This page separates information reported in the ClinicalTrials.gov record from statistical interpretation. The registry reports that results were posted, but the ClinicalTrials.gov record contains no formal statistical analyses, effect estimates, confidence intervals, or p-values.
1. Trial at a Glance
EXAM was a randomized, parallel-group, quadruple-masked phase 3 oncology trial evaluating XL184 (cabozantinib) versus placebo in thyroid cancer. The registered primary endpoint was progression-free survival, defined using progression determined by an Independent Review Committee according to mRECIST or death due to any cause.
| Feature | EXAM |
|---|---|
| Trial name | EXAM |
| NCT ID | NCT00704730 |
| Brief title | Efficacy of XL184 (Cabozantinib) in Advanced Medullary Thyroid Cancer |
| Therapeutic area | Oncology |
| Condition | Thyroid Cancer |
| Phase | Phase 3 |
| Status | COMPLETED |
| Allocation | RANDOMIZED |
| Design model | PARALLEL |
| Masking | QUADRUPLE |
| Primary purpose | TREATMENT |
| Enrollment | 330 |
| Arms | 2 |
| Interventions | XL184 (drug); Placebo (drug) |
| Lead sponsor | Exelixis |
| Sponsor type | INDUSTRY |
| Start | 2008-06 |
| Primary completion | 2011-10 |
| Results posted | Yes |
| Outcome measures posted | 6 |
2. Clinical Question
The registered study title describes an efficacy comparison of XL184 (cabozantinib) in advanced medullary thyroid cancer. Statistically, the primary question is represented by the registered progression-free survival endpoint: compare the time from randomization until an IRC-determined progression event or death due to any cause between the XL184 and placebo groups.
Population
Patients enrolled in the phase 3 EXAM study for thyroid cancer, with the trial brief title specifying advanced medullary thyroid cancer.
Intervention
XL184 (cabozantinib), recorded in the registry as a drug intervention.
Comparator
Placebo, recorded in the registry as a drug intervention.
Primary question
How does progression-free survival compare between randomized treatment groups under the registered PFS definition?
3. Trial Design
XL184
- XL184 (cabozantinib)
- Drug intervention
- Randomized parallel-group comparison
Placebo
- Placebo
- Drug intervention
- Randomized parallel-group comparison
The registry classifies the allocation as randomized, the design model as parallel, and the masking as quadruple. Those design features are statistically important because randomization establishes the basis for a between-group causal comparison, while masking can reduce the influence of treatment knowledge on trial conduct and assessment.
4. Trial Timeline
Trial start
The registered EXAM trial began in 2008-06.
Primary completion
The registered primary completion date was 2011-10.
Results posted
The registry indicates that results were posted, with 6 outcome measures posted.
5. Primary Endpoint
| Endpoint | Registered definition and time frame | Endpoint type |
|---|---|---|
| Progression-Free Survival (PFS) | The duration of Progression-Free Survival (PFS) using progression events as determined by Independent Review Committee (IRC) per mRECIST, or death due to any cause. Treatment period consisted of 4-week cycles with radiologic tumor assessment every 12 weeks from date of randomization. | Time-to-event |
The registered analysis was to be conducted after at least 315 subjects were randomized and at least 138 events were observed. These are design-information thresholds reported by the registry, not observed treatment-effect results.
6. Planned Analysis
ClinicalTrials.gov reports that results were posted for the registered primary PFS endpoint, but the ClinicalTrials.gov record contains no formal statistical analyses. Therefore, there is no reported hazard ratio, confidence interval, p-value, median PFS, event count by treatment arm, or other formal comparative estimate available in the ClinicalTrials.gov record.
For a randomized trial with a time-to-event primary endpoint such as this one, a conventional analysis would typically estimate the distribution of PFS with the Kaplan-Meier method, compare randomized groups with a log-rank test, and estimate a relative treatment effect with a Cox proportional-hazards model. Those are standard analytical approaches for this endpoint type; the ClinicalTrials.gov record does not establish that each of these procedures was actually used for EXAM.
A Kaplan-Meier analysis estimates the probability of remaining progression-free over time while accounting for right censoring. A Cox model can summarize the relative event rate with a hazard ratio, while a log-rank test can compare the survival distributions.
7. Statistical Methodology
Randomization
EXAM used randomized allocation. Randomization is the key design mechanism that allows the treatment groups to be compared as groups rather than treating observed differences as inherently causal. In a properly conducted randomized trial, treatment assignment is determined independently of the outcome that will subsequently be observed.
Parallel-group comparison
The design model was parallel. Each randomized participant belongs to one treatment pathway rather than serving as a repeated randomized comparison of both interventions. The treatment-group contrast is therefore the primary unit of inference.
