This page separates reported trial information from statistical interpretation. The numerical trial information used here is restricted to the registry-reported FREEDOMS trial data.
1. Trial at a Glance
FREEDOMS was a phase 3 randomized, parallel-group, triple-blinded treatment trial in relapsing-remitting multiple sclerosis. The study compared two fingolimod doses with placebo and enrolled 1272 participants.
| Feature | FREEDOMS |
|---|---|
| Trial name | FREEDOMS |
| ClinicalTrials.gov identifier | NCT00289978 |
| Phase | Phase 3 |
| Status | Completed |
| Therapeutic area | Neurology |
| Condition | Relapsing-remitting Multiple Sclerosis |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Triple |
| Primary purpose | Treatment |
| Enrollment | 1272 |
| Lead sponsor | Novartis |
| Sponsor type | Industry |
| Start | 2006-01 |
| Primary completion | 2009-07 |
2. Clinical Question
The central statistical question was whether the annualized aggregate relapse rate over the study period differed among patients randomized to fingolimod 1.25 mg, fingolimod 0.5 mg, and placebo.
Population
Patients with relapsing-remitting multiple sclerosis.
Interventions
Fingolimod 1.25 mg and fingolimod 0.5 mg.
Comparator
Placebo.
Primary question
How does the annualized aggregate relapse rate compare across the two fingolimod dose groups and placebo through Month 24?
3. Trial Design
The ClinicalTrials.gov record describes a randomized, parallel-group, triple-masked phase 3 trial. Three treatment groups were evaluated under a treatment-oriented study design.
Fingolimod 1.25 mg
- Drug intervention
- 1272 total participants randomized across the three arms
Fingolimod 0.5 mg
- Drug intervention
- 1272 total participants randomized across the three arms
Placebo
- Drug comparator
- 1272 total participants randomized across the three arms
The ClinicalTrials.gov record does not provide randomized allocation ratios or arm-specific randomized sample sizes as part of the ClinicalTrials.gov record. Accordingly, this page does not infer an allocation ratio or reconstruct one from other information.
4. Randomization and Blinding
Randomization is one of the central design features of FREEDOMS. By assigning participants to treatment groups through a randomized process, the design is intended to separate treatment assignment from systematic differences that might otherwise affect the comparison.
The registry classifies the allocation as randomized.
Participants were assigned to one of three parallel treatment arms.
The registry-reported design classification identifies triple masking.
The primary purpose classification is treatment.
Why randomization matters statistically
Randomization does not guarantee that every measured baseline variable will be numerically identical between treatment groups. Its principal statistical value is different: it provides a design-based mechanism for making treatment assignment independent of prognostic characteristics in expectation. That makes between-group outcome differences interpretable as evidence about treatment assignment rather than simply as associations with pre-existing characteristics.
Why triple masking matters
Blinding can reduce the risk that knowledge of treatment assignment influences behavior, assessment, or other aspects of the trial. In a relapse-based endpoint, this distinction is important because outcome ascertainment and participant reporting can potentially be influenced by expectations about treatment. The registry's triple-masking classification therefore represents an important protection in the trial design.
5. Endpoints
| Endpoint | Registry definition / description | Time frame |
|---|---|---|
| Estimated Annualized Aggregate Relapse Rate (ARR) | The ARR is defined as the number of confirmed relapses in a year. A relapse is defined as the appearance of a new or worsening of a previously stable or improving pre-existing neurological abnormality, separated by at least 30 days from onset of a preceding relapse. The abnormality must be present for at least 24 hours and occur in the absence of fever or infection. The annualized ARR for each treatment group was defined in the registry as an annualized rate calculated from confirmed relapses and time on study. | Baseline to end of study (Month 24) |
The primary endpoint is a count / rate endpoint rather than a binary endpoint or a time-to-first-event endpoint. That distinction determines the statistical framework that is most natural for analysis.
6. Planned Analysis
The ClinicalTrials.gov record identifies results as posted, but no formal statistical analyses were posted. Consequently, the numerical treatment-effect estimates, confidence intervals, and p-values for the primary endpoint are not reproduced here.
What would normally be analyzed?
For an annualized relapse-rate endpoint, the analysis would compare the frequency of confirmed relapse events while accounting for the amount of time participants contribute to the study. The registry's endpoint definition itself expresses the annualization concept through total confirmed relapses relative to total time on study.
