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
FACT OA2 was a randomized, parallel-group, triple-masked, phase 3 trial evaluating fasinumab against placebo and nonsteroidal anti-inflammatory drugs in adults with pain from osteoarthritis of the knee or hip. The registry reports 1650 participants, 4 arms, 18 posted outcome measures, and 6 posted statistical analyses.
| Feature | FACT OA2 |
|---|---|
| Phase | Phase 3 |
| Conditions | Osteoarthritis, Knee; Osteoarthritis, Hip |
| Design | Randomized, parallel-group, triple-masked |
| Primary purpose | Treatment |
| Enrollment | 1650 |
| Interventions | Fasinumab; diclofenac; celecoxib; matching placebo |
| Primary endpoints | Change from baseline in WOMAC Pain and WOMAC Physical Function scores up to Week 24 |
| Hypothesis type | Superiority |
| Primary endpoint type | Binary, as recorded in the registry analysis data |
| Statistical methods posted | MMRM and Cochran-Mantel-Haenszel test |
| Results | Posted |
2. Clinical Question
The central efficacy question was whether fasinumab produced greater improvement in osteoarthritis-related pain and physical function than placebo through Week 24. The primary analyses reported in the registry specifically compared placebo with fasinumab 1 mg.
Population
Adults with pain from osteoarthritis of the knee or hip.
Intervention
Fasinumab, including the 1 mg Q4W arm used in the posted primary efficacy analyses.
Comparator
Matching placebo for the primary efficacy comparisons reported in the registry analyses.
Primary question
Does treatment with fasinumab improve WOMAC pain and physical-function scores compared with placebo from baseline through Week 24?
3. Trial Design
Fasinumab treatment
- Fasinumab 1 mg Q4W
- Fasinumab 3 mg Q4W
- Fasinumab 6 mg Q8W
Placebo and NSAIDs
- Matching placebo
- Diclofenac
- Celecoxib
The registry describes the overall study as a 4-arm trial, while the listed intervention information includes fasinumab, diclofenac, celecoxib, and matching placebo. The posted primary statistical analyses examine only placebo versus fasinumab 1 mg.
4. Trial Timeline
Study start
The FACT OA2 trial began.
Primary completion
The registry reports primary study completion on this date.
5. Primary Endpoints
The trial had two registered primary endpoints, both assessed from baseline up to Week 24. Both use the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC), but they measure different dimensions of osteoarthritis-related symptoms.
| Endpoint | Registry definition | Time frame |
|---|---|---|
| WOMAC Pain | Change From Baseline in the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) Pain Subscale Scores up to Week 24 in Participants Treated With Fasinumab Compared to Placebo | Baseline up to Week 24 |
| WOMAC Physical Function | Change From Baseline in WOMAC Physical Function Subscale Scores up to Week 24 in Participants Treated With Fasinumab Compared to Placebo | Baseline up to Week 24 |
WOMAC Pain
The WOMAC pain subscale is a 5-item questionnaire assessing the amount of pain experienced because of osteoarthritis in the index knee or hip during the past 48 hours. Each question uses a numerical rating scale from 0, meaning no pain, to 10, with higher pain. The subscale is calculated as the mean of the 5 individual question scores.
WOMAC Physical Function
The WOMAC physical-function subscale contains 17 items assessing the degree of difficulty experienced because of osteoarthritis in the index knee or hip during the past 48 hours. Each question is scored from 0, representing minimum difficulty, to 10, representing maximum difficulty. The subscale is the mean of the 17 individual question scores.
6. Statistical Methodology
Mixed Model for Repeated Measures
The primary continuous-score analyses used a mixed-effect model with repeated measures, or MMRM. This is appropriate for longitudinal outcomes because the same participants contribute measurements over multiple visits and the analysis can model treatment differences while accounting for the repeated structure of the observations.
The registry describes analyses using a multiple-imputation approach with baseline, randomization strata, treatment, visit, and treatment-by-visit interaction in the MMRM framework.
Least-squares mean difference
The primary results are reported as least-squares mean differences. A least-squares mean is a model-adjusted mean rather than simply the arithmetic mean observed in each treatment group. The corresponding difference summarizes the model-based separation between the two randomized treatment groups over the specified analysis framework.
Multiple imputation
The registry states that the MMRM analyses were based on a multiple-imputation approach. Missing outcome information is therefore handled within a prespecified modeling and imputation framework rather than by simply discarding every participant who lacks a measurement at a particular visit.
Cochran-Mantel-Haenszel analysis
The secondary binary responder endpoint was analyzed using the Cochran-Mantel-Haenszel test. This approach provides a stratified comparison of categorical outcomes and is commonly used when the analysis needs to account for stratification factors rather than treating all observations as coming from a single unstratified 2-by-2 table.
Odds ratio
The responder analysis reports an odds ratio. An odds ratio above 1 indicates higher odds of achieving the specified response in the fasinumab group than in the placebo group. It is not the same as a risk ratio or a difference in response probabilities.
Superiority testing
The primary hypotheses were superiority hypotheses. The objective was to determine whether the fasinumab treatment produced a statistically distinguishable improvement relative to placebo rather than merely demonstrating that the treatments were sufficiently close to one another.
7. Analysis Populations
The registry makes an important distinction between the randomized study population and the participants evaluable for a particular endpoint.
| Population | Registry description | Role in analysis |
|---|---|---|
| Modified full analysis set | All randomized participants in the full analysis set, excluding participants from 4 sites for which there were potential concerns regarding data quality identified prior to database lock. | Used for the specified primary mFAS analyses. |
| Full analysis set | For the responder endpoint, included all randomized participants excluding participants affected by the urgent safety measure and analyzed according to randomized treatment. | Used for the categorical responder analysis. |
| Endpoint-evaluable participants | The registry notes that the number analyzed represents participants evaluable for the particular outcome measure. | Defines the denominator for the posted analysis rather than automatically equating it with total enrollment. |
The distinction is statistically important. A randomized trial can enroll one number of participants while the number contributing to a particular model or endpoint is smaller because of prespecified population definitions, data-quality exclusions, safety-related exclusions, or missing evaluable information.
8. Primary Results: WOMAC Pain
The registry reports two closely related primary MMRM analyses of change from baseline in the WOMAC Pain Subscale through Week 24, both comparing placebo with fasinumab 1 mg.
Primary MMRM analysis
Least-squares mean difference
95% CI: -0.971 to -0.286 · P = 0.0003
Placebo vs Fasinumab 1 mg · MMRM · Baseline up to Week 24
The negative estimate indicates that the modeled change in WOMAC pain score was lower with fasinumab 1 mg than with placebo under the analysis framework reported in the registry. The 95% confidence interval remains below 0, extending from -0.971 to -0.286.
What the estimate means: the LS mean difference of -0.63 represents a model-adjusted difference in the WOMAC pain outcome between fasinumab 1 mg and placebo. Given the direction of the WOMAC scale, the negative value favors lower pain scores with fasinumab.
What it does not mean: it is not a 0.63-point change experienced by every participant, nor is it a statement that every patient had the same treatment response. It is a population-level model estimate.
What the confidence interval says: the 95% CI of -0.971 to -0.286 quantifies statistical uncertainty around the estimated mean difference under the specified model and analysis population. Because the interval does not include 0, it is consistent with a statistically detectable difference under a two-sided 95% confidence framework.
What the p-value says: P = 0.0003 addresses the statistical evidence against the null hypothesis of no modeled difference. It does not measure the size, clinical importance, or probability that an individual participant will benefit.
Model caution: the result depends on the longitudinal MMRM and multiple-imputation framework, including treatment, visit, treatment-by-visit interaction, baseline, and randomization strata. The registry does not provide enough information here to independently assess every modeling assumption.
Modified full analysis set analysis
Least-squares mean difference
95% CI: -1.154 to -0.383 · P > 0.0001
Placebo vs Fasinumab 1 mg · mFAS · MMRM
The mFAS analysis gives an LS mean difference of -0.77, again indicating a lower modeled WOMAC pain score with fasinumab 1 mg than placebo. Its 95% CI of -1.154 to -0.383 is entirely below 0.
The confidence interval is useful because it describes the precision of the estimated treatment difference. It does not describe the range of individual patient responses. The reported P > 0.0001 provides evidence against a zero mean difference under the reported test, but the p-value itself should not be interpreted as a measure of treatment magnitude.
The fact that this estimate is -0.77 rather than -0.63 reflects the different analysis specification/population described by the registry. These estimates should therefore not be treated as two independent randomized experiments or averaged together.
9. Primary Results: WOMAC Physical Function
The second primary endpoint evaluated change from baseline in the WOMAC Physical Function Subscale through Week 24, again comparing placebo with fasinumab 1 mg.
Primary MMRM analysis
Least-squares mean difference
95% CI: -0.981 to -0.290 · P = 0.0003
Placebo vs Fasinumab 1 mg · MMRM · Baseline up to Week 24
What the estimate means: the LS mean difference of -0.64 represents the model-adjusted difference in physical-function score between fasinumab 1 mg and placebo. Because higher scores indicate greater difficulty, the negative estimate corresponds to lower modeled difficulty with fasinumab.
Precision: the 95% CI is -0.981 to -0.290. The interval stays below 0, indicating that the estimated difference is separated from the no-difference value within this confidence framework.
P-value: P = 0.0003 quantifies evidence against a zero treatment difference under the reported statistical test. It does not quantify the magnitude or clinical importance of the treatment effect.
Interpretive caution: because the endpoint is longitudinal and analyzed using MMRM with multiple imputation, the result should be understood as a model-based treatment comparison rather than a simple comparison of two observed Week 24 means.
Modified full analysis set analysis
Least-squares mean difference
95% CI: -1.198 to -0.443 · P < 0.0001
Placebo vs Fasinumab 1 mg · mFAS · MMRM
The mFAS analysis estimates a difference of -0.82. The 95% CI of -1.198 to -0.443 is entirely below 0, so the estimated treatment difference is directionally consistent with lower physical-function difficulty in the fasinumab group.
P < 0.0001 indicates strong statistical evidence against the null hypothesis specified for the analysis. It does not mean that the probability of benefit for an individual participant is greater than 99.99%, and it does not establish a particular magnitude of clinical importance.
As with the pain endpoint, the mFAS result incorporates the registry's stated population definition and multiple-imputation MMRM framework. The estimate therefore represents the output of a specified longitudinal analysis rather than a raw observed mean difference.
10. Summary of Primary Efficacy Results
| Endpoint | Analysis | Estimate | 95% CI | P-value |
|---|---|---|---|---|
| WOMAC Pain | MMRM | -0.63 | -0.971 to -0.286 | = 0.0003 |
| WOMAC Pain | mFAS MMRM | -0.77 | -1.154 to -0.383 | > 0.0001 |
| WOMAC Physical Function | MMRM | -0.64 | -0.981 to -0.290 | = 0.0003 |
| WOMAC Physical Function | mFAS MMRM | -0.82 | -1.198 to -0.443 | < 0.0001 |
Across the four posted primary analyses, all estimated differences are negative and all reported 95% confidence intervals exclude 0. The registry therefore reports statistically detectable differences favoring lower WOMAC scores for fasinumab 1 mg relative to placebo under each of these specified analyses.
11. Secondary Endpoint Results
WOMAC Pain: ≥30% Reduction From Baseline
The registry also reports the percentage of participants achieving a greater than or equal to 30% reduction from baseline through Week 24 in the WOMAC Pain Subscale score. For this analysis, efficacy data from fasinumab 1 mg Q4W and placebo were included as prespecified.
Odds ratio for achieving ≥30% reduction
95% CI: 1.195 to 2.092 · P = 0.0013
Cochran-Mantel-Haenszel analysis · Placebo vs Fasinumab 1 mg
An odds ratio of 1.581 means that the estimated odds of achieving at least a 30% reduction in WOMAC pain were higher in the fasinumab 1 mg group than in the placebo group under the reported Cochran-Mantel-Haenszel analysis.
The odds ratio is not a percentage-point difference in response rates. It also should not automatically be interpreted as a 58.1% increase in the probability of response. Odds and probabilities are related but are not numerically interchangeable.
The 95% CI of 1.195 to 2.092 indicates uncertainty around the odds-ratio estimate. Because the interval is entirely above 1, it is consistent with higher odds of response in the fasinumab group under the reported two-sided confidence framework.
The reported P = 0.0013 measures evidence against the null odds ratio of 1; it does not measure the clinical magnitude of the responder difference.
Patient Global Assessment
A second posted secondary analysis examined change from baseline in Patient Global Assessment score through Week 24 using MMRM.
Least-squares mean difference
95% CI: -0.28 to -0.009 · P = 0.0365
Placebo vs Fasinumab 1 mg · MMRM
The estimated LS mean difference of -0.14 indicates a lower modeled Patient Global Assessment score with fasinumab 1 mg than placebo under the reported analysis. The 95% CI, -0.28 to -0.009, lies below 0 but approaches the null value closely at its upper boundary.
The reported P = 0.0365 is below 0.05, but its interpretation depends on the trial's prespecified hierarchical testing procedure. The registry states that secondary endpoints were tested sequentially and that testing continued only when the preceding endpoint was statistically significant at the 0.05 level.
The effect estimate itself should be considered separately from the p-value. Statistical significance does not establish that an estimated difference is clinically important, and the relatively narrow numerical scale of a model-adjusted difference does not by itself tell us whether participants would perceive the change as meaningful.
12. Multiplicity and Hierarchical Testing
The registry states that a hierarchical testing procedure was used to control type I error and handle multiple secondary endpoint analyses. Secondary endpoints were tested sequentially in the order reported, and the hierarchy continued only when the preceding endpoint was statistically significant at the 0.05 level.
| Feature | Registry-reported approach | Statistical implication |
|---|---|---|
| Primary hypothesis type | Superiority | Tests whether fasinumab differs from placebo in the specified direction of the endpoint analysis. |
| Secondary endpoint testing | Hierarchical testing | Controls the interpretation of sequential secondary hypothesis tests within the stated hierarchy. |
| Continuation criterion | Previous endpoint statistically significant at 0.05 | A failure at an earlier step stops the formal sequential testing process. |
| Binary responder analysis | Cochran-Mantel-Haenszel | Provides the reported stratified categorical comparison and odds ratio. |
Multiplicity is important because each additional hypothesis test creates another opportunity to obtain a small p-value by chance alone. A hierarchical procedure addresses that problem by defining an ordered testing strategy before interpreting the sequence of results.
13. Statistical Methods Explained
Why was MMRM used for the WOMAC outcomes?
WOMAC pain and physical-function scores are measured longitudinally from baseline through Week 24. MMRM is designed for repeated measurements because it models the outcome across visits while accounting for the correlation among measurements from the same participant. The registry specifically reports treatment, visit, treatment-by-visit interaction, baseline, and randomization strata within the model.
What does an LS mean difference of -0.63 mean?
It means that the model estimates a difference of -0.63 between the fasinumab 1 mg and placebo groups for the analyzed WOMAC pain outcome. Because higher WOMAC pain scores indicate more pain, the negative direction corresponds to lower modeled pain with fasinumab. It is a model-adjusted group comparison, not an individual-level prediction.
Why are there two WOMAC pain estimates?
The registry reports both an MMRM analysis and an mFAS MMRM analysis. The mFAS excludes participants from 4 sites for which there were potential data-quality concerns identified before database lock. Because the analysis population differs, the resulting estimates need not be identical.
What does the odds ratio of 1.581 mean?
The odds ratio compares the odds of achieving at least a 30% reduction in WOMAC pain between fasinumab 1 mg and placebo. An odds ratio of 1.581 indicates higher estimated odds in the fasinumab group. It is not equivalent to a 58.1 percentage-point increase in the response rate.
Why use the Cochran-Mantel-Haenszel test?
The Cochran-Mantel-Haenszel approach provides a stratified analysis of categorical outcomes. It is useful when treatment groups need to be compared while accounting for the study's stratified analysis structure rather than relying solely on an unadjusted comparison.
Why does the p-value not measure effect size?
A p-value measures the compatibility of the observed data with a specified null hypothesis under the statistical model. It depends on both the size of the estimated effect and the amount of information available. A very small p-value can accompany a modest effect when precision is high, while a larger p-value can accompany a larger estimate when uncertainty is greater.
Why does multiplicity matter?
When many hypotheses are tested, the probability of obtaining at least one apparently positive result by chance increases unless the testing strategy accounts for the multiplicity. FACT OA2 used a hierarchical procedure for the reported secondary analyses, with sequential testing continuing only after the preceding endpoint met the specified 0.05 criterion.
14. Confidence Intervals and Effect Size
The confidence intervals provide important information that p-values alone cannot convey. For the primary WOMAC analyses, the intervals quantify the uncertainty around the estimated LS mean differences.
| Analysis | Estimate | 95% confidence interval | Interpretation of interval |
|---|---|---|---|
| WOMAC Pain, MMRM | -0.63 | -0.971 to -0.286 | Entire interval is below 0. |
| WOMAC Pain, mFAS | -0.77 | -1.154 to -0.383 | Entire interval is below 0. |
| WOMAC Physical Function, MMRM | -0.64 | -0.981 to -0.290 | Entire interval is below 0. |
| WOMAC Physical Function, mFAS | -0.82 | -1.198 to -0.443 | Entire interval is below 0. |
The important statistical distinction is between precision and clinical importance. A confidence interval can be relatively narrow while the underlying effect remains of uncertain practical importance. Clinical interpretation requires understanding the measurement scale and the context in which the difference would matter to patients.
A 95% confidence interval is an interval estimate generated by a statistical procedure. It is not a statement that there is a 95% probability that the fixed treatment effect lies inside this particular interval. The interval communicates how compatible different parameter values are with the observed data and the assumptions of the analysis procedure.
15. Missing Data and Multiple Imputation
The registry explicitly states that the MMRM analyses used a multiple-imputation approach. Missing longitudinal measurements are a central issue in repeated-measures clinical trials because participants may have incomplete follow-up or may not contribute measurements at every visit.
Multiple imputation handles missing information by creating multiple plausible completed datasets under a specified missing-data model, analyzing each dataset, and combining the resulting estimates according to standard rules. The resulting uncertainty reflects both the sampling variability in the observed data and uncertainty associated with the missing values.
Why this matters
Simply analyzing participants with complete observations can change the target population and may introduce bias when missingness is related to observed or unobserved characteristics.
What the model does not guarantee
Multiple imputation does not automatically eliminate missing-data bias. Its validity depends on the assumptions and model used to describe the missing observations.
The registry identifies multiple imputation as part of the reported analysis but does not provide sufficient detail here to evaluate the full missing-data sensitivity-analysis framework.
16. Randomization and Masking
Randomization is the foundation of the causal comparison because treatment assignment is determined by the study design rather than by the participant's baseline characteristics or subsequent response.
Participants were assigned within a randomized parallel-group design.
The registry identifies the study as triple masked.
The intervention groups were evaluated as parallel treatment arms.
The registry identifies treatment as the primary purpose of the study.
Masking can reduce the influence of expectations on participant-reported outcomes such as pain and global assessment. This is especially relevant when outcomes are subjective and depend on participant responses to questionnaires.
17. Safety Results
The registry reports serious adverse events by treatment arm. The available counts include placebo, the combined NSAID category, and the three fasinumab dose schedules.
| Arm | Serious adverse events | At risk |
|---|---|---|
| Placebo | 20 | 309 |
| NSAIDs | 45 | 609 |
| Fasinumab 1 mg Q4W | 35 | 609 |
| Fasinumab 3 mg Q4W | 5 | 58 |
| Fasinumab 6 mg Q8W | 2 | 59 |
These are reported affected/at-risk counts. They should not be collapsed into a single overall safety comparison because the denominators differ and the trial contains multiple treatment arms with different intervention schedules.
The at-risk denominator is essential to interpreting these counts. For example, 35 affected participants in a group with 609 at risk and 5 affected participants in a group with 58 at risk represent different denominators, so raw event counts alone are not directly comparable measures of incidence.
18. Interpreting the Primary Findings
All four posted primary MMRM estimates are negative: -0.63, -0.77, -0.64, and -0.82. Because higher WOMAC scores indicate greater pain or difficulty, the direction of these estimates corresponds to lower modeled symptom scores with fasinumab 1 mg relative to placebo.
The reported confidence intervals for all four primary analyses exclude 0, and the corresponding p-values are = 0.0003, > 0.0001, = 0.0003, and < 0.0001. These results provide statistical evidence of differences under the respective analysis specifications.
A statistically detectable difference does not by itself establish how meaningful the change is to an individual patient. The WOMAC scale, the magnitude of the estimated difference, the uncertainty interval, and the responder analysis provide different perspectives on treatment effect.
The continuous WOMAC analyses estimate differences in modeled mean scores, while the ≥30% responder analysis asks whether participants achieved a predefined amount of improvement. These endpoints complement one another rather than representing interchangeable effect measures.
19. Important Limitations and Interpretation Issues
- Analysis population: the primary analyses are not simply based on all 1650 enrolled participants. The registry describes endpoint-evaluable populations and an mFAS excluding participants from 4 sites because of potential data-quality concerns identified before database lock.
- Endpoint type: the registry's analysis metadata labels the primary endpoints as binary even though the posted primary estimates are LS mean differences for WOMAC score changes. The analysis method and effect measure reported in the statistical-analysis records are the more informative descriptions of the posted results.
- Multiple analyses: two primary endpoints and additional analyses of those endpoints require careful interpretation of the testing framework rather than treating every posted estimate as an independent confirmatory hypothesis.
- Multiple imputation: the MMRM analyses rely on a multiple-imputation approach. Interpretation therefore depends partly on assumptions concerning missing outcome information.
- Hierarchical testing: secondary endpoint interpretation depends on the sequential testing procedure described in the registry.
- Odds ratio interpretation: an odds ratio is not a risk ratio or a percentage-point response difference. Translating it into an absolute treatment effect requires the underlying response probabilities.
- Subjective outcomes: WOMAC pain, physical function, and Patient Global Assessment are participant-reported measures, making appropriate masking particularly relevant to interpretation.
- Safety denominators: serious adverse-event counts are reported with different at-risk denominators across the treatment arms, so raw counts should not be interpreted as equivalent incidence measures.
20. Why This Trial Matters Statistically
FACT OA2 is a useful statistical teaching case because it combines randomized treatment allocation, masking, longitudinal patient-reported outcomes, mixed models, multiple imputation, categorical responder analysis, odds ratios, stratified testing, and hierarchical multiplicity control within one phase 3 study.
| Concept | How it appears in FACT OA2 |
|---|---|
| Randomization | Randomized parallel-group phase 3 design with 1650 participants. |
| Blinding | Triple masking is reported in the registry. |
| Repeated measures | WOMAC pain and physical-function outcomes were analyzed through Week 24. |
| MMRM | Primary continuous-score analyses used mixed-effect models with repeated measures. |
| Multiple imputation | The MMRM analyses were based on a multiple-imputation approach. |
| Mean difference | Primary analyses reported LS mean differences. |
| Cochran-Mantel-Haenszel test | The ≥30% WOMAC pain responder analysis used a Cochran-Mantel-Haenszel model. |
| Odds ratio | The responder analysis reported an OR of 1.581. |
| Confidence intervals | Primary and secondary estimates were accompanied by two-sided 95% confidence intervals. |
| Multiplicity | Secondary endpoints were evaluated using a hierarchical testing procedure. |
21. Related Tutorials
Learn more about the methods used in this trial:
22. Related Calculators
23. Sources
- ClinicalTrials.gov: NCT03304379.
- Publication: PubMed record 40001000.
- Publication: PubMed record 39004211.
Continue through the Clinical Biostats statistical library
Explore related tutorials and statistical calculators for longitudinal outcomes, categorical endpoints, confidence intervals, missing data, and multiplicity.
24. Record Summary
FACT OA2 illustrates how a randomized phase 3 trial can combine longitudinal continuous outcomes with categorical responder endpoints. The primary WOMAC pain and physical-function analyses used MMRM with multiple imputation and reported LS mean differences, while the ≥30% WOMAC pain responder endpoint used a Cochran-Mantel-Haenszel analysis with an odds ratio. The posted estimates consistently favor lower WOMAC scores with fasinumab 1 mg relative to placebo, with two-sided 95% confidence intervals that exclude the null value for all four posted primary analyses.
The statistical interpretation is strongest when the treatment estimates are considered together with their confidence intervals, analysis populations, missing-data approach, and multiplicity framework. The secondary analyses add a complementary responder perspective, while the serious-adverse-event counts provide a separate safety dimension that must be interpreted using the corresponding at-risk denominators.