Quadruple masking
The registry classifies the study as quadruple-masked. Masking is particularly relevant in trials containing subjective decisions or assessments because knowledge of assignment can potentially influence behavior, assessment, reporting, or other trial processes. Here, progression was determined by an Independent Review Committee per mRECIST, providing an additional structured assessment mechanism.
Independent review and mRECIST
The registered PFS definition specifies progression determined by an Independent Review Committee using mRECIST. This is important statistically because the endpoint is not simply a clinician's unstructured declaration of progression; it is defined through a prespecified radiologic assessment framework and independent review.
Repeated radiologic assessment
The registry states that radiologic tumor assessment occurred every 12 weeks from the date of randomization during a treatment period consisting of 4-week cycles. Repeated assessment creates discrete opportunities at which progression can be identified. Time-to-event methods then incorporate both events and censored follow-up rather than reducing each patient to a simple binary responder/non-responder classification.
Event-driven information
The registered PFS analysis was to be conducted after at least 315 subjects had been randomized and at least 138 events had been observed. This illustrates an important feature of survival-trial design: statistical information depends not only on the number of participants but also on the number of observed endpoint events.
8. Statistical Methods Explained
Why is PFS a time-to-event endpoint?
PFS records when the first qualifying event occurs rather than only whether an event occurred. Participants who remain progression-free at the end of their observable follow-up do not simply disappear from the analysis; their information contributes up to the point at which they are censored.
Why use Kaplan-Meier estimation?
Kaplan-Meier estimation is designed for incomplete follow-up caused by censoring. It produces an estimated progression-free survival curve over time without requiring every participant to have experienced progression or death.
Why would a log-rank test be useful?
A log-rank test provides a formal way to compare time-to-event distributions between randomized groups. Unlike a simple comparison of the proportion of patients who eventually progressed, it uses the ordering and timing of observed events while accounting for censored observations.
What would a Cox hazard ratio mean?
If a Cox model were used, its hazard ratio would summarize the relative instantaneous rate of the PFS event between the treatment groups under the model. A hazard ratio below 1 would indicate a lower estimated event hazard in the numerator group, while a hazard ratio above 1 would indicate a higher estimated event hazard. A hazard ratio would not mean that the same percentage of patients experienced that relative change in their individual probability of progression or death.
Why does the number of events matter?
Time-to-event precision depends substantially on the number of events. The EXAM registry specified a minimum of at least 138 observed events for the planned analysis. More events generally provide more information for estimating relative treatment effects than an equivalent increase in participants who have not yet experienced the event.
Why does the independent review matter?
Progression is part of the primary endpoint, so how progression is determined directly affects the endpoint. The registry's use of an Independent Review Committee per mRECIST establishes a prespecified framework for determining progression rather than leaving the primary endpoint entirely to an unstructured assessment.
9. Understanding Censoring in PFS
Censoring is fundamental to the interpretation of any PFS analysis. A participant who has not experienced IRC-determined progression or death by the relevant end of follow-up does not provide an exact event time. Instead, the analysis uses the information available up to the censoring point.
Observed event
The participant reaches the PFS endpoint through an IRC-determined progression event or death due to any cause.
Censored observation
The participant has not experienced the defined event during the period for which usable follow-up is available.
The statistical challenge is that censoring must be handled appropriately. Kaplan-Meier and Cox methods are designed for right-censored time-to-event data, but their validity depends on assumptions about the relationship between censoring and the event process. The ClinicalTrials.gov record does not describe its specific censoring rules or missing-data procedures.
10. Safety Information Reported by Arm
The ClinicalTrials.gov record reports serious adverse events by treatment arm as the number affected divided by the number at risk. These figures provide descriptive safety information but are not accompanied in the ClinicalTrials.gov record by a formal comparative statistical analysis.
| Treatment arm | Serious adverse events | Affected / at risk |
|---|---|---|
| XL184 (Cabozantinib) | Serious adverse events | 90/214 |
| Placebo | Serious adverse events | 25/109 |
The denominator is important when interpreting these figures: the registry reports the number affected relative to the number at risk, rather than simply reporting the affected counts. The ClinicalTrials.gov record does not provide a confidence interval, hypothesis test, exposure-adjusted incidence rate, time-to-first-serious-adverse-event analysis, or a definition of the safety analysis population beyond these reported affected/at-risk figures.