The key statistical feature is that participants may contribute different amounts of observation time. A rate-based analysis therefore differs from simply dividing the number of participants with at least one relapse by the total number randomized.
Why a count/rate model is different from a survival model
A time-to-event analysis asks when an event occurs, often focusing on the first event. An annualized relapse-rate analysis instead concerns the number of confirmed events accumulated over observed time. A participant can therefore contribute information through multiple relapse events, subject to the endpoint's confirmation rules.
For repeated event counts, investigators commonly consider count-regression approaches such as Poisson or negative-binomial models, frequently with an exposure-time offset. The appropriate choice depends on the prespecified statistical analysis plan, the event distribution, and whether the observed variability is adequately described by the chosen model. The registry-reported FREEDOMS data do not identify the formal model, so no specific model is attributed to the trial here.
7. Results
ClinicalTrials.gov results are identified in the ClinicalTrials.gov record, and the ClinicalTrials.gov record also identify a primary endpoint and safety results. However, no formal statistical analyses were posted. The primary endpoint therefore cannot be presented here as a formal treatment-effect analysis with an estimate, confidence interval, and p-value without importing information beyond the ClinicalTrials.gov record.
| Result category | What the ClinicalTrials.gov record establishes | Statistical interpretation |
|---|---|---|
| Primary endpoint | Estimated Annualized Aggregate Relapse Rate (ARR), measured from baseline to end of study (Month 24) | A formal between-arm estimate, confidence interval, and p-value are not included in the ClinicalTrials.gov record. |
| Outcome measures posted | 3 | the ClinicalTrials.gov record states that three outcome measures were posted. |
| Results posted | Yes | Results availability is recorded, but availability of results is not equivalent to having a registry-reported formal statistical-analysis record for every endpoint. |
The important distinction is between a reported outcome and a formal comparative estimate. An annualized relapse rate can be summarized separately within each randomized group, but a treatment-effect statement requires an explicitly defined comparison and its associated uncertainty. Because those formal statistical-analysis quantities are absent from the ClinicalTrials.gov record, no hazard ratio, rate ratio, confidence interval, or p-value is inferred here.
A rate itself also does not mean that every patient experienced that number of relapses. It is an aggregate measure of event frequency relative to observed time. Its interpretation depends on the event definition, follow-up, censoring or exposure time, and the statistical model used for comparison.
8. Secondary Endpoint Information
the ClinicalTrials.gov record states that three outcome measures were posted, but only the primary endpoint is fully represented in the registry-reported endpoint array. No secondary endpoint definitions or numerical secondary endpoint analyses are included in the ClinicalTrials.gov record.
Accordingly, this page does not reconstruct secondary efficacy estimates from publications or other external sources. That restriction is important because secondary endpoints can have different estimands, analysis populations, time frames, and multiplicity considerations from the primary endpoint.
9. Safety Results
The ClinicalTrials.gov record provides serious adverse event counts by treatment arm. These are reported as affected participants over participants at risk.
| Safety measure | Fingolimod 1.25 mg | Fingolimod 0.5 mg | Placebo |
|---|---|---|---|
| Serious adverse events | 51/429 | 43/425 | 56/418 |
The bars above display the number of affected participants, not a formal risk comparison. The denominators differ by arm, so visual comparison of counts alone should not be interpreted as a comparison of serious-adverse-event incidence.
A serious-adverse-event count is a numerator. The ClinicalTrials.gov record also provide the corresponding at-risk denominator, which is essential for interpreting frequency across treatment groups. A formal comparative safety analysis would require a prespecified definition of the estimand, an appropriate denominator and follow-up framework, and an uncertainty measure if a between-group effect were to be reported.
These figures should therefore be read as 51/429, 43/425, and 56/418, respectively, rather than as evidence of a statistically significant or nonsignificant difference. No such formal comparison is included in the ClinicalTrials.gov record.
10. Statistical Methodology
Annualized relapse rate
The primary endpoint is a rate based on confirmed relapse events over study time. Unlike a simple proportion, an annualized rate incorporates the amount of observation contributed by participants.
Annualization expresses that rate on a common yearly scale. The resulting quantity is an event-frequency measure, not a probability that a particular patient will experience an event.
