This page separates reported registry information from statistical interpretation. The ClinicalTrials.gov record contains a posted primary outcome measure but no formal statistical analyses with treatment-effect estimates, confidence intervals, or p-values.
1. Trial at a Glance
CONFIRM was a randomized, double-blind, parallel phase 3 treatment trial in relapsing-remitting multiple sclerosis. The trial enrolled 1417 participants and evaluated BG00012, placebo, and glatiramer acetate across 4 arms.
| Feature | CONFIRM |
|---|---|
| Trial name | CONFIRM |
| ClinicalTrials.gov identifier | NCT00451451 |
| Brief title | Efficacy and Safety Study of Oral BG00012 With Active Reference in Relapsing-Remitting Multiple Sclerosis |
| Phase | Phase 3 |
| Status | COMPLETED |
| Therapeutic area | Neurology |
| Condition | Relapsing-Remitting Multiple Sclerosis |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Double |
| Primary purpose | Treatment |
| Enrollment | 1417 |
| Interventions | BG00012; Placebo; Glatiramer Acetate |
| Lead sponsor | Biogen |
| Sponsor type | Industry |
2. Clinical Question
The trial was designed to evaluate the efficacy and safety of oral BG00012 with an active reference in patients with relapsing-remitting multiple sclerosis. The primary registered endpoint was annualized relapse rate over 2 years.
Population
Patients with relapsing-remitting multiple sclerosis enrolled in a phase 3 treatment trial.
Intervention
BG00012, evaluated in 240 mg Twice Daily (BID) and 240 mg 3 Times Daily (TID) arms.
Comparators
Placebo and glatiramer acetate (GA) 20 mg injection served as comparator arms.
Primary question
How does the trial compare annualized relapse rate over 2 years across the randomized treatment groups?
3. Trial Design
Participants were assigned to randomized treatment groups.
The registry identifies the study design model as parallel.
The trial used double masking.
The registry identifies treatment as the primary purpose.
Study arms
BG00012 Twice Daily
Serious adverse events were reported in 61 of 359 participants in this arm.
BG00012 3 Times Daily
Serious adverse events were reported in 54 of 344 participants in this arm.
Placebo
Serious adverse events were reported in 79 of 363 participants in this arm.
GA 20 mg Injection
Serious adverse events were reported in 60 of 351 participants in this arm.
4. Endpoints
| Endpoint | Time frame | Registry definition / information reported |
|---|---|---|
| Annualized Relapse Rate | 2 years | A protocol-defined relapse was defined as new or recurrent neurologic symptoms not associated with fever or infection that lasted at least 24 hours, and were separated by at least 30 days from onset of a preceding relapse. All protocol-defined relapses were evaluated by an independent neurologic evaluation committee. The adjusted annualized relapse rate was calculated from a negative binomial regression model , adjusted for baseline Expanded Disability Status Scale (EDSS ) score(≤2.0 versus>2.0), age (\<40 versus ≥40 years), region, and the number of relapses in the 1 year prior to enrollment. |
What an annualized relapse rate measures
Annualized relapse rate is a rate-based endpoint rather than a conventional binary endpoint. It summarizes the frequency of qualifying relapses over the observation period while accounting for the amount of follow-up represented in the analysis. This is different from simply asking whether a participant experienced at least one relapse.
The distinction matters because two participants can both be classified as having experienced a relapse while contributing different numbers of observed relapse events. A rate model is designed to retain that event-count information rather than reducing the outcome to a single yes/no indicator.
5. Statistical Results Status
No formal statistical-analysis estimates are reported
The ClinicalTrials.gov record indicates that results are posted and that 5 outcome measures are posted.
However, no formal statistical analyses were posted. Therefore no treatment-effect estimate, confidence interval, or p-value can be reported from the ClinicalTrials.gov record.
This distinction is important. A registry can contain posted outcome information without the structured statistical-analysis record needed to reproduce a formal inferential comparison. The absence of a registry-reported hazard ratio, rate ratio, confidence interval, or p-value is not evidence that an effect was absent; it means that those inferential results are not present in the ClinicalTrials.gov record.
6. Planned Analysis
A rate model relates the number of qualifying events to the amount of time over which those events could be observed. For an annualized relapse endpoint, this makes the analysis fundamentally different from a simple comparison of proportions.
For an endpoint of this type, a negative binomial regression is commonly used when relapse counts show greater variability than would be accommodated by a simple Poisson model. The regression framework can estimate treatment-group differences in relapse rates while accounting for the observation period through an exposure or follow-up-time component.
In a typical rate-ratio interpretation, a treatment-group rate ratio below 1 would indicate a lower estimated relapse rate relative to the reference group, whereas a value above 1 would indicate a higher estimated rate. The registry-reported CONFIRM data do not provide such an estimate, so no numerical treatment effect is presented here.
Why negative binomial regression is appropriate to understand
Relapses are recurrent events. One participant may experience several qualifying relapses while another experiences none. Treating only the first relapse as the endpoint would discard information about subsequent events. A count model preserves the distinction between zero, one, and multiple events.
The negative binomial model also allows the event counts to be more dispersed than a Poisson distribution assumes. In clinical relapse data, that flexibility can be important because patient-to-patient variability in relapse frequency may be substantial.
Adjusted annualized relapse rate
The registry specifically describes the primary endpoint as an adjusted annualized relapse rate. Adjustment means that the reported analysis is not merely a raw division of the total number of observed relapses by a calendar period. The exact adjustment variables and model specification are not provided in the registry-reported statistical-analysis field, so they are not added here.
7. Statistical Methods Explained
Why use a rate instead of a simple proportion?
A proportion answers whether a participant experienced an event. An annualized relapse rate instead concerns the frequency of qualifying events during follow-up. Because relapses can recur, the rate framework can use more of the observed event information.
What does a negative binomial model do?
A negative binomial regression models event counts while allowing more variability in the counts than a basic Poisson model. For recurrent clinical events, this can provide a more flexible representation of between-participant differences in event frequency.
What would a rate ratio mean?
A rate ratio compares estimated event rates between groups. For example, a rate ratio of 0.70 would ordinarily be interpreted as an estimated rate 30% lower in the numerator group than in the reference group. That numerical example is an explanation of the statistic, not a reported CONFIRM result.
Why does follow-up time matter?
A participant observed for a longer period has more opportunity to experience a relapse than a participant observed for a shorter period. Annualization and exposure-time handling are therefore central to interpreting a rate endpoint.
Why define a relapse so precisely?
The registry definition requires new or recurrent neurologic symptoms, excludes symptoms associated with fever or infection, requires symptoms to last at least 24 hours, and requires separation of at least 30 days from onset of a preceding relapse. These rules reduce ambiguity about which clinical events enter the primary count.
Why was an independent neurologic evaluation committee used?
The registry states that all protocol-defined relapses were evaluated by an independent neurologic evaluation committee. Independent assessment can reduce the influence of treatment assignment or local investigator judgment on whether an event satisfies the protocol definition.
Why is a p-value not enough?
A p-value addresses evidence against a specified null hypothesis under the statistical model and testing framework. It does not describe the magnitude of a treatment effect or its clinical importance. For a count/rate endpoint, an effect estimate such as a rate ratio together with a confidence interval would provide information about both direction and precision. No such CONFIRM estimate is reported in the ClinicalTrials.gov record for this page.
8. Analysis Population and Interpretation
the ClinicalTrials.gov record identifies the study as randomized, double masked, and parallel. Randomization is important because it establishes the treatment groups before outcomes are observed, supporting comparison of outcomes according to randomized assignment.
However, the ClinicalTrials.gov record does not specify the exact efficacy analysis population, treatment-discontinuation rules, censoring rules, missing-data strategy, or sensitivity analyses. Those details can materially affect the analysis of recurrent-event outcomes. They should therefore not be reconstructed from generic clinical-trial conventions.
What is known
The primary endpoint was annualized relapse rate over 2 years, categorized as a count/rate endpoint, with an adjusted annualized relapse rate calculated from a negative binomial regression.
What is not reported
No formal treatment-effect estimate, confidence interval, or p-value is included in the registry-reported statistical-analysis field.
9. Safety Results
The ClinicalTrials.gov record provides serious adverse events by randomized treatment arm. These counts are descriptive safety information and should not be converted into comparative statistical conclusions without the corresponding analysis population and inferential framework.
| Arm | Serious adverse events affected | At risk |
|---|---|---|
| Placebo | 79 | 363 |
| BG00012 240 mg Twice Daily (BID) | 61 | 359 |
| BG00012 240 mg 3 Times Daily (TID) | 54 | 344 |
| Total BG00012 | 115 | 703 |
| Glatiramer Acetate (GA) 20 mg Injection | 60 | 351 |
The chart displays the number of affected participants, not the proportion of participants affected. Because the arms have different denominators, raw counts should not be interpreted as if they represented equivalent exposure groups.
10. How to Read the Safety Data
The serious-adverse-event information illustrates an important statistical principle: a numerator without its denominator is incomplete. The data therefore report both affected participants and participants at risk for each arm.
Even with denominators, a descriptive comparison is not automatically an inferential comparison. A formal safety analysis could depend on definitions, exposure, follow-up, event timing, recurrent events, and the prespecified analysis population. None of those additional specifications are reported here.
This is a useful descriptive quantity, but it is not by itself a statistical test and does not establish a treatment-related difference.
11. Statistical Methodology
Negative binomial regression
The registry identifies negative binomial regression as the basis for calculating the adjusted annualized relapse rate. Conceptually, the model treats the number of qualifying relapses as a count outcome and relates that count to treatment and the observation period.
Here, Y represents a relapse count, X represents a treatment-group indicator, and T represents observation time. This is a conceptual representation of a count-rate model, not a reconstruction of the CONFIRM statistical-analysis specification.
The logarithm of follow-up time can function as an offset in a rate model, allowing the analysis to distinguish a higher event count caused simply by longer observation from a genuinely higher underlying event rate. The exact offset, covariates, and model parameterization used in CONFIRM are not provided in the ClinicalTrials.gov record.
Rate ratios
In a negative binomial model, the exponentiated treatment coefficient is commonly interpreted as a rate ratio. A rate ratio below 1 corresponds to a lower modeled event rate for the treatment group relative to its reference group; a value above 1 corresponds to a higher rate.
The registry-reported CONFIRM dataset does not contain a numerical rate ratio. Accordingly, this page explains the interpretation without assigning a value to the trial.
Randomization and masking
Randomization is a design feature rather than a statistical estimator, but it is fundamental to interpretation. By assigning treatment before outcome assessment, randomization provides the framework for comparing outcome distributions across treatment groups. Double masking can further reduce the potential influence of treatment knowledge on behavior, assessment, or reporting.
Independent relapse assessment
The protocol-defined relapse endpoint was evaluated by an independent neurologic evaluation committee. For an outcome that depends on applying clinical criteria, independent assessment is relevant because endpoint classification itself can otherwise introduce an additional source of variation.
12. Multiplicity and the Four-Arm Design
CONFIRM has 4 arms, including two BG00012 dosing schedules, placebo, and glatiramer acetate. That structure creates several possible pairwise or grouped comparisons.
| Comparison type | Statistical issue |
|---|---|
| BG00012 240 mg BID vs placebo | Direct comparison of one BG00012 dosing group with placebo. |
| BG00012 240 mg TID vs placebo | Direct comparison of the second BG00012 dosing group with placebo. |
| BG00012 vs active reference | Potential comparison involving glatiramer acetate, depending on the prespecified analysis. |
| Multiple treatment comparisons | Repeated confirmatory testing can require multiplicity control to preserve the intended type I error rate. |
The ClinicalTrials.gov record does not state which treatment comparisons were formally confirmatory, whether the two BG00012 regimens were tested separately or jointly for the primary endpoint, or what multiplicity adjustment was prespecified. Therefore no claim about statistical significance of any particular comparison is made.
13. Missing Data, Censoring, and Follow-Up
Annualized relapse rate is particularly sensitive to the amount of time during which relapses can be observed. A participant who contributes less observation time should not automatically be treated as equivalent to a participant observed for the entire 2-year period.
The ClinicalTrials.gov record does not specify the detailed rules for handling treatment discontinuation, withdrawal, missing assessments, or incomplete follow-up. Those rules are therefore not described as if they were known.
Observation time
Rate-based analyses account for the period over which events are observed. This is one reason a rate model is useful for recurrent relapse outcomes.
Missing-data assumptions
The ClinicalTrials.gov record does not specify the imputation or sensitivity-analysis strategy for missing or incomplete outcome information.
Censoring
The detailed censoring rules for the primary relapse-rate analysis are not included in the registry-reported statistical-analysis field.
Analysis population
The exact primary efficacy analysis population is not specified in the ClinicalTrials.gov record.
14. What a Confidence Interval Would Add
No confidence interval for the primary annualized relapse rate comparison is included in the ClinicalTrials.gov record. Nevertheless, confidence intervals are an important part of interpreting a count-rate treatment effect.
A 95% confidence interval around a rate ratio would describe statistical uncertainty around the estimated treatment effect under the specified model and sampling framework. It would not describe the range of relapse rates that individual patients could experience.
A p-value and a confidence interval answer related but different questions. The p-value concerns compatibility with a null hypothesis, whereas the confidence interval communicates the range of effect estimates supported by the statistical analysis. Neither alone determines clinical importance.
Without a rate ratio and its confidence interval, the ClinicalTrials.gov record cannot establish the magnitude or precision of a between-group efficacy effect. The correct response is to preserve that uncertainty rather than substitute an external result.
15. Clinical Biostats Interpretation of the Primary Endpoint
The primary endpoint measures the frequency of protocol-defined relapses over 2 years. It is therefore a recurrent-event outcome rather than a one-time binary endpoint.
Annualized relapse rate is not the same as the proportion of patients who relapse, time to first relapse, disability progression, or another clinical endpoint. A rate summarizes event frequency over observed follow-up.
The registry identifies negative binomial regression for the adjusted annualized relapse rate. The choice of a count model determines how recurrent events, variability in event counts, and observation time contribute to the estimated treatment effect.
Because no statistical-analysis estimates are reported, this page cannot state the numerical size of the treatment effect, the uncertainty around it, or whether a particular comparison met a prespecified significance threshold.
16. Safety Interpretation
The safety data show serious adverse events in each reported treatment group. The figures are best read first as arm-specific descriptive counts with denominators.
| Group | Serious AEs | At risk | Statistical reading |
|---|---|---|---|
| Placebo | 79 | 363 | Descriptive count and denominator; no formal comparative estimate reported. |
| BG00012 240 mg BID | 61 | 359 | Descriptive count and denominator; no formal comparative estimate reported. |
| BG00012 240 mg TID | 54 | 344 | Descriptive count and denominator; no formal comparative estimate reported. |
| GA 20 mg Injection | 60 | 351 | Descriptive count and denominator; no formal comparative estimate reported. |
The separate Total BG00012 value of 115/703 combines the two BG00012 dose groups. It should therefore be interpreted as an aggregate exposure group, not as an additional randomized arm.
17. Limitations
- No formal statistical analyses reported: the structured no formal statistical analyses were posted, so no numerical efficacy estimate, confidence interval, or p-value can be reported from the ClinicalTrials.gov record.
- Incomplete endpoint wording: the registry-reported annualized relapse rate definition ends with “negative binomial regression model , adjusted for baseline Expanded Disability Status Scale (EDSS ) score(≤2.0 versus>2.0), age (\<40 versus ≥40 years), region, and the number of relapses in the 1 year prior to enrollment.”, so the remainder of the registry sentence is not reconstructed.
- Four-arm multiplicity: the study contains 4 arms, but the ClinicalTrials.gov record does not specify the formal confirmatory comparison hierarchy or multiplicity adjustment.
- Analysis population not specified: the registry-reported profile does not identify the exact efficacy analysis population or detailed safety analysis population.
- Missing-data rules not reported: treatment discontinuation, withdrawal, incomplete follow-up, and imputation procedures are not described in the ClinicalTrials.gov record.
- Exposure and follow-up: the detailed rules used to account for incomplete observation time are not reported beyond the identification of annualized relapse rate.
- Safety inference: serious adverse-event counts by arm are descriptive and should not be treated as a formal comparative safety analysis without the corresponding statistical framework.
- External results excluded: no efficacy results from publications or other external sources have been imported because the task restricts numerical trial facts to the ClinicalTrials.gov record.