11. What Can and Cannot Be Concluded From the ClinicalTrials.gov record
| Question | Information available in the ClinicalTrials.gov record |
|---|---|
| Was EXAM randomized? | Yes. The allocation is recorded as RANDOMIZED. |
| Was the design parallel? | Yes. The design model is PARALLEL. |
| Was the trial masked? | Yes. The registry classification is QUADRUPLE. |
| What was the primary endpoint? | Progression-Free Survival (PFS). |
| Was PFS a time-to-event endpoint? | Yes. |
| How was progression defined? | Progression determined by an Independent Review Committee per mRECIST, or death due to any cause. |
| When were radiologic assessments scheduled? | Every 12 weeks from date of randomization. |
| Was a formal statistical analysis reported? | No. No formal statistical analyses were posted. |
| Were results posted? | Yes. |
| Are a PFS hazard ratio and confidence interval available in the ClinicalTrials.gov record? | No. |
| Are PFS p-values available in the ClinicalTrials.gov record? | No. |
| Are median PFS values available in the ClinicalTrials.gov record? | No. |
| Are subgroup efficacy estimates available? | No. |
12. Limitations
- No formal statistical analyses reported: the ClinicalTrials.gov record indicates that results were posted, but provides no formal statistical analyses.
- No effect estimate: the ClinicalTrials.gov record does not provide a hazard ratio, risk ratio, median PFS, confidence interval, or p-value for the primary endpoint.
- No event breakdown by arm: the registry data provides the threshold of at least 138 events for the planned analysis but does not provide an observed PFS event count by treatment arm.
- No censoring details: the ClinicalTrials.gov record does not specify the detailed censoring algorithm, follow-up rules, or handling of potentially informative censoring.
- No missing-data or imputation information: the ClinicalTrials.gov record does not describe a missing-data or imputation strategy for the primary endpoint.
- No stratification information: the ClinicalTrials.gov record does not identify randomization strata or stratified analysis factors.
- No multiplicity information: the ClinicalTrials.gov record does not describe an alpha hierarchy, multiplicity adjustment, or testing strategy beyond identifying one registered primary endpoint.
- No interim-analysis details: the ClinicalTrials.gov record does not report an interim monitoring plan, stopping boundaries, or alpha-spending method.
- No non-inferiority margin: the ClinicalTrials.gov record does not identify a non-inferiority design or margin.
- No crossover information: the ClinicalTrials.gov record does not report crossover between treatment groups.
- No subgroup estimates: subgroup treatment effects cannot be reconstructed from the ClinicalTrials.gov record.
- Safety is descriptive here: serious adverse-event counts by arm are reported, but no formal comparative safety analysis accompanies them.
13. Why This Trial Matters Statistically
EXAM is a useful teaching example because the registry record illustrates how a randomized oncology trial can be structured around a time-to-event endpoint even when the numerical treatment-effect analysis is not available in the ClinicalTrials.gov record.
| Concept | How it appears in EXAM |
|---|---|
| Randomization | The allocation is recorded as RANDOMIZED. |
| Parallel design | The design model is PARALLEL. |
| Blinding | The masking classification is QUADRUPLE. |
| Time-to-event endpoint | PFS is the single registered primary endpoint. |
| Independent review | Progression is determined by an Independent Review Committee. |
| mRECIST | The registered PFS definition uses progression determined per mRECIST. |
| Repeated assessment | Radiologic tumor assessment occurs every 12 weeks from randomization. |
| Event-driven analysis | The planned analysis required at least 138 observed events and at least 315 randomized subjects. |
| Kaplan-Meier framework | A standard approach for estimating PFS distributions from censored time-to-event data. |
| Cox model | A standard approach for estimating a relative hazard when appropriate for the endpoint and assumptions. |
| Safety analysis | Serious adverse events are reported descriptively as affected / at risk by arm. |
14. Statistical Interpretation of the Primary Endpoint
The registered PFS endpoint combines two clinically important events: IRC-determined progression per mRECIST and death due to any cause. It therefore measures the time until the first qualifying event rather than measuring tumor response alone.
A formal time-to-event comparison would estimate whether the distribution of PFS differed between the randomized XL184 and placebo groups. Depending on the prespecified statistical analysis, this could be expressed using Kaplan-Meier estimates, a log-rank test, a Cox-model hazard ratio, or related time-to-event quantities.
Without a reported hazard ratio and its uncertainty interval, there is no numerical basis in the ClinicalTrials.gov record for describing the magnitude or precision of the treatment effect. The fact that results are marked as posted does not substitute for the actual statistical estimate.
Even if a p-value were available, it would primarily address evidence against a specified null hypothesis under the chosen testing framework. It would not by itself describe the magnitude of the treatment effect, its clinical importance, or the precision of the estimate. Those questions require an effect estimate and an uncertainty measure.
15. Analysis Population and Causal Interpretation
The ClinicalTrials.gov record confirms randomization but does not provide the detailed analysis-population definitions used for the primary endpoint. That distinction matters because a randomized trial can contain several analysis populations, and the choice of population can affect the interpretation of the resulting estimate.
For a primary efficacy endpoint in a randomized trial, an intention-to-treat approach is commonly important because it preserves the comparison created by randomization. However, the registry-reported EXAM data does not explicitly identify the primary analysis population, so this page does not attribute a specific analysis population to the trial.
16. Time Frames and Measurement Schedule
| Trial feature | Registered information |
|---|---|
| Treatment cycle | 4-week cycles |
| Radiologic assessment | Every 12 weeks |
| Assessment origin | Date of randomization |
| PFS event | IRC-determined progression per mRECIST or death due to any cause |
| Planned analysis threshold | At least 315 subjects randomized and at least 138 events observed |
The relationship between treatment cycles and radiologic assessment is statistically relevant because these are different time scales. Treatment was organized into 4-week cycles, while radiologic tumor assessment was scheduled every 12 weeks from randomization. PFS is defined continuously as a time-to-event measure even though progression may only become observable at scheduled assessment times.
17. Serious Adverse Events: Statistical Perspective
The ClinicalTrials.gov record reports serious adverse events as affected participants divided by participants at risk: 90/214 for XL184 (cabozantinib) and 25/109 for placebo.
The visual is intentionally based on the affected counts relative to the reported XL184 affected count and is not a statistical comparison. A formal safety comparison would require prespecified definitions of the safety population, follow-up or exposure time, event ascertainment, and the appropriate statistical method.
For example, if the scientific question were simply whether the proportion of participants experiencing at least one serious adverse event differed between randomized groups, a two-group proportion analysis could be considered. If the timing of serious adverse events and differential follow-up were central, a time-to-first-event approach could be more informative. The ClinicalTrials.gov record does not specify which approach was used.
18. What the Registry Record Does Not Establish
Effect size
The ClinicalTrials.gov record does not establish a numerical PFS treatment effect.
Precision
No confidence interval for the primary endpoint is reported.
Statistical significance
No PFS p-value or formal hypothesis-test result is reported.
Median PFS
No median PFS estimate is posted on ClinicalTrials.gov for either treatment arm.
This distinction is important for an educational statistical record. A registry can identify an endpoint, define it carefully, and indicate that results have been posted without the ClinicalTrials.gov recordset containing the statistical estimates necessary to reconstruct the comparative result.
19. Clinical Biostats Interpretation
EXAM's primary endpoint is structurally suited to survival-analysis methods because it records the duration of progression-free survival and includes censoring. The event definition is also clinically specific: progression must be determined by an Independent Review Committee per mRECIST, or the participant can reach the endpoint through death due to any cause.
Randomization and quadruple masking establish important safeguards for a comparative clinical trial. The independent review component adds structure to determination of the progression component of PFS.
The central statistical limitation of the ClinicalTrials.gov record is not that the trial lacks a defined primary endpoint; rather, it is that the ClinicalTrials.gov record does not include the formal statistical analysis of that endpoint.
20. Related Tutorials
Learn more about the methods used in this trial:
21. Related Calculators
22. Sources
- ClinicalTrials.gov: EXAM — NCT00704730.
- Linked publication: PubMed record for PMID 29045520.
- Linked publication: PubMed record for PMID 24002501.
Continue through Clinical Biostats
Explore the broader statistical methods used to design, analyze, and interpret randomized clinical trials.
23. Record Summary
EXAM was a randomized, parallel-group, quadruple-masked phase 3 trial of XL184 (cabozantinib) versus placebo in advanced medullary thyroid cancer, with 330 participants enrolled and a single registered primary endpoint: progression-free survival. The endpoint was defined as the duration from randomization to IRC-determined progression per mRECIST or death due to any cause, with radiologic tumor assessment every 12 weeks during a treatment period consisting of 4-week cycles.
From a statistical perspective, the trial illustrates the structure of an event-driven time-to-event analysis. The registry specified an analysis after at least 315 subjects had been randomized and at least 138 events had been observed. A conventional analysis of this type of endpoint would typically use Kaplan-Meier estimation and a comparative time-to-event model such as Cox regression, with an appropriate hypothesis test such as the log-rank test. However, the ClinicalTrials.gov record contains no formal statistical analysis, so no numerical PFS treatment effect, confidence interval, or p-value is reported here.
The registry does provide descriptive serious-adverse-event information: 90/214 for XL184 (cabozantinib) and 25/109 for placebo. These figures should remain separate from the efficacy analysis because they answer a different statistical question and are not accompanied in the ClinicalTrials.gov record by a formal comparative safety analysis.