Why exposure time matters
Suppose two groups experienced the same number of relapse events but contributed different total amounts of follow-up. A raw event count would treat those groups as equivalent, even though one group may have been observed for substantially longer. A rate analysis addresses this by incorporating time into the denominator.
Count models
Count outcomes are often analyzed with regression models designed for event counts. A Poisson model is one basic framework, while a negative-binomial model can be useful when the variability of counts exceeds what a simple Poisson distribution would imply. An exposure-time offset can account for unequal follow-up.
These are general statistical methods for count/rate endpoints. The registry-reported FREEDOMS data do not identify which formal model was used for the primary comparison, so these methods are presented as methodological context rather than as claims about the trial's prespecified analysis.
Confidence intervals for rates
A confidence interval around a rate or rate ratio quantifies statistical uncertainty under the chosen model. It does not describe the range of relapse counts that individual patients might experience. Precision depends on the amount of information available, including the number of observed events and the amount of follow-up.
Analysis population
The ClinicalTrials.gov record does not specify a formal statistical-analysis population for the primary endpoint beyond identifying the randomized trial design. A rigorous analysis should distinguish the population used for efficacy estimation from any population used for safety summaries, because the estimand can change when treatment exposure or post-randomization events determine who is included.
11. Statistical Methods Explained
Why is randomization important?
Randomization creates the treatment comparison at the design stage. If treatment assignment is randomized, systematic differences in prognostic factors are less likely to explain a treatment-group difference than they would in a purely observational comparison. Randomization therefore supports causal interpretation, although it does not eliminate every source of uncertainty or bias after randomization.
Why was blinding important for a relapse endpoint?
The trial was triple masked. Blinding can reduce the possibility that knowledge of treatment assignment affects participant behavior, investigator decisions, or outcome assessment. This is particularly relevant when the endpoint depends on identification and confirmation of clinical events rather than on a purely mechanical measurement.
What does an annualized relapse rate measure?
It measures the frequency of confirmed relapse events relative to observed study time, expressed on an annual scale. It is therefore different from the percentage of participants who ever relapse. A participant-level probability and a population-level event rate answer different questions.
Why can the same patient contribute more than one event?
The endpoint is defined as the number of confirmed relapses in a year. That makes the outcome fundamentally different from an analysis that records only the first relapse. Recurrent-event data can contain information about multiple events experienced by the same participant, which creates statistical dependence that a simple binomial analysis would not necessarily address.
Why might a Poisson model be insufficient?
A Poisson model makes a specific assumption about the relationship between the mean and variance of the event count. Clinical event data can show more variability than that assumption allows. A negative-binomial model is one common alternative when overdispersion is present. The appropriate choice should be determined from the prespecified analysis strategy and the observed data structure.
Why is the denominator as important as the numerator?
Fifty events observed over a short period and fifty events observed over a much longer period do not represent the same event frequency. The annualized rate incorporates total time on study, making the denominator central to interpretation.
12. Analysis Populations and Missing Data
The ClinicalTrials.gov record does not provide a detailed analysis-population specification for the primary endpoint, nor do they describe a formal missing-data or imputation strategy. These omissions matter because rate endpoints can be influenced by unequal follow-up and by the way participants who discontinue early contribute observation time.
Unequal follow-up
Participants who contribute different amounts of observation time should not automatically be treated as though they had identical exposure time.
Early discontinuation
Leaving the study early reduces the amount of observed time available for counting subsequent relapse events.
Missing event information
Incomplete ascertainment can affect event counts and therefore the estimated rate.
Imputation
No imputation method is specified in the ClinicalTrials.gov record, so none is attributed to FREEDOMS.
For a rate endpoint, missing-data handling is closely connected to the definition of observation time. An analysis should make clear whether participants contribute person-time until withdrawal, last assessment, another prespecified censoring point, or some other endpoint-specific time.
13. Multiplicity and the Three-Arm Design
FREEDOMS has three randomized arms and one registered primary endpoint. Even without importing an external testing strategy, the structure creates an important statistical issue: a three-arm trial can involve multiple pairwise treatment comparisons.
| Potential comparison | Role in the design | Statistical issue |
|---|---|---|
| Fingolimod 1.25 mg vs placebo | Dose-versus-control comparison | Contributes to the multiple-comparison structure |
| Fingolimod 0.5 mg vs placebo | Dose-versus-control comparison | Contributes to the multiple-comparison structure |
| Fingolimod 1.25 mg vs fingolimod 0.5 mg | Between-dose comparison | Would be a separate treatment contrast if formally tested |
The presence of several possible contrasts does not by itself establish that all three were confirmatory hypotheses. A multiplicity strategy must be defined prospectively. Because the ClinicalTrials.gov record does not provide the alpha-allocation or multiplicity procedure, this page does not assign one to FREEDOMS.
If several independent hypotheses are tested at the same nominal significance level, the probability of obtaining at least one apparently positive result by chance can exceed the nominal level for an individual test. Prespecified multiplicity control can address this problem through methods such as hierarchical testing, adjusted significance thresholds, or other error-control procedures.
14. Interim Analysis and Sequential Testing
The ClinicalTrials.gov record does not specify an interim-analysis schedule, alpha-spending method, stopping boundary, or information fraction. These features therefore cannot be attributed to FREEDOMS on the basis of the ClinicalTrials.gov record.
More generally, an interim analysis can create a statistical problem if accumulating data are repeatedly examined without an appropriate sequential-testing framework. If the investigators formally planned interim efficacy or futility decisions, the analysis plan would normally specify when those looks occurred and how the overall type I error was controlled.
Because those details are absent from the ClinicalTrials.gov record, no interim boundary or sequential-analysis claim is made here.
15. Non-Inferiority, Factorial, and Bayesian Methods
Non-inferiority
the ClinicalTrials.gov record does not identify a non-inferiority objective or margin. No non-inferiority interpretation is therefore applied.
Factorial design
The design model is parallel, with three arms. The ClinicalTrials.gov record does not identify a factorial structure.
Bayesian analysis
The ClinicalTrials.gov record does not identify a Bayesian model, prior distribution, or posterior analysis.
Parallel comparison
The registry classifies the intervention model as parallel assignment, making randomized between-arm comparison the central design structure.
16. Interpreting the Serious-Adverse-Event Data
The serious-adverse-event counts provide an opportunity to distinguish descriptive safety reporting from inferential statistics.
| Arm | Affected | At risk | What can be said directly |
|---|---|---|---|
| Fingolimod 1.25 mg | 51 | 429 | 51 participants are reported as affected among 429 at risk. |
| Fingolimod 0.5 mg | 43 | 425 | 43 participants are reported as affected among 425 at risk. |
| Placebo | 56 | 418 | 56 participants are reported as affected among 418 at risk. |
The numbers can be used to construct descriptive incidence proportions, but the ClinicalTrials.gov record does not provide confidence intervals, hypothesis tests, exposure-adjusted incidence rates, or a prespecified safety estimand. Accordingly, this page retains the reported affected/at-risk format rather than calculating new safety estimates.
The three affected/at-risk counts do not, by themselves, establish that one treatment has a statistically different serious-adverse-event risk from another. Such a conclusion would require a defined comparison, uncertainty assessment, and appropriate treatment of follow-up and event timing where relevant.
17. Trial Timeline
Trial start
the ClinicalTrials.gov record records the study start in January 2006.
Randomized parallel-group study
The trial was conducted as a phase 3 randomized, parallel-group, triple-masked treatment study with three arms.
Primary endpoint time frame
The registered primary endpoint, Estimated Annualized Aggregate Relapse Rate, was assessed from baseline to the end of study at Month 24.
Primary completion
the ClinicalTrials.gov record records primary completion in July 2009.
18. What an Annualized Relapse Rate Does — and Does Not — Mean
An annualized relapse rate is a population-level event-frequency measure. It summarizes the number of confirmed relapse events in relation to the amount of study time contributed by participants.
It does not mean that every patient experienced that number of relapses during one year. It also does not directly give the probability that an individual patient will experience a relapse.
The registry defines a relapse using clinical criteria, including the requirement that the neurological abnormality be present for at least 24 hours and occur in the absence of fever or infection, together with separation from the onset of a preceding relapse. Those rules determine which events enter the numerator.
Because the endpoint is annualized, the amount of observed study time is part of the statistical quantity itself. The denominator is therefore not merely the number of randomized participants.
19. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The trial uses a randomized three-arm design with a primary count/rate endpoint. A rigorous analysis requires comparison of relapse-event frequency while accounting for observation time and uncertainty.
Clinical interpretation
The endpoint is intended to quantify relapse activity over the 24-month study period. The clinical meaning of any treatment contrast depends on the magnitude, precision, event definition, and analysis population used.
20. Important Limitations and Interpretation Issues
- Incomplete the ClinicalTrials.gov record: no formal statistical analyses were posted, so formal treatment-effect estimates, confidence intervals, and p-values are not reproduced.
- Three-arm multiplicity: multiple potential treatment contrasts exist, but the ClinicalTrials.gov record does not specify a multiplicity-control strategy.
- Rate versus risk: annualized relapse rate is not equivalent to the probability that a participant experiences at least one relapse.
- Repeated events: a relapse-rate endpoint can incorporate multiple events per participant, creating a different statistical structure from a first-event endpoint.
- Observation time: unequal follow-up can affect the amount of information contributed to a rate analysis.
- Missing-data strategy: the ClinicalTrials.gov record does not specify an imputation or missing-data method.
- Analysis population: the ClinicalTrials.gov record does not identify the formal primary efficacy analysis population.
- Safety inference: serious-adverse-event affected/at-risk counts are descriptive unless accompanied by a prespecified inferential comparison.
- External results restriction: well-known published results are intentionally not substituted for the numerical statistical-analysis information contained in the ClinicalTrials.gov record.
21. Why This Trial Matters Statistically
FREEDOMS is a useful teaching case because its primary endpoint illustrates a different branch of clinical-trial statistics from the familiar binary and time-to-event frameworks. The core outcome is an event rate accumulated over study time, embedded in a randomized three-arm design.
| Concept | How it appears in FREEDOMS |
|---|---|
| Randomization | Randomized allocation across three treatment arms |
| Blinding | Triple masking |
| Parallel design | Parallel-group interventional model |
| Count/rate endpoint | Estimated Annualized Aggregate Relapse Rate |
| Repeated events | The primary endpoint counts confirmed relapses rather than simply recording whether a participant ever relapsed |
| Exposure time | The annualized rate incorporates time on study |
| Multiple arms | Two fingolimod doses and placebo create multiple potential treatment contrasts |
| Safety analysis | Serious adverse events are reported as affected participants over participants at risk |
| Phase 3 | Large randomized confirmatory treatment study structure |
The most important statistical lesson is that an endpoint's measurement scale determines the analysis. A relapse rate is not simply a binary outcome with a different label. The numerator represents event counts, the denominator incorporates observation time, and repeated events can contribute information from the same participant.
22. Statistical Concepts in This Trial
Learn more about the methods used in this trial:
23. Related Statistical Calculators
24. Sources
- ClinicalTrials.gov: NCT00289978 — FREEDOMS.
- Linked PubMed record: PubMed PMID 35698787.
- Linked PubMed record: PubMed PMID 27733070.
- Linked PubMed record: PubMed PMID 26121423.
- Linked PubMed record: PubMed PMID 25876471.
- Linked PubMed record: PubMed PMID 25245812.
Continue through the Clinical Biostats statistical pathway
Explore foundational clinical-trial concepts related to randomized and blinded study designs, then connect those concepts to statistical tools.
25. Record Summary
FREEDOMS is a completed phase 3 randomized, parallel-group, triple-masked trial in relapsing-remitting multiple sclerosis, with 1272 participants assigned across fingolimod 1.25 mg, fingolimod 0.5 mg, and placebo. Its registered primary endpoint was the Estimated Annualized Aggregate Relapse Rate from baseline to the end of study at Month 24.
From a statistical perspective, the trial is particularly useful for understanding the distinction between event counts, event rates, and individual-level risk. The primary endpoint requires attention to observation time and potentially repeated events, while the three-arm structure introduces a multiple-comparison question. The triple-masked randomized design also illustrates how trial design and outcome measurement interact with statistical interpretation.
The ClinicalTrials.gov record reports serious adverse events as 51/429 for fingolimod 1.25 mg, 43/425 for fingolimod 0.5 mg, and 56/418 for placebo. Those figures are descriptive safety data rather than formal comparative estimates.