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
SOURCE was a randomized, parallel-group, triple-masked phase 3 trial evaluating tezepelumab in adults with oral corticosteroid-dependent asthma. The registered primary endpoint assessed the categorized percent reduction from baseline in daily oral corticosteroid dose at Week 48 while not losing asthma control.
| Feature | SOURCE |
|---|---|
| Trial name | SOURCE |
| ClinicalTrials.gov identifier | NCT03406078 |
| Phase | Phase 3 |
| Condition | Asthma |
| Population | Adults with oral corticosteroid-dependent asthma |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Triple |
| Primary purpose | Treatment |
| Enrollment | 150 |
| Interventions | Tezepelumab; placebo |
| Primary endpoint type | Binary in the registry classification; analyzed using a proportional odds model for the categorized ordinal response |
| Results posted | Yes |
| Statistical analyses posted | 1 |
2. Clinical Question
The central question was whether treatment with tezepelumab, compared with placebo, was associated with greater reduction in daily oral corticosteroid dose while maintaining asthma control over the registered period from baseline to Week 48.
Population
Adults with oral corticosteroid-dependent asthma.
Intervention
Tezepelumab, classified in the registry as a biological intervention.
Comparator
Placebo, classified in the registry as an other intervention.
Primary question
Does tezepelumab increase the odds of achieving a greater categorized reduction in daily oral corticosteroid dose without losing asthma control?
3. Trial Design
Tezepelumab
- Tezepelumab was the biological intervention.
- The primary comparison was against placebo.
- The primary endpoint assessed categorized oral corticosteroid-dose reduction.
Placebo
- Placebo was the comparator intervention.
- The primary analysis compared its response distribution with that of the tezepelumab group.
- The endpoint was assessed over the same baseline-to-Week-48 period.
4. Trial Timeline
Study start
The SOURCE trial began on March 5, 2018.
Primary completion
The registry lists September 25, 2020 as the primary completion date.
Final registry status
The trial is listed as completed, with results posted on ClinicalTrials.gov.
5. Primary Endpoint
| Endpoint | Registry definition | Time frame | Analysis |
|---|---|---|---|
| Categorized Percent Reduction From Baseline in the Daily OCS Dose While Not Losing Asthma Control | Categorized percent reduction from baseline at Week 48. Percent change from baseline is defined as {final dose-baseline dose)/baseline dose}*100, with categories of ≥90% to ≤100% reduction, ≥75% to <90% reduction, ≥50% to <75% reduction, >0% to <50% reduction, and no change or any increase. | Baseline to Week 48 | Proportional odds model with covariate adjustment |
The endpoint is fundamentally a ranked response outcome. Instead of reducing corticosteroid response to a simple yes/no classification, the registry defines ordered categories according to the magnitude of dose reduction. That ordering is what makes a proportional odds model appropriate to the reported analysis.
The five response categories
| Category | Definition |
|---|---|
| Highest reduction | ≥90% to ≤100% reduction |
| High reduction | ≥75% to <90% reduction |
| Moderate reduction | ≥50% to <75% reduction |
| Lower reduction | >0% to <50% reduction |
| No reduction / increase | No change or any increase |
The endpoint also includes the requirement of not losing asthma control. Thus, the statistical target is not simply corticosteroid reduction in isolation; the registered endpoint explicitly combines dose reduction with preservation of asthma control.
6. Statistical Methodology
Proportional odds regression
The registry reports a proportional odds model as the method for the primary endpoint. This is an ordinal regression framework designed for outcomes with a natural ordering, such as the five categories of oral corticosteroid-dose reduction used here.
The model estimates cumulative odds across ordered thresholds. The proportional-odds assumption means that the treatment effect is represented by a common odds ratio across the modeled category boundaries.
For SOURCE, the response variable was the categorized percent reduction from baseline in final oral corticosteroid dose. The registry reports treatment, region, and daily oral corticosteroid dose at baseline as covariates.
Covariate adjustment
The reported model included treatment, region, and daily oral corticosteroid dose at baseline. This means the treatment comparison was not simply a raw comparison of the observed response categories. Instead, the model estimated the treatment association while accounting for these specified covariates.
Treatment
Identifies the randomized comparison between tezepelumab and placebo.
Region
Accounts for the region variable specified in the registry analysis model.
Baseline OCS dose
Accounts for the daily oral corticosteroid dose measured at baseline.
Outcome structure
Uses the ordering of the corticosteroid-reduction categories rather than treating them as unrelated nominal groups.
Cumulative odds ratio
The reported effect measure is a cumulative odds ratio. An odds ratio above 1 indicates higher modeled odds of being in a more favorable portion of the ordered response distribution, assuming the proportional-odds model is appropriate and the direction of the coding corresponds to greater reduction.
The meaning depends on the ordinal coding and cumulative-probability convention used by the analysis. The registry reports the cumulative odds ratio but does not provide the full model coding specification in the ClinicalTrials.gov record.
7. Primary Result
The registry reports one formal statistical analysis for the primary endpoint. It compared tezepelumab with placebo using a proportional odds model with treatment, region, and baseline daily oral corticosteroid dose as covariates.
Cumulative odds ratio
95% CI: 0.69–2.35 · P = 0.434
Tezepelumab vs placebo · Baseline to Week 48
| Primary endpoint | Comparison | Method | Effect | 95% CI | P-value |
|---|---|---|---|---|---|
| Categorized Percent Reduction From Baseline in the Daily OCS Dose While Not Losing Asthma Control | Tezepelumab vs placebo | Proportional odds model | Cumulative OR 1.28 | 0.69–2.35 | 0.434 |
The estimated cumulative odds ratio of 1.28 indicates that the fitted ordinal model estimated higher cumulative odds for the more favorable categorized response distribution with tezepelumab relative to placebo, under the model's coding and proportional-odds assumptions.
It does not mean that 28% more participants achieved the endpoint, that the probability of success increased by 28%, or that each participant experienced a 28% reduction in oral corticosteroid dose. An odds ratio is not a risk ratio, probability difference, or percent-change measure.
The 95% confidence interval of 0.69–2.35 is relatively broad compared with the point estimate. It represents uncertainty around the estimated model parameter under the statistical framework; it is not the range of individual treatment effects and does not mean that individual participants' outcomes must lie between 0.69 and 2.35.
The P-value of 0.434 measures the compatibility of the observed result with the null hypothesis under the specified model and testing framework. It does not measure the size of the treatment effect, the clinical importance of the effect, or the probability that the null hypothesis is true.
Because the endpoint is ordinal, interpretation also depends on the proportional-odds assumption. That assumption states that the treatment effect can be represented by a common odds ratio across the cumulative category thresholds. The registry analysis does not report a separate diagnostic assessment of that assumption.
8. How to Read the Primary Endpoint
The most important statistical feature of SOURCE is that the primary endpoint contains ordered categories. This changes how the treatment effect should be interpreted.
| Question | What the SOURCE endpoint does | What it does not do |
|---|---|---|
| Magnitude of OCS reduction | Places participants into ordered categories according to percent reduction. | Does not report the primary endpoint as one continuous average percentage. |
| Asthma control | Requires the categorized reduction to occur while not losing asthma control. | Does not treat corticosteroid reduction as independent of asthma control. |
| Treatment effect | Uses a cumulative odds ratio from an ordinal regression model. | Does not represent a simple risk ratio or percentage-point difference. |
| Adjustment | Includes treatment, region, and baseline daily OCS dose as covariates. | Is not merely an unadjusted comparison of category frequencies. |
9. Statistical Methods Explained
Why was proportional odds regression used?
The primary outcome is categorized into five ordered levels of oral corticosteroid-dose reduction. A proportional odds model is designed for exactly this type of ordinal response because it can use the ordering information rather than treating every category as unrelated.
What does a cumulative odds ratio of 1.28 mean?
A cumulative odds ratio of 1.28 represents the estimated treatment association on the cumulative-odds scale. Subject to the proportional-odds assumption and the model's response coding, an odds ratio above 1 corresponds to higher odds of a more favorable ordered response with tezepelumab. It is not equivalent to saying that the probability of achieving a particular category increased by 28%.
Why is the confidence interval important?
The 95% confidence interval of 0.69–2.35 shows the uncertainty surrounding the estimated cumulative odds ratio. Because the interval spans 1, the reported estimate is compatible with both a lower and a higher cumulative-odds association under the specified statistical framework.
Why does the P-value not measure effect size?
The P-value of 0.434 is a measure used for hypothesis testing. It does not tell us whether an effect is large or small. Effect magnitude is described by the odds ratio, while the confidence interval provides information about the precision of that estimate.
Why adjust for baseline oral corticosteroid dose?
Participants may begin the trial at different daily oral corticosteroid doses. The registry specifies baseline daily OCS dose as a model covariate, allowing the treatment comparison to account statistically for that baseline measure rather than relying only on an unadjusted category comparison.
What does the proportional-odds assumption add?
The model summarizes treatment effects across multiple cumulative thresholds with a common odds ratio. That is statistically efficient when the assumption is reasonable, but the interpretation becomes less straightforward if treatment effects differ substantially across the individual category thresholds. The ClinicalTrials.gov record does not report a separate test or diagnostic for this assumption.
Why is randomization important?
Randomization is the design feature that establishes the treatment groups before outcomes are observed. It supports a causal comparison of assigned treatments by reducing systematic differences between groups in expectation. Regression adjustment can improve precision or account for prespecified covariates, but it does not replace the fundamental role of randomization.
10. Confidence Interval and Hypothesis Testing
Point estimate
The cumulative odds ratio was 1.28. This is the single best-fitting estimate reported by the model.
Uncertainty
The 95% CI was 0.69–2.35, indicating substantial uncertainty around the point estimate.
Hypothesis test
The reported P-value was 0.434 for the superiority analysis.
Interpretive boundary
An odds ratio of 1 represents equal cumulative odds under the relevant null formulation. The confidence interval crosses that value.
The three quantities answer different questions. The odds ratio describes the estimated magnitude and direction on the model scale. The confidence interval describes uncertainty around that estimate. The P-value addresses the statistical test. None of these quantities alone describes the full clinical experience of individual participants.
11. Safety
The ClinicalTrials.gov record reports serious adverse events by randomized arm as affected participants divided by participants at risk.
| Safety measure | Tezepelumab | Placebo |
|---|---|---|
| Serious adverse events | 12/74 | 16/76 |
These counts describe the number of participants with serious adverse events and the corresponding number at risk in each arm. The ClinicalTrials.gov record does not provide a formal statistical comparison of serious adverse events, confidence interval, or P-value, so the safety counts should not be converted into a comparative hypothesis-test conclusion.
12. Statistical Analysis Record
| Analysis characteristic | Reported value |
|---|---|
| Endpoint role | Primary |
| Outcome measure | Categorized Percent Reduction From Baseline in the Daily OCS Dose While Not Losing Asthma Control |
| Time frame | Baseline to Week 48 |
| Outcome unit | Participants |
| Endpoint type | Binary in the registry classification |
| Groups compared | Tezepelumab vs Placebo |
| Method as reported | Proportional odds model |
| Effect measure | Cumulative Odds Ratio |
| Estimate | 1.28 |
| Confidence interval | 95% CI 0.69–2.35, two-sided |
| P-value | 0.434 |
| Hypothesis type | Superiority |
| Covariates | Treatment, region, and daily OCS dose at baseline |
The registry lists the primary endpoint as binary in its endpoint-type classification, while the actual definition contains five ordered categories and the posted statistical method is proportional odds regression. For statistical interpretation, the ordered categorical structure and the reported proportional odds model are the more informative features.
13. What the Primary Result Does — and Does Not — Mean
The cumulative odds ratio of 1.28 is a model-based estimate comparing the ordered corticosteroid-reduction response between tezepelumab and placebo. Under the proportional-odds framework, it summarizes the treatment association across cumulative response thresholds.
It does not mean that tezepelumab produced a 28% reduction in oral corticosteroid dose, a 28-percentage-point increase in responders, or a 28% increase in the probability of success. Odds and probabilities are different quantities.
The 95% CI of 0.69–2.35 quantifies uncertainty around the estimated cumulative odds ratio. Its width indicates that the point estimate should not be treated as a highly precise measurement of the underlying treatment association.
The P-value of 0.434 is evidence from the specified hypothesis test, not a measure of the size or clinical importance of the treatment effect. It should be read alongside the effect estimate and confidence interval.
14. Design Features and What They Contribute
| Design feature | Role in interpretation |
|---|---|
| Randomization | Creates the randomized treatment comparison between tezepelumab and placebo. |
| Parallel design | Participants remain in their randomized treatment groups rather than serving as their own crossover controls. |
| Triple masking | Reduces the potential for knowledge of treatment assignment to influence trial conduct or assessment. |
| Phase 3 | Places the trial within a late-stage confirmatory clinical-development framework. |
| Covariate adjustment | The reported primary model adjusts for treatment, region, and baseline daily OCS dose. |
| Ordinal regression | Uses the ordered structure of the corticosteroid-reduction categories. |
The ClinicalTrials.gov record does not report a non-inferiority margin, crossover design, factorial structure, Bayesian methodology, or interim-analysis method. Those features are therefore not incorporated into this analysis.
15. Multiplicity and Analysis Scope
The ClinicalTrials.gov record identifies 1 primary endpoint and 1 posted statistical analysis. The analysis in the ClinicalTrials.gov record is the formal primary analysis reported in the trial data.
Primary endpoint
One registered primary endpoint focused on categorized daily OCS-dose reduction while not losing asthma control.
Formal analysis
One statistical analysis was posted for that endpoint.
Hypothesis
The reported hypothesis type was superiority.
Multiplicity information
The ClinicalTrials.gov record does not report a multiplicity-adjustment procedure beyond the primary analysis information provided.
Because the ClinicalTrials.gov record does not specify a larger confirmatory hierarchy, alpha-spending scheme, or multiplicity adjustment, no such procedure is inferred here.
16. Limitations
- Single reported formal analysis: the ClinicalTrials.gov record contains one statistical analysis for the primary endpoint. Other posted outcome measures are not accompanied by formal statistical analyses in the ClinicalTrials.gov record.
- Ordinal-model assumption: proportional odds regression relies on a common treatment-effect structure across cumulative thresholds. The ClinicalTrials.gov record does not report a separate assessment of this assumption.
- Confidence-interval width: the 95% CI of 0.69–2.35 is substantially wider than the point estimate alone and should be considered when assessing precision.
- Endpoint-type classification: the registry classifies the endpoint as binary, while its definition contains five ordered response categories and the posted analysis is proportional odds regression. The statistical interpretation therefore follows the actual reported model and endpoint definition.
- Incomplete safety inference: serious adverse event counts are reported by arm, but the ClinicalTrials.gov record does not provide a formal comparative safety analysis.
- No subgroup results reported: subgroup estimates and interaction analyses are not part of the ClinicalTrials.gov record and are not inferred.
- No interim-analysis information reported: the ClinicalTrials.gov record does not specify an interim monitoring method or alpha-spending procedure.
- No missing-data method reported: the registry-reported statistical analysis does not specify an imputation method or detailed handling of missing endpoint observations.
- No non-inferiority framework: the hypothesis type is superiority, so a non-inferiority margin is not part of the reported primary analysis.
17. Why This Trial Matters Statistically
SOURCE is a useful teaching case because its primary endpoint illustrates a problem that appears frequently in clinical-trial statistics: outcomes may contain clinically meaningful ordered categories rather than a single binary event or continuous measurement.
| Concept | How it appears in SOURCE |
|---|---|
| Randomization | Participants were randomized between tezepelumab and placebo. |
| Blinding | The registry specifies triple masking. |
| Ordinal outcome | The primary endpoint uses five ordered categories of percent OCS-dose reduction. |
| Proportional odds regression | The primary endpoint was analyzed using a proportional odds model. |
| Odds ratio | The treatment effect was reported as a cumulative odds ratio of 1.28. |
| Confidence interval | The reported 95% CI was 0.69–2.35. |
| P-value | The reported P-value was 0.434. |
| Covariate adjustment | The model included treatment, region, and baseline daily OCS dose. |
| Superiority testing | The registry identifies the hypothesis type as superiority. |
| Safety analysis | Serious adverse events were reported as 12/74 for tezepelumab and 16/76 for placebo. |
The broader statistical lesson is that an endpoint's measurement scale should guide the analysis. If response categories are naturally ordered, discarding that ordering by using an inappropriate nominal or binary analysis can lose information. Conversely, an ordinal model introduces assumptions—especially the proportional-odds assumption—that must be understood when interpreting the result.
18. A Worked Statistical Reading of the SOURCE Analysis
Step 1: Identify the outcome scale
The primary endpoint ranks participants according to the extent of daily oral corticosteroid reduction while not losing asthma control. The categories run from very large reductions through smaller reductions to no change or an increase.
Step 2: Identify the comparison
The treatment comparison is tezepelumab versus placebo. Because allocation was randomized, the comparison is anchored to the randomized treatment groups.
Step 3: Identify the model
The registry reports a proportional odds model. The model incorporates treatment, region, and baseline daily oral corticosteroid dose.
Step 4: Identify the effect measure
The effect measure is the cumulative odds ratio. The estimate is 1.28.
Step 5: Quantify uncertainty
The 95% confidence interval is 0.69–2.35. Because the interval includes 1, the data are compatible with no difference on the cumulative-odds scale under the specified model.
Step 6: Read the hypothesis test separately
The P-value is 0.434. This is evidence for the specified statistical test, not a measurement of clinical magnitude.
Step 7: Return to the clinical endpoint
The statistical result must ultimately be interpreted in terms of the actual endpoint: categorized reduction in daily oral corticosteroid dose while not losing asthma control. An odds ratio should never be substituted for the clinical endpoint itself.
19. What Is Not Reported in the registry Registry Analysis
The ClinicalTrials.gov record does not report formal statistical results for the remaining posted outcome measures. They also do not provide subgroup estimates, baseline characteristics, median time-to-event outcomes, forest plots, a missing-data or imputation strategy, or an interim-analysis procedure.
20. Sources
- ClinicalTrials.gov: SOURCE, NCT03406078.
- PubMed record: PMID 36702146.
- PubMed record: PMID 35364018.
- PubMed record: PMID 33050928.
21. Related Tutorials
Learn more about the methods used in this trial:
22. Related Calculators
Continue through the Clinical Biostats statistical tutorials
Connect the trial's endpoint and regression method to broader explanations of odds ratios, confidence intervals, covariate adjustment, randomization, and ordinal regression.
23. Record Summary
SOURCE is a randomized, parallel-group, triple-masked phase 3 trial enrolling 150 participants with asthma and evaluating tezepelumab versus placebo. Its registered primary endpoint measured categorized percent reduction from baseline in daily oral corticosteroid dose while not losing asthma control from baseline to Week 48. The posted primary analysis used a proportional odds model with treatment, region, and baseline daily oral corticosteroid dose as covariates.
The reported treatment effect was a cumulative odds ratio of 1.28, with a 95% confidence interval of 0.69–2.35 and a P-value of 0.434. Statistically, the key lesson is not to translate that odds ratio directly into a percentage change in probability or corticosteroid dose. Instead, it should be interpreted as an ordinal-regression treatment effect, together with its confidence interval and the assumptions underlying proportional odds regression.
The trial also provides a useful example of why endpoint construction matters. The outcome is not simply whether corticosteroid dose decreased. It ranks the magnitude of reduction across ordered categories and simultaneously requires that asthma control not be lost. That structure motivates the proportional odds analysis and makes careful interpretation of the cumulative odds ratio essential.