← Clinical Trials
Ovarian Cancer Phase 3 Randomized NCT01281254

TRINOVA-2: Complete Statistical Analysis of AMG 386 (Trebananib) in Ovarian Cancer

An independent statistical review of the randomized, double-blind, phase 3 TRINOVA-2 trial evaluating AMG 386 (trebananib) plus pegylated liposomal doxorubicin (PLD) versus placebo plus PLD in ovarian cancer and related gynecologic cancers.

TRINOVA-2  ·  NCT01281254  ·  Phase 3  ·  Enrollment 223
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.

1. Trial at a Glance

TRINOVA-2 was a randomized, double-blind, parallel-design phase 3 trial evaluating AMG 386 plus PLD against placebo plus PLD in participants with fallopian tube cancer, ovarian cancer, or primary peritoneal cancer.

223
Enrollment
Randomized trial
2
Treatment arms
Parallel design
3
Trial phase
Phase 3
2011–2014
Trial period
Start to primary completion
FeatureTRINOVA-2
PhasePhase 3
StatusTerminated
ConditionsFallopian Tube Cancer; Ovarian Cancer; Primary Peritoneal Cancer
DesignRandomized, double-blind, parallel
Primary purposeTreatment
Enrollment223
Number of arms2
Lead sponsorAmgen
Sponsor typeIndustry
ClinicalTrials.govNCT01281254

2. Clinical Question

The central question was whether AMG 386 plus PLD was superior to placebo plus PLD with respect to progression-free survival in participants with ovarian cancer and related gynecologic cancers represented in the registry.

Population

Participants with fallopian tube cancer, ovarian cancer, or primary peritoneal cancer enrolled in the phase 3 TRINOVA-2 study.

Intervention

AMG 386 plus PLD.

Comparator

Placebo plus PLD.

Primary question

Is AMG 386 plus PLD superior to placebo plus PLD as measured by progression-free survival?

3. Trial Design

01
Enroll223 participants
02
Randomize2 treatment arms
03
Double-blindAMG 386 or placebo
04
CombineBoth arms receive PLD
05
Assess PFSRadiologic progression
Allocation
Randomized allocation to one of two parallel treatment arms.
Masking
Double-blind design.
Primary purpose
Treatment.
Study period
Start date: April 18, 2011. Primary completion date: August 29, 2014.
ARM 1

AMG 386 plus PLD

  • AMG 386
  • Pegylated liposomal doxorubicin (PLD)
ARM 2

Placebo plus PLD

  • Placebo
  • Pegylated liposomal doxorubicin (PLD)

The parallel design makes the randomized treatment assignment the principal comparison between the two groups. Because both arms contain PLD, the contrast is specifically the addition of AMG 386 rather than a comparison of completely different treatment backbones.

4. Primary Endpoint

EndpointRegistry definitionAssessment schedule
Progression-free survival To determine if AMG 386 plus PLD is superior to placebo plus PLD as measured by progression-free survival, defined as the time from randomization to the earliest of the dates of first radiologic disease progression per RECIST 1.1 with modifications Radiological imaging will be performed 8 weeks ± 1 week, starting from date of randomization for the first 64 weeks, then every 16 weeks ± 1 week for the next 32 weeks, and then every 24 weeks ± 4 weeks thereafter.

The endpoint is a time-to-event outcome. Its clock begins at randomization, and the event is defined through radiologic disease progression using RECIST 1.1 with modifications. The registry therefore defines PFS around the occurrence of a specified disease-progression event rather than around a simple binary assessment at a single follow-up visit.

Assessment timing matters. The registry specifies increasingly spaced radiological assessments over follow-up: every 8 weeks ± 1 week for the first 64 weeks, every 16 weeks ± 1 week for the next 32 weeks, and every 24 weeks ± 4 weeks thereafter. For a time-to-progression endpoint, the observation schedule is part of how the event process becomes measurable.

5. Planned Analysis

The ClinicalTrials.gov record identifies progression-free survival as the primary endpoint and defines it as the time from randomization to the earliest date of first radiologic disease progression per RECIST 1.1 with modifications. The registry does not post a formal statistical analysis for this endpoint, and it does not provide an estimated treatment effect, confidence interval, or p-value.

For a randomized trial with a time-to-event primary endpoint such as this one, a conventional analysis would estimate the distribution of progression-free survival in each randomized group using the Kaplan-Meier method. A treatment comparison would commonly be performed using a log-rank test, with a Cox proportional-hazards model commonly used to estimate a hazard ratio and its confidence interval when the proportional-hazards framework is appropriate.

The resulting analysis would distinguish three related but different quantities: the estimated progression-free survival function over time, a relative comparison such as a hazard ratio, and an absolute time-specific difference between treatment groups. These measures describe different aspects of the same time-to-event process and should not be treated as interchangeable.

Conceptual time-to-event framework
S(t) = P(T > t)

Here, S(t) represents the probability of remaining free of the defined event beyond time t. In a clinical trial, patients who have not experienced the event by their last evaluable observation may contribute censored follow-up rather than being treated as having experienced progression at that time.

Registry reporting status: The ClinicalTrials.gov record does not post statistical analyses for the primary endpoint. Accordingly, there is no reported PFS hazard ratio, confidence interval, p-value, median PFS, or other numerical efficacy estimate in the registry record.

6. Statistical Methodology

Kaplan-Meier estimation

Progression-free survival is naturally analyzed with methods that account for differing follow-up times and right censoring. The Kaplan-Meier estimator provides an estimate of the probability of remaining event-free over time without requiring every participant to experience the event.

Kaplan-Meier estimator
Ŝ(t) = ∏ti ≤ t (1 − di/ni)

At each observed event time, di represents the number of events and ni represents the number of participants at risk immediately beforehand.

For TRINOVA-2, the radiological assessment schedule means that progression is observed through periodic imaging rather than continuously. The exact timing of a radiologic progression event therefore depends on the protocol's assessment process and the registry's endpoint definition.

Log-rank comparison

A log-rank test provides a way to compare two time-to-event distributions across follow-up. Conceptually, it compares the observed number of events in each treatment group with the number expected under the hypothesis that the groups have the same underlying event experience.

The log-rank framework is particularly useful when the question concerns whether the treatment groups differ across the follow-up period rather than at one prespecified time point.

Cox proportional-hazards model

A Cox model can summarize the relative event rate between treatment groups using a hazard ratio. In a simple two-group formulation, the treatment coefficient is exponentiated to obtain the estimated hazard ratio.

Interpretation of a hazard ratio
HR = exp(βtreatment)

A hazard ratio below 1 would correspond to a lower estimated instantaneous event rate in the AMG 386 plus PLD group relative to the placebo plus PLD group, while a value above 1 would correspond to a higher estimated event rate. The hazard ratio is not itself a probability of progression-free survival.

Censoring

Time-to-event analysis must distinguish an observed progression event from a participant whose progression status is not observed after a particular point in follow-up. Such participants can be censored according to prespecified rules. Correct handling of censoring is important because treating every censored participant as if an event occurred at the last observation would distort the estimated survival distribution.

Radiologic endpoint assessment

The registry defines the primary endpoint using radiologic disease progression per RECIST 1.1 with modifications. This means that the endpoint is tied to a structured tumor-assessment framework rather than an investigator's unrestricted clinical impression alone.

7. Statistical Methods Explained

Why is progression-free survival a time-to-event endpoint?

Progression-free survival incorporates both whether progression occurs and when it occurs. Two participants may both eventually experience progression, but the participant who progresses later has a different PFS outcome from the participant who progresses earlier. Time-to-event methods preserve that temporal information.

Why use Kaplan-Meier rather than simply compare the proportion progressed?

A simple proportion requires a fixed observation time and can discard information from participants followed for different lengths of time. Kaplan-Meier estimation instead uses the available follow-up while accounting for censoring, allowing the estimated event-free probability to be described across time.

What would a hazard ratio represent?

A hazard ratio summarizes the relative instantaneous event rate between the treatment groups under a fitted survival model. For example, an HR of 0.70 would be interpreted as a 30% lower estimated hazard under the model. It would not mean that 30% fewer participants necessarily experienced progression, nor would it imply that every participant had the same proportional reduction in risk.

Why does the confidence interval matter?

A point estimate alone does not communicate how precisely the treatment effect has been estimated. A confidence interval describes uncertainty around the estimated parameter under the statistical model and sampling framework. A wide interval indicates greater uncertainty than a narrow interval, even when the point estimates are identical.

Why is the imaging schedule statistically important?

Progression is detected through radiological assessment, and the registry specifies when those assessments occur. If progression is discovered at an imaging visit, the recorded event timing is connected to that assessment process. The prespecified imaging schedule therefore forms part of the measurement system for the PFS endpoint.

What is the difference between PFS and a response rate?

PFS measures time until the defined progression event or other event incorporated into the endpoint definition. A response rate is generally a proportion of participants achieving a specified tumor-response category. PFS therefore contains a temporal component that a single response proportion does not.

8. Reading the Primary Endpoint Correctly

Time origin

The registry defines PFS from the date of randomization. Randomization therefore establishes the starting point for the primary time-to-event measurement.

Event definition

The event is based on the earliest date of first radiologic disease progression according to RECIST 1.1 with modifications.

Repeated assessment

Radiological imaging is scheduled repeatedly over follow-up rather than being performed only once.

Treatment comparison

The registry frames the primary question as superiority of AMG 386 plus PLD over placebo plus PLD.

This structure is important because a time-to-progression analysis is not equivalent to comparing the number of participants who have progressed by the end of the study. The statistical analysis uses the timing of events and the available follow-up information to characterize the treatment groups.

9. Trial Timeline

April 18, 2011

Study start

The ClinicalTrials.gov record lists April 18, 2011 as the trial start date.

Phase 3

Randomized treatment comparison

The study used randomized allocation, a parallel design, double masking, and two treatment arms.

August 29, 2014

Primary completion

The ClinicalTrials.gov record lists August 29, 2014 as the primary completion date.

Current registry status

Terminated

The ClinicalTrials.gov record lists the study status as terminated.

10. What the Registry Does and Does Not Establish Statistically

The registry establishes the design framework and the definition of the primary endpoint, including the randomization structure, masking, treatment arms, enrollment, and radiological assessment schedule. These details are sufficient to identify the principal statistical problem: comparison of two randomized groups on a radiologically defined time-to-event endpoint.

The registry does not post statistical analyses for TRINOVA-2. Consequently, the registry record does not provide a numerical estimate of the treatment effect on PFS or its associated statistical uncertainty.

Statistical quantityRegistry status
Primary endpointProgression-free survival
Endpoint time originRandomization
Progression definitionFirst radiologic disease progression per RECIST 1.1 with modifications
PFS hazard ratioNot reported in posted statistical analyses
PFS confidence intervalNot reported in posted statistical analyses
PFS p-valueNot reported in posted statistical analyses
Median PFSNot reported in posted statistical analyses

This distinction is central to reading a registry record. A clearly specified endpoint can tell the reader what the investigators intended to measure, but the endpoint definition itself is not evidence that one treatment performed better than another.

11. Limitations

Important distinction: absence of a posted numerical analysis is different from evidence of no treatment effect. The registry status describes what is publicly reported in the ClinicalTrials.gov record; it does not, by itself, establish the direction or magnitude of the trial's efficacy result.

12. Why This Trial Matters Statistically

TRINOVA-2 is a useful teaching example because its primary endpoint illustrates a central problem in clinical-trial biostatistics: how to compare treatments when the outcome is defined not merely by whether an event occurs, but by when that event occurs.

ConceptHow it appears in TRINOVA-2
RandomizationThe trial uses randomized allocation to two treatment arms.
Double maskingThe registry identifies the study as double-blind.
Parallel designThe two interventions are evaluated in parallel treatment groups.
Time-to-event analysisThe primary endpoint is progression-free survival measured from randomization.
Kaplan-Meier estimationA standard approach for describing progression-free survival over time.
Log-rank testingA standard framework for comparing time-to-event distributions between randomized groups.
Cox regressionA standard model for estimating a hazard ratio when its assumptions are appropriate.
CensoringParticipants may require censoring rules when progression is not observed during available follow-up.
Radiologic assessmentProgression is defined using radiologic disease progression per RECIST 1.1 with modifications.
Repeated assessmentsThe registry specifies imaging intervals that change over the course of follow-up.

13. Clinical Interpretation vs Statistical Interpretation

Statistical interpretation

The registry defines a randomized comparison of progression-free survival between AMG 386 plus PLD and placebo plus PLD. A time-to-event framework is appropriate because the endpoint incorporates the time from randomization to radiologic progression.

Clinical interpretation

The clinical question is whether adding AMG 386 to PLD improves the duration of time before the registry-defined radiologic progression event compared with PLD plus placebo.

The two perspectives should remain connected but distinct. Statistical analysis determines how the observed event-time data are summarized and compared. Clinical interpretation asks what those estimates mean in the context of disease progression and treatment. Neither perspective should be substituted for the other.

14. A Practical Framework for Analysing TRINOVA-2

If a complete patient-level or formal results analysis were available, a rigorous statistical workflow for the primary endpoint would proceed through several linked steps.

01
DefineRandomization time zero
02
ClassifyProgression events
03
CensorNon-event follow-up
04
EstimateKaplan-Meier curves
05
CompareEffect and uncertainty

Step 1: Define the analysis time origin

The registry specifies randomization as the starting point for progression-free survival. Every participant's event-time calculation therefore begins from the date of randomization.

Step 2: Classify progression

Radiologic progression is identified using RECIST 1.1 with modifications. A statistical dataset would need to preserve the assessment history sufficiently to determine the first qualifying progression date according to the protocol's rules.

Step 3: Apply censoring rules

Participants without an observed qualifying event require protocol-defined censoring rules. These rules are important because censoring determines how much follow-up information contributes to the estimated survival distribution.

Step 4: Estimate the survival functions

Kaplan-Meier methods can summarize the estimated probability of remaining progression-free across follow-up for each randomized group. Graphical presentation can reveal differences that a single summary statistic may obscure.

Step 5: Quantify the treatment comparison

A hazard ratio from a Cox model can summarize the relative event rate, while a confidence interval communicates statistical uncertainty. A log-rank test can provide a formal comparison of the time-to-event distributions.

The key statistical principle: the primary endpoint should be analyzed according to its prespecified definition. Changing the event definition, time origin, censoring rules, or assessment schedule can change the estimand and therefore change the scientific question being answered.

15. Sources

Continue through the Clinical Biostats trial-analysis collection

Explore additional clinical trial statistical analyses and connect trial endpoints with broader biostatistical concepts.

16. Record Summary

TRINOVA-2 was a randomized, double-blind, parallel-design phase 3 trial with 223 participants and two treatment arms: AMG 386 plus PLD and placebo plus PLD. The registry identifies progression-free survival as the primary endpoint and defines it as the time from randomization to the earliest of the dates of first radiologic disease progression per RECIST 1.1 with modifications.

The statistical structure is therefore fundamentally a time-to-event comparison. Kaplan-Meier estimation can describe progression-free survival over time, log-rank methods can compare the treatment groups, and a Cox model can provide a hazard-ratio summary when its assumptions are appropriate. The registry's specified radiological assessment schedule is an important part of the endpoint measurement process.

The ClinicalTrials.gov record lists the study as terminated and does not post statistical analyses for the primary endpoint. As a result, the registry establishes the design and planned endpoint framework but does not provide a numerical PFS treatment effect, confidence interval, or p-value from which an efficacy conclusion can be drawn.

Clinical Biostats methodology: A trial-results page should distinguish the statistical question posed by a study from the numerical evidence available to answer it. For TRINOVA-2, the registry provides a clearly defined randomized time-to-event endpoint, while the posted record does not provide the corresponding formal statistical results.