18. Why This Trial Matters Statistically
CONFIRM is a useful teaching example because it combines randomized allocation, double masking, multiple treatment arms, a recurrent count/rate endpoint, independent endpoint assessment, and arm-specific safety data.
| Concept | How it appears in CONFIRM |
|---|---|
| Randomization | The registry identifies the allocation as randomized. |
| Double masking | The registry identifies the trial as double masked. |
| Parallel design | The design model is parallel. |
| Multiple treatment groups | There are 4 arms, including two BG00012 dosing schedules, placebo, and glatiramer acetate. |
| Count/rate endpoint | The primary endpoint is Annualized Relapse Rate. |
| Recurrent events | The relapse definition permits multiple protocol-defined relapse events to contribute to the rate endpoint. |
| Negative binomial regression | The registry definition identifies this model for the adjusted annualized relapse rate. |
| Independent assessment | Protocol-defined relapses were evaluated by an independent neurologic evaluation committee. |
| Multiplicity | The 4-arm design creates multiple possible comparisons, making the prespecified testing strategy important. |
| Safety denominators | Serious adverse events are reported as affected participants over participants at risk for each arm. |
19. Statistical Interpretation vs Clinical Interpretation
Statistical interpretation
The registered primary outcome is a recurrent count/rate endpoint analyzed using an adjusted annualized relapse-rate framework based on negative binomial regression. The ClinicalTrials.gov recordset does not contain the resulting treatment-effect estimates or inferential statistics.
Clinical interpretation
The endpoint is clinically oriented toward relapse frequency in relapsing-remitting multiple sclerosis. Its interpretation depends on the precise definition of qualifying relapses and the duration of follow-up.
Keeping these two layers separate is important. Statistical methodology describes how evidence is quantified; clinical interpretation asks what the measured endpoint represents. Neither layer should be filled with unsupported numerical conclusions.
20. A Worked Conceptual Example of the Primary Analysis
The following is a statistical illustration, not a CONFIRM result. It demonstrates how an annualized relapse-rate comparison can be interpreted without introducing any trial-specific estimate that is absent from the ClinicalTrials.gov record.
If an analysis produced a rate ratio of 0.75, the modeled relapse rate would be interpreted as 25% lower in the treatment group than in the reference group. The value 0.75 is purely illustrative and is not reported for CONFIRM here.
The important lesson is that the rate ratio compares rates, not the percentage of participants who experience any relapse. The distinction becomes particularly important when participants can experience multiple events.
21. Trial Timeline
Trial start
the ClinicalTrials.gov record lists 2007-06 as the study start.
Randomized treatment study
The trial is identified as a randomized, double-masked, parallel phase 3 study with 1417 enrolled participants.
Annualized relapse rate
The registered primary endpoint was Annualized Relapse Rate with a time frame of 2 years.
Primary completion
the ClinicalTrials.gov record lists 2011-08 as the primary completion date.
Registry status
The current registry-reported ClinicalTrials.gov record identifies the study status as COMPLETED.
22. Related Tutorials
Learn more about the methods used in this trial:
23. Related Calculators
24. Sources
- ClinicalTrials.gov: CONFIRM — NCT00451451. Official registry record used as the source of the ClinicalTrials.gov record and registered endpoint ClinicalTrials.gov record.
- Linked publication: PubMed PMID 37061656.
- Linked publication: PubMed PMID 32426039.
- Linked publication: PubMed PMID 30918100.
- Linked publication: PubMed PMID 28770420.
- Linked publication: PubMed PMID 28751099.
Continue through the Clinical Biostats learning pathway
Use the related tutorials and statistical calculators to explore randomized trial design, recurrent-event outcomes, count models, rate ratios, and safety analysis.
25. Record Summary
CONFIRM is a phase 3, randomized, double-masked, parallel clinical trial in relapsing-remitting multiple sclerosis with 1417 enrolled participants and 4 treatment arms. Its registered primary endpoint was Annualized Relapse Rate over 2 years, a recurrent count/rate outcome defined using protocol-specific criteria and evaluated by an independent neurologic evaluation committee. The registry definition identifies an adjusted annualized relapse rate calculated from negative binomial regression.
From a statistical perspective, the central teaching point is the distinction between a recurrent-event rate and a simple binary outcome. Negative binomial regression provides a framework for modeling relapse counts while accommodating overdispersion, and annualization connects the event count to the observation period. The four-arm structure also makes the prespecified comparison and multiplicity strategy important.
The ClinicalTrials.gov record also provide serious adverse-event counts and denominators for placebo, the two BG00012 dosing groups, Total BG00012, and glatiramer acetate. These are useful descriptive safety quantities, but the registry-reported statistical-analysis field contains no formal comparative analyses.