported statistical uncertainty under the analysis framework.

P-value: evidence

The value <0.0001 quantifies statistical evidence against the null hypothesis. It is not a measure of how large or clinically important the treatment effect is.

ITT population: analysis principle

All randomized participants were included in the primary efficacy population regardless of whether they received study treatment, preserving treatment assignment as the basis of comparison.

Hazard-ratio interpretation
HR = 0.63   →   estimated hazard in mirvetuximab group ≈ 63% of comparator hazard

The corresponding relative complement is 1 − 0.63 = 0.37, or a 37% lower estimated hazard. This is a direct interpretation of the reported hazard ratio; it is not an estimate of absolute risk reduction.

7. Statistical Methodology

Intention-to-treat analysis

The primary analysis used the ITT population. The registry defines this population as all participants randomized to the study, regardless of whether or not participants received study treatment. This is an important design principle because the comparison remains tied to the original randomization.

ITT analysis is particularly relevant in superiority trials because excluding randomized participants after allocation can introduce selection into the treatment comparison. Retaining randomized participants according to their assigned group helps preserve the comparability created by randomization.

Log-rank test

The reported primary statistical method was the log-rank test. The log-rank test is designed for comparing survival or time-to-event distributions between randomized groups while accounting for different follow-up times and right censoring.

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

The log-rank procedure compares the observed pattern of events with the pattern expected under the null hypothesis across event times. It is therefore fundamentally different from a simple comparison of proportions at one fixed time point.

Cox proportional-hazards model

The registry reports the effect measure as a Cox proportional hazard and the effect measure as a hazard ratio. The Cox model is commonly used to estimate a relative hazard while retaining censored observations in the time-to-event analysis.

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

For a two-group treatment comparison, exponentiating the treatment coefficient produces the hazard ratio. The registry's reported estimate of 0.63 therefore summarizes the relative hazard under the fitted Cox model.

Right censoring

Time-to-event analyses allow some participants to reach their last known assessment without having experienced the event by that time. Such participants are censored rather than treated as if the event had occurred. This lets their available follow-up contribute information without assigning an unobserved event date.

For PFS, the event definition is especially important because either progressive disease or death can end event-free follow-up. The registered time frame begins at randomization and continues until PD or death, whichever occurs first.

8. Statistical Methods Explained

Why was a log-rank test used?

PFS is a time-to-event endpoint, not merely a binary outcome. Participants can have different follow-up times, and some may be censored before experiencing progression or death. The log-rank test is designed for this setting because it compares event-time distributions while incorporating the available follow-up information.

What does a hazard ratio of 0.63 mean?

A hazard ratio of 0.63 indicates an estimated instantaneous event rate approximately 37% lower in the mirvetuximab soravtansine group relative to investigator's choice chemotherapy, under the Cox model. It is a relative time-to-event measure, not an absolute percentage of patients who avoid progression.

Why is the confidence interval important?

The point estimate alone gives only one estimate of the treatment effect. The 95% confidence interval of 0.513–0.785 shows the statistical uncertainty around that estimate. A narrower interval would indicate greater precision, whereas a wider interval would indicate less precision.

Why does the p-value not measure effect size?

The p-value and the hazard ratio answer different questions. The hazard ratio describes the estimated relative treatment effect. The p-value describes the degree of statistical evidence against the null hypothesis used by the test. A very small p-value can occur with a relatively modest effect when the data are highly informative, while a larger effect estimate can be imprecisely estimated in a smaller or more variable dataset.

Why does the ITT population matter?

Analyzing all randomized participants according to their randomized assignment preserves the principal comparison created by randomization. In MIRASOL, the registry explicitly defines the ITT population as all randomized participants regardless of whether they received study treatment.

What does censoring mean for PFS?

Censoring occurs when a participant's event-free follow-up ends without observing progression or death during the available observation period. The participant contributes information up to the censoring time. Censoring is therefore not the same thing as a PFS event.

Why is the proportional-hazards assumption relevant?

The Cox hazard ratio is most naturally interpreted as a common relative hazard over time. If the relative hazard changes substantially over time, a single hazard ratio can compress a more complicated time-dependent treatment effect into one summary number. The ClinicalTrials.gov record does not report an assessment of this assumption.

9. What the Confidence Interval Tells Us

Precision of the primary estimate

The reported hazard ratio is 0.63, with a two-sided 95% CI of 0.513–0.785. The interval is entirely below 1, so the full reported interval represents hazard-ratio values below the no-difference value of 1.

The lower endpoint, 0.513, corresponds to a stronger estimated relative reduction in hazard than the point estimate, while the upper endpoint, 0.785, corresponds to a smaller relative reduction. Thus the confidence interval communicates both the direction and the statistical precision of the estimated treatment effect.

QuantityReported valueHow to interpret it
Hazard ratio0.63Estimated relative hazard of progression or death under the Cox model
95% CI lower0.513Lower endpoint of the reported uncertainty interval
95% CI upper0.785Upper endpoint of the reported uncertainty interval
P-value<0.0001Statistical evidence from the reported comparison

A confidence interval should not be interpreted as saying that there is a 95% probability that the true hazard ratio lies between 0.513 and 0.785. In the frequentist framework, the interval is a property of the statistical procedure and the observed data. Its practical role here is to show the uncertainty associated with the reported point estimate.

10. Primary Endpoint Analysis Population

ITT definition from the registry: The ITT population included all participants randomized to the study, regardless of whether or not participants received study treatment (MIRV or IC Chemo).

This definition has an important statistical consequence. The primary comparison does not redefine the groups based on treatment exposure after randomization. Instead, the treatment groups remain the randomized groups. This protects the principal advantage of randomization: treatment assignment is determined independently of subsequent treatment receipt and other post-randomization events.

For a superiority analysis, this distinction is central. A comparison restricted only to participants who completed treatment could differ systematically from the randomized comparison because treatment completion can itself be related to prognosis, tolerability, disease progression, or other post-randomization factors.

11. Safety Results

The ClinicalTrials.gov record reports serious adverse events by randomized arm. The reported counts are the number affected divided by the number at risk for each arm.

Safety measureMirvetuximab SoravtansineInvestigator's Choice (IC) Chemotherapy
Serious adverse events62 / 22773 / 226
Serious adverse events: affected / at risk
Mirvetuximab Soravtansine
62/227
IC Chemotherapy
73/226

The safety information should be kept conceptually separate from the PFS analysis. The primary efficacy endpoint is a time-to-event comparison analyzed in the ITT population, whereas the ClinicalTrials.gov record is presented as affected participants over those at risk in each arm. The ClinicalTrials.gov record does not provide a formal statistical comparison of serious adverse-event rates, so no comparative p-value or effect estimate is presented for this safety outcome.

Statistical caution: 62/227 and 73/226 are descriptive safety counts. They should not be converted into an inferred hypothesis test or described as statistically different without a reported formal safety comparison and its prespecified analysis framework.

12. What Is and Is Not Reported in the Supplied Registry Data

The ClinicalTrials.gov record contains a formal statistical analysis for the primary PFS endpoint. They do not provide additional formal statistical analyses that can be reproduced here for other outcome measures.

Analysis elementSupplied dataInterpretation
Primary PFS analysisReportedLog-rank test with Cox proportional-hazard effect estimate
Hazard-ratio estimate0.63Reported for PFS
95% CI0.513–0.785Reported for PFS
P-value<0.0001Reported for PFS
Secondary endpoint formal analysesNot reportedNo additional comparative result is presented
Subgroup analysesNot reportedNo subgroup effect estimates are presented
Median PFSNot reportedNo median is reported here
Kaplan-Meier estimatesNot reportedNo time-specific survival probabilities are reported here
Interim-analysis detailsNot reportedNo interim alpha-spending or stopping rule is described here
Multiplicity procedureNot reportedNo multiplicity adjustment is described here
Missing-data or imputation procedureNot reportedNo imputation method is reported here
Stratification factorsNot reportedNo stratified analysis factors are identified here
Bayesian analysisNot reportedNo Bayesian method is reported

This distinction is important for an educational statistical record. A time-to-event trial may use several additional analytical components in its full protocol or statistical analysis plan, but they should not be attributed to MIRASOL unless they are documented in the ClinicalTrials.gov record.

13. A Closer Look at the Primary PFS Result

The primary result combines a very small p-value with a hazard ratio substantially below 1 and a confidence interval entirely below 1. Those three pieces of information are complementary rather than interchangeable.

Direction

The hazard ratio of 0.63 is below 1, indicating a lower estimated hazard of progression or death in the mirvetuximab soravtansine group relative to investigator's choice chemotherapy.

Magnitude

The point estimate corresponds to a 37% lower estimated hazard, calculated directly as the complement of the reported hazard ratio.

Precision

The 95% CI extends from 0.513 to 0.785, providing a range of plausible effect estimates under the reported statistical framework.

Evidence

The two-sided p-value is <0.0001, indicating strong statistical evidence against the null comparison used by the primary test.

It is also important not to confuse the hazard ratio with an absolute treatment effect. Without reported median PFS, fixed-time PFS probabilities, or the underlying event and censoring data, the ClinicalTrials.gov record does not support a calculation of an absolute difference in PFS duration or a number-needed-to-treat measure.

14. Why the Time Origin Matters

The registry defines PFS from randomization until PD or death. This time origin is statistically consequential. Randomization occurs before subsequent treatment exposure and before the future course of disease is known, making it a clean anchor for the randomized comparison.

If time were instead measured only among participants who began or completed treatment, the analysis population and time origin could become conditional on post-randomization events. Such conditioning can compromise the interpretation of the randomized comparison.

Registered PFS clock
Randomization  →  first PD or death  →  event time

If neither event is observed during available follow-up, the participant contributes follow-up information up to the applicable censoring point under the analysis framework.

15. The Role of RECIST in the Endpoint

The primary endpoint was not simply "time until treatment failure." It was specifically PFS as assessed by RECIST Version 1.1. The registry definition identifies progressive disease using a relative increase in the sum of the longest diameters of target lesions, together with an absolute increase requirement.

This makes endpoint definition part of the statistical design. A time-to-event analysis can only be interpreted correctly when the event itself is defined consistently. Here, the event is either RECIST-defined PD or death, whichever occurs first.

Educational distinction: the statistical method and the clinical event definition solve different problems. RECIST defines what counts as progression; the log-rank test and Cox model define how the resulting time-to-event observations are compared statistically.

16. Superiority Hypothesis

The registered hypothesis type was superiority. This means the statistical question was whether the treatment groups differed in the prespecified direction of interest rather than whether the new treatment stayed within a predefined non-inferiority margin.

Hypothesis featureMIRASOL
Hypothesis typeSuperiority
Primary endpointProgression-free survival
EventPD or death, whichever occurred first
Primary testLog-rank
Effect measureHazard ratio
Reported hazard ratio0.63
Reported two-sided 95% CI0.513–0.785
Reported p-value<0.0001

Because this is a superiority analysis, the central null reference for the hazard ratio is 1. A hazard ratio of 1 would correspond to equal hazards under the model. The reported estimate is below 1, and the entire reported confidence interval remains below 1.

17. Limitations

18. Why This Trial Matters Statistically

MIRASOL is a useful teaching example because it connects a randomized phase 3 treatment comparison directly to the core machinery of time-to-event analysis. The primary endpoint is defined from randomization, the event combines progression and death, the analysis uses a log-rank test, and the treatment effect is summarized using a Cox proportional-hazards hazard ratio.

Statistical conceptHow it appears in MIRASOL
RandomizationParticipants were randomized between two parallel treatment arms.
Intention-to-treat analysisThe primary analysis included all randomized participants regardless of whether study treatment was received.
Time-to-event endpointPFS was measured from randomization until PD or death.
RECIST endpoint definitionProgression was defined using the registered RECIST Version 1.1 criteria.
Log-rank testUsed as the reported primary statistical comparison.
Hazard ratioReported Cox proportional-hazard estimate of 0.63.
Confidence intervalTwo-sided 95% CI of 0.513–0.785.
P-valueTwo-sided p-value <0.0001.
CensoringRelevant because PFS is a time-to-event endpoint with follow-up potentially ending before an event.
Superiority testingThe registered hypothesis type was superiority.
Open-label designMasking was listed as none.
Safety analysisSerious adverse events are reported descriptively by arm.

19. Statistical Concepts in This Trial

Learn more about the methods used in this trial:

20. Related Statistical Calculators

21. Sources

Continue with the statistical methods behind MIRASOL

Explore the underlying survival-analysis concepts, then apply them with Clinical Biostats statistical calculators.

22. Record Summary

MIRASOL is a randomized phase 3 superiority trial with a primary time-to-event endpoint: progression-free survival from randomization until RECIST Version 1.1-defined progressive disease or death, whichever occurred first. The primary analysis used the ITT population and a log-rank test, with a Cox proportional-hazard effect measure.

The reported hazard ratio of 0.63 corresponds to a 37% lower estimated hazard of progression or death in the mirvetuximab soravtansine group relative to investigator's choice chemotherapy. The two-sided 95% confidence interval of 0.513–0.785 describes the statistical uncertainty around that estimate, while the reported p-value <0.0001 describes the statistical evidence against the null hypothesis. These quantities should be interpreted together rather than treating the p-value as a measure of treatment magnitude.

The ClinicalTrials.gov record also report serious adverse events of 62/227 in the mirvetuximab soravtansine arm and 73/226 in the investigator's choice chemotherapy arm. No formal comparative safety analysis is reported here. Likewise, the ClinicalTrials.gov record does not provide median PFS, time-specific Kaplan-Meier estimates, subgroup effects, multiplicity procedures, interim-analysis details, or missing-data methods, so those elements are not inferred.

Clinical Biostats methodology: A trial-results page should distinguish the registered endpoint definition, analysis population, statistical test, effect estimate, confidence interval, and p-value from interpretations that require additional assumptions or data. MIRASOL illustrates how a randomized time-to-event result can be explained without treating a hazard ratio, confidence interval, and p-value as interchangeable quantities.