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
FLAURA2 was a randomized, open-label, parallel phase 3 trial comparing osimertinib with chemotherapy against osimertinib alone as first-line treatment in patients with mutated epidermal growth factor receptor non-small cell lung cancer. The registry reports an enrollment of 587 participants, two randomized treatment groups, three registered primary endpoints, 32 posted outcome measures, and 12 posted statistical analyses.
| Feature | FLAURA2 |
|---|---|
| Phase | Phase 3 |
| Condition | Non-Small Cell Lung Cancer |
| Population | Patients with mutated epidermal growth factor receptor non-small cell lung cancer receiving first-line treatment |
| Design | Randomized, parallel, unmasked |
| Allocation | Randomized |
| Primary purpose | Treatment |
| Enrollment | 587 |
| Registered primary endpoints | 3 |
| Outcome measures posted | 32 |
| Statistical analyses posted | 12 |
| Lead sponsor | AstraZeneca |
| Sponsor type | Industry |
| ClinicalTrