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
contRAst 2 was a randomized, double-blind, phase 3 trial evaluating GSK3196165 (otilimab) in participants with moderately to severely active rheumatoid arthritis who had an inadequate response to conventional synthetic or biologic DMARDs. The registry reports 1,764 enrolled participants, 12 arms, two registered primary endpoints, and seven posted statistical analyses.
| Feature | contRAst 2 |
|---|---|
| Phase | Phase 3 |
| Condition | Arthritis, Rheumatoid |
| Therapeutic area | Immunology & Rheumatology; Musculoskeletal |
| Design | Randomized, double-blind, parallel |
| Allocation | Randomized |
| Primary purpose | Treatment |
| Enrollment | 1,764 |
| Interventions | GSK3196165 (otilimab), tofacitinib, and placebo |
| Status | Terminated |
| Start | June 5, 2019 |
| Primary completion | October 29, 2021 |
| Lead sponsor | GlaxoSmithKline |
2. Clinical Question
The central efficacy question was whether participants receiving GSK3196165 achieved a higher proportion of ACR20 responders at Week 12 than participants receiving pooled placebo in the global cohort, and how the two GSK3196165 doses compared with tofacitinib 5 mg. A later secondary endpoint assessed whether GSK3196165 was non-inferior to tofacitinib on ACR20 at Week 24.
Population
Participants with moderately to severely active rheumatoid arthritis who had an inadequate response to conventional synthetic or biologic DMARDs.
Intervention
GSK3196165, also identified as otilimab, at 90 mg or 150 mg with csDMARD therapy.
Comparators
Pooled placebo with csDMARD therapy and tofacitinib 5 mg with csDMARD therapy.
Primary question
What was the Week 12 ACR20 treatment effect for each active treatment compared with placebo and for GSK3196165 compared with tofacitinib?
3. Trial Design
Otilimab treatment groups
- GSK3196165 90 mg + csDMARD
- GSK3196165 150 mg + csDMARD
- Global and Asia cohort analyses were reported separately where specified.
Placebo and tofacitinib
- Tofacitinib 5 mg + csDMARD
- Pooled placebo + csDMARD
- The registry also reports cohort-specific placebo and active-treatment groups within the 12-arm design.
4. Primary Endpoints
| Endpoint | Time frame | Population / cohort | Statistical role |
|---|---|---|---|
| Percentage of participants with 20% improvement in ACR criteria (ACR20) | Week 12 | Global Cohort | Superiority comparison with placebo; additional comparison with tofacitinib |
| Percentage of participants with 20% improvement in ACR criteria (ACR20) | Week 12 | Asia Cohort | Registered primary endpoint |
ACR20 definition
ACR20 requires a 20% improvement from baseline in Tender Joint Count 68 and Swollen Joint Count 66, together with a 20% improvement in at least 3 of 5 additional measures: Patient's Global Assessment of Arthritis Disease Activity, Physician Global Assessment of Arthritis Disease Activity, Patient Assessment of Arthritis Pain, Health Assessment Questionnaire-Disability Index, and an acute-phase reactant represented by high-sensitivity C-reactive protein.
The registry specifies that Patient's Global Assessment, Physician Global Assessment, and Patient Assessment of Arthritis Pain use visual analogue scales. The Patient's Global Assessment and Physician Global Assessment range from 0 = best to 100 = worst; the pain assessment ranges from 0 = no pain to 100 = most severe pain. HAQ-DI ranges from 0 to 3, with 0 representing least difficulty and 3 extreme difficulty.
5. Statistical Methodology
Logistic regression for binary response
ACR20 is a binary endpoint: each participant either meets the ACR20 response definition at the specified time point or does not. The registry reports logistic regression as the statistical method for the posted analyses. The reported odds ratios were generated from models adjusted for baseline Tender Joint Count 68, baseline Swollen Joint Count 66, and treatment group.
For a treatment comparison, the exponentiated treatment coefficient is an odds ratio. An odds ratio above 1 indicates higher estimated odds of ACR20 response for the first-listed treatment group relative to its comparator; an odds ratio below 1 indicates lower estimated odds.
Covariate adjustment
The global-cohort primary models adjusted for baseline TJC68 and baseline SJC66 as well as treatment group. Covariate adjustment can improve precision when baseline variables are prognostically informative, while the randomized treatment assignment remains the basis for the between-group comparison.
Intention-to-treat analysis
The primary analyses were performed on the ITT set defined in the registry as all randomized participants who received at least one dose of study treatment. Participants were analyzed according to the treatment to which they were randomized.
Multiple imputation
The registry states that multiple imputation was used to handle missing data for the primary Week 12 analyses. This is important because a binary response endpoint requires an analysis value for each participant included in the modeled comparison. Multiple imputation incorporates uncertainty associated with missing responses rather than simply treating every missing observation as a particular outcome.
6. Primary Results: Week 12 ACR20
The registry posts five primary statistical analyses for the global-cohort Week 12 ACR20 endpoint. All five use two-sided 95% confidence intervals and logistic regression with adjustment for baseline TJC68, baseline SJC66, and treatment group.
GSK3196165 90 mg vs pooled placebo
Odds ratio for ACR20 response
95% CI: 1.87–3.53 · P < 0.0001
Week 12 · Global Cohort · ITT population with multiple imputation
An odds ratio of 2.57 means that the estimated odds of achieving ACR20 at Week 12 were 2.57 times as high with GSK3196165 90 mg plus csDMARD as with pooled placebo plus csDMARD under the adjusted logistic regression model.
This does not mean that 2.57 times as many participants responded, nor does it mean that the probability of response was 157% higher. Odds and probabilities are different quantities, particularly when the outcome is not rare.
The 95% CI of 1.87–3.53 describes uncertainty around the estimated odds ratio under the analysis model. It does not describe the range of individual treatment effects.
The P-value of <0.0001 addresses the compatibility of the observed result with the null hypothesis specified for the comparison. It does not measure the size or clinical importance of the effect. The result also reflects the prespecified analysis population, covariate adjustment, and multiple-imputation approach.
GSK3196165 150 mg vs pooled placebo
Odds ratio for ACR20 response
95% CI: 1.85–3.50 · P < 0.0001
Week 12 · Global Cohort · ITT population with multiple imputation
An odds ratio of 2.55 indicates that the estimated odds of ACR20 response at Week 12 were 2.55 times as high for GSK3196165 150 mg plus csDMARD compared with pooled placebo plus csDMARD in the adjusted logistic regression model.
The estimate is an odds ratio, not a risk ratio or risk difference. Without the underlying response probabilities, the odds ratio alone cannot be translated into a percentage-point increase in ACR20 response.
The 95% CI of 1.85–3.50 quantifies statistical precision around the estimated odds ratio. The P-value of <0.0001 addresses the null hypothesis of no difference specified for this comparison and should not be interpreted as a measure of effect magnitude.
Because this is one of several reported treatment comparisons, the interpretation of individual P-values also depends on the trial's multiplicity framework. The registry identifies the hypothesis type as superiority.
Tofacitinib 5 mg vs pooled placebo
Odds ratio for ACR20 response
95% CI: 3.66–7.90 · P < 0.0001
Week 12 · Global Cohort · ITT population with multiple imputation
The reported odds ratio of 5.38 means that the estimated odds of ACR20 response were 5.38 times as high for tofacitinib 5 mg plus csDMARD compared with pooled placebo plus csDMARD under the adjusted model.
This is an odds comparison, so it should not be described as a 438% increase in the probability of response. The absolute difference in response probabilities would require the corresponding probabilities or model-based predicted probabilities.
The 95% CI of 3.66–7.90 indicates uncertainty around the estimated odds ratio. The P-value of <0.0001 is evidence against the specified null hypothesis in this analysis; it is not a measure of how large the treatment effect is.
GSK3196165 90 mg vs tofacitinib 5 mg
Odds ratio for ACR20 response
95% CI: 0.34–0.66 · P < 0.0001
Week 12 · Global Cohort · ITT population with multiple imputation
An odds ratio of 0.48 means that the estimated odds of achieving ACR20 at Week 12 were 0.48 times the odds under tofacitinib 5 mg, comparing GSK3196165 90 mg plus csDMARD with tofacitinib 5 mg plus csDMARD in the adjusted model. Expressed as a relative odds comparison, this corresponds to approximately 52% lower estimated odds for GSK3196165 90 mg because 1 − 0.48 = 0.52.
The statement concerns odds, not a 52-percentage-point difference and not necessarily a 52% lower probability of response. The distinction becomes especially important when response probabilities are substantial.
The 95% CI of 0.34–0.66 provides the statistical uncertainty around the odds ratio. The P-value of <0.0001 tests the specified null hypothesis and does not quantify clinical importance.
GSK3196165 150 mg vs tofacitinib 5 mg
Odds ratio for ACR20 response
95% CI: 0.34–0.66 · P < 0.0001
Week 12 · Global Cohort · ITT population with multiple imputation
An odds ratio of 0.47 means that the estimated odds of ACR20 response with GSK3196165 150 mg plus csDMARD were 0.47 times the corresponding odds with tofacitinib 5 mg plus csDMARD in the adjusted model. In relative-odds terms, this is approximately 53% lower estimated odds because 1 − 0.47 = 0.53.
Again, an odds ratio should not be translated directly into a probability difference. The confidence interval of 0.34–0.66 describes uncertainty around the modeled odds ratio rather than variability among individual patients.
The P-value of <0.0001 addresses the hypothesis of no difference in the modeled odds. It does not indicate that the observed effect is clinically important by itself. Interpretation should also account for the multiple comparisons represented by the overall primary-endpoint testing framework.
| Week 12 Global Cohort comparison | Effect | 95% CI | P-value | Hypothesis |
|---|---|---|---|---|
| GSK3196165 90 mg vs pooled placebo | OR 2.57 | 1.87–3.53 | <0.0001 | Superiority |
| GSK3196165 150 mg vs pooled placebo | OR 2.55 | 1.85–3.50 | <0.0001 | Superiority |
| Tofacitinib 5 mg vs pooled placebo | OR 5.38 | 3.66–7.90 | <0.0001 | Superiority |
| GSK3196165 90 mg vs tofacitinib 5 mg | OR 0.48 | 0.34–0.66 | <0.0001 | Superiority |
| GSK3196165 150 mg vs tofacitinib 5 mg | OR 0.47 | 0.34–0.66 | <0.0001 | Superiority |
7. Secondary Results: Week 24 Non-Inferiority Comparison
The registry also posts two secondary Week 24 analyses comparing GSK3196165 with tofacitinib 5 mg in the global cohort. These analyses use the difference in percentage of ACR20 responders, expressed as GSK3196165 minus tofacitinib. Both analyses use logistic regression adjusted for baseline TJC68, baseline SJC66, and treatment group, with multiple imputation for missing data.
GSK3196165 90 mg vs tofacitinib 5 mg
Difference in ACR20 response percentage
95% CI: −21.3 to −8.1
Week 24 · Global Cohort · Non-inferiority analysis
The reported difference is −14.7 percentage points, defined as the ACR20 response percentage with GSK3196165 90 mg minus the response percentage with tofacitinib 5 mg. Thus, the estimated response percentage was 14.7 percentage points lower for GSK3196165 90 mg under this analysis.
The key feature of a non-inferiority analysis is the prespecified margin. The registry states that non-inferiority would be concluded if the lower limit of the multiplicity-corrected 95% CI for the difference in proportions was greater than −12%.
Here, the reported lower confidence limit is −21.3, which is below the stated −12% margin. Therefore, the reported confidence interval does not satisfy the registry's stated non-inferiority criterion.
This is why non-inferiority cannot be judged simply by asking whether a conventional P-value is above or below 0.05. The relevant question is whether the confidence interval remains entirely above the prespecified acceptable-loss boundary.
GSK3196165 150 mg vs tofacitinib 5 mg
Difference in ACR20 response percentage
95% CI: −23.9 to −10.6
Week 24 · Global Cohort · Non-inferiority analysis
The estimated ACR20 response percentage difference was −17.2 percentage points, again defined as GSK3196165 minus tofacitinib. The estimate therefore represents a lower ACR20 response percentage for GSK3196165 under the reported analysis.
The 95% CI extends from −23.9 to −10.6 percentage points. Because the lower limit is below the prespecified −12% non-inferiority margin, the confidence interval does not satisfy the stated criterion for non-inferiority.
The upper confidence limit of −10.6 is above the margin, but that alone is not sufficient. The entire confidence interval must be on the acceptable side of the non-inferiority margin for the stated criterion to be met.
| Week 24 comparison | Risk difference | 95% CI | Non-inferiority margin | Registry criterion |
|---|---|---|---|---|
| GSK3196165 90 mg vs tofacitinib 5 mg | −14.7% | −21.3% to −8.1% | −12% | Lower CI limit must be > −12% |
| GSK3196165 150 mg vs tofacitinib 5 mg | −17.2% | −23.9% to −10.6% | −12% | Lower CI limit must be > −12% |
8. Safety: Serious Adverse Events
The registry reports serious adverse events by arm and cohort. These are counts of participants affected among those at risk in each listed treatment group. Because the registry provides counts rather than comparative effect estimates for these safety measures, the numbers below are best interpreted descriptively.
Global Cohort
| Group | Participants with serious adverse events | Participants at risk |
|---|---|---|
| GSK3196165 90 mg + csDMARD | 44 | 545 |
| GSK3196165 150 mg + csDMARD | 43 | 539 |
| Tofacitinib 5 mg + csDMARD | 31 | 286 |
| Pooled placebo | 6 | 255 |
| Placebo + csDMARD and GSK3196165 90 mg + csDMARD | 5 | 85 |
| Placebo + csDMARD and GSK3196165 150 mg + csDMARD | 3 | 80 |
| Placebo + csDMARD and tofacitinib 5 mg + csDMARD | 2 | 67 |
Asia Cohort
| Group | Participants with serious adverse events | Participants at risk |
|---|---|---|
| GSK3196165 90 mg + csDMARD | 5 | 47 |
| GSK3196165 150 mg + csDMARD | 4 | 49 |
| Tofacitinib 5 mg + csDMARD | 2 | 19 |
| Pooled placebo | 1 | 23 |
| Placebo + csDMARD and GSK3196165 90 mg + csDMARD | 0 | 6 |
| Placebo + csDMARD and GSK3196165 150 mg + csDMARD | 1 | 8 |
| Placebo + csDMARD and tofacitinib 5 mg + csDMARD | 0 | 8 |
Serious adverse-event counts should not be compared using raw counts alone because the groups have different numbers of participants at risk. The registry's presentation also includes both global and Asia-cohort groups, so the denominator must always be kept with the numerator when describing these safety results.
9. Statistical Methods Explained
Why was logistic regression used?
ACR20 at a fixed time point is binary: a participant either achieves the response criterion or does not. Logistic regression is therefore a natural model for the endpoint. In this trial, the model also allowed adjustment for baseline TJC68 and SJC66 while estimating treatment-group differences.
What does an odds ratio of 2.57 mean?
An odds ratio of 2.57 means the estimated odds of ACR20 response are 2.57 times as large in the GSK3196165 90 mg group as in the pooled placebo group, after the specified adjustment. If the placebo probability were known, odds could be converted back to a probability, but the odds ratio alone does not give the absolute probability difference.
Why is covariate adjustment used?
The primary models adjusted for baseline TJC68 and SJC66. Adjustment incorporates baseline information into the treatment-effect model and can improve precision when these variables are related to the outcome. The interpretation remains tied to the randomized treatment comparison.
Why was multiple imputation used?
Some participants may not have an observed ACR20 response at the scheduled assessment. Multiple imputation creates multiple plausible completed datasets under an assumed missing-data framework, fits the analysis model in each dataset, and combines the resulting estimates. This propagates uncertainty associated with missing observations rather than treating imputed values as known facts.
Why does the non-inferiority margin matter more than a conventional P-value?
The Week 24 question is not simply whether GSK3196165 and tofacitinib differ. The registry defines an acceptable loss of no more than 12 percentage points. The reported confidence interval must therefore be compared directly with the −12% boundary. A conventional superiority P-value does not answer that non-inferiority question.
How should a risk difference be interpreted?
A risk difference is an absolute difference in response probabilities. Here it is defined as GSK3196165 minus tofacitinib. A value of −14.7 means the estimated ACR20 response percentage was 14.7 percentage points lower in the GSK3196165 90 mg group under the reported Week 24 analysis.
10. Primary vs Secondary Hypothesis Frameworks
The statistical story of contRAst 2 contains two distinct comparison frameworks. The Week 12 primary analyses are superiority comparisons expressed primarily as odds ratios. The Week 24 secondary analyses use risk differences and a non-inferiority margin.
| Feature | Week 12 primary analyses | Week 24 secondary analyses |
|---|---|---|
| Endpoint | ACR20 | ACR20 |
| Time point | Week 12 | Week 24 |
| Model | Logistic regression | Logistic regression |
| Effect measure | Odds ratio | Risk difference |
| Hypothesis | Superiority | Non-inferiority |
| Covariates | Baseline TJC68, baseline SJC66, treatment group | Baseline TJC68, baseline SJC66, treatment group |
| Missing data | Multiple imputation | Multiple imputation |
| Key interpretive reference | OR = 1 for no modeled odds difference | −12% non-inferiority margin |
This distinction is statistically important. The null reference for an odds ratio is 1, whereas the relevant reference for the Week 24 non-inferiority analysis is the prespecified −12% margin. The direction of the effect also needs to be read from the analysis definition: the Week 24 difference is explicitly GSK3196165 minus tofacitinib.
11. Intention-to-Treat and Missing Data
The primary Week 12 analyses use an ITT set consisting of randomized participants who received at least one dose of study treatment, analyzed according to randomized treatment. This preserves the randomized comparison rather than redefining groups according to subsequent treatment behavior.
Multiple imputation is explicitly incorporated into the primary analysis. Statistically, this means that the treatment estimate is conditional on the missing-data assumptions underlying the imputation procedure. The resulting confidence interval therefore incorporates uncertainty from the imputed outcomes rather than treating missing responses as automatically equivalent to responders or nonresponders.
It is to propagate uncertainty about missing responses through the treatment-effect analysis.
12. Non-Inferiority Logic in Detail
The Week 24 secondary endpoint provides a useful example of how a non-inferiority trial is interpreted. The registry defines the treatment difference as the ACR20 response percentage for GSK3196165 minus that for tofacitinib and specifies a non-inferiority margin of −12%.
Acceptable region
If the lower limit of the multiplicity-corrected 95% CI were greater than −12%, the registry states that non-inferiority would be concluded.
90 mg estimate
The estimated difference was −14.7 percentage points, with a 95% CI from −21.3 to −8.1 percentage points.
150 mg estimate
The estimated difference was −17.2 percentage points, with a 95% CI from −23.9 to −10.6 percentage points.
Boundary comparison
For both doses, the lower confidence limit is below −12%, so the reported confidence intervals do not satisfy the stated non-inferiority criterion.
The visual logic can be represented simply:
The comparison is based on the entire confidence interval, not merely whether the point estimate lies above or below the margin.
13. Why Odds Ratios and Risk Differences Tell Different Stories
The Week 12 and Week 24 analyses use different effect measures because they answer somewhat different statistical questions.
| Measure | Interpretation | Reference value |
|---|---|---|
| Odds ratio | Ratio of the odds of ACR20 between treatment groups | 1 |
| Risk difference | Absolute difference in ACR20 response proportions | 0 percentage points for equality |
| Non-inferiority margin | Maximum prespecified acceptable loss relative to comparator | −12 percentage points |
An odds ratio can show the relative separation between treatment groups without directly displaying the absolute difference in response probability. A risk difference, by contrast, is expressed directly in percentage points and therefore aligns naturally with a clinically specified non-inferiority margin.
14. Interpreting the Week 12 Results as a Set
The five primary analyses form a structured set of comparisons rather than five unrelated statistical tests. Both GSK3196165 doses were compared with pooled placebo, tofacitinib was compared with pooled placebo, and each GSK3196165 dose was compared directly with tofacitinib.
The important statistical point is that an odds ratio must always be interpreted with its comparison direction. An OR of 2.57 for GSK3196165 versus placebo and an OR of 0.48 for GSK3196165 versus tofacitinib are not contradictory measures: they use different reference groups.
15. Statistical Precision and Confidence Intervals
All five posted primary analyses report two-sided 95% confidence intervals. The intervals provide information that a point estimate alone cannot provide.
| Comparison | Estimate | 95% CI width / range | Statistical reference |
|---|---|---|---|
| 90 mg vs placebo | OR 2.57 | 1.87–3.53 | OR = 1 |
| 150 mg vs placebo | OR 2.55 | 1.85–3.50 | OR = 1 |
| Tofacitinib vs placebo | OR 5.38 | 3.66–7.90 | OR = 1 |
| 90 mg vs tofacitinib | OR 0.48 | 0.34–0.66 | OR = 1 |
| 150 mg vs tofacitinib | OR 0.47 | 0.34–0.66 | OR = 1 |
Confidence intervals should not be read as a range containing 95% of individual patient effects. They describe uncertainty in the estimated population parameter under the statistical model and sampling framework. Their width is affected by information available for the comparison, including sample size and outcome variability.
16. Limitations
- Registry-level reporting: the ClinicalTrials.gov record provides the posted statistical analyses and selected safety counts, but it does not provide the complete participant-level dataset needed to reproduce every analysis.
- Binary endpoint interpretation: odds ratios should not be interpreted as risk ratios or percentage-point differences without the underlying response probabilities.
- Missing data: multiple imputation depends on assumptions about the missing-data mechanism and the imputation model.
- Non-inferiority: the Week 24 conclusion depends on the prespecified −12% margin and the multiplicity-corrected confidence interval described in the registry.
- Multiple comparisons: the primary endpoint includes several treatment comparisons. Individual P-values should therefore be interpreted within the trial's overall multiplicity framework rather than treated as isolated tests.
- Analysis population: the primary ITT definition includes randomized participants who received at least one dose, and participants are analyzed according to randomized treatment.
- Safety denominators: serious adverse-event counts are presented with different numbers at risk across groups and cohorts; raw counts alone are not appropriate for direct rate comparisons.
- Cohort structure: global and Asia-cohort results should not be combined without preserving the cohort definitions used in the registry.
- Endpoint definition: ACR20 is a composite criterion requiring improvement in multiple clinical and laboratory measures, so the endpoint represents a multidimensional response rather than a single measurement.
17. Why This Trial Matters Statistically
contRAst 2 is a useful statistical teaching case because the registry illustrates several important ideas in modern randomized clinical-trial analysis within the same endpoint family. The trial combines randomization and blinding with covariate-adjusted logistic regression, multiple imputation, odds ratios, risk differences, superiority testing, and a prespecified non-inferiority margin.
| Concept | How it appears in contRAst 2 |
|---|---|
| Randomization | Randomized phase 3 parallel-group design |
| Blinding | Double-blind design |
| Binary endpoint | ACR20 response at specified time points |
| Logistic regression | Primary and secondary posted analyses |
| Covariate adjustment | Baseline TJC68 and SJC66 included in the reported models |
| Intention-to-treat | Primary Week 12 analyses performed in the defined ITT set |
| Multiple imputation | Used to handle missing data |
| Odds ratio | Primary Week 12 effect measure |
| Risk difference | Week 24 non-inferiority effect measure |
| Superiority | Primary Week 12 comparisons |
| Non-inferiority | Secondary Week 24 comparisons with a −12% margin |
| Confidence intervals | 95% two-sided intervals for reported treatment effects |
18. Overall Statistical Interpretation
The registry reports odds ratios of 2.57 and 2.55 for GSK3196165 90 mg and 150 mg, respectively, versus pooled placebo, with two-sided 95% confidence intervals of 1.87–3.53 and 1.85–3.50. It also reports an odds ratio of 5.38 for tofacitinib versus pooled placebo and odds ratios of 0.48 and 0.47 for GSK3196165 90 mg and 150 mg, respectively, versus tofacitinib.
The registry reports risk differences of −14.7 and −17.2 percentage points for GSK3196165 90 mg and 150 mg versus tofacitinib, respectively. Their 95% confidence intervals are −21.3 to −8.1 and −23.9 to −10.6. The registry's stated non-inferiority margin is −12%, and the lower limits of both reported confidence intervals are below that margin.
The effect estimates do not by themselves describe absolute ACR20 response probabilities, individual patient benefit, long-term clinical outcomes, or the complete safety profile. Those interpretations require the corresponding outcome distributions, follow-up information, and safety analyses. The posted statistical analyses provide specific treatment comparisons at Week 12 and Week 24 under their defined models and populations.
19. Related Tutorials
Learn more about the methods used in this trial:
20. Related Calculators
21. Sources
- ClinicalTrials.gov: contRAst 2, NCT03970837.
- Linked publication: PubMed record for PMID 37699654.
Continue through the Clinical Biostats knowledge graph
Explore the statistical methods underlying randomized trials, binary endpoints, regression models, confidence intervals, and non-inferiority analyses.
22. Record Summary
contRAst 2 provides a compact example of several important principles in clinical-trial statistics. The Week 12 primary endpoint was a binary ACR20 response analyzed with covariate-adjusted logistic regression in an ITT population with multiple imputation for missing data. The resulting treatment effects were reported as odds ratios for superiority comparisons. The Week 24 secondary endpoint used a risk difference and a prespecified −12% non-inferiority margin, requiring the confidence interval to be evaluated against that boundary rather than against the conventional null value for a superiority test.
The trial also illustrates why statistical interpretation requires more than reading a P-value. Odds ratios require careful attention to the reference group; confidence intervals quantify uncertainty around the estimate; covariate adjustment defines the fitted comparison; multiple imputation affects the missing-data analysis; and non-inferiority depends directly on a clinically specified margin. Together, these elements define the statistical meaning of the reported contRAst 2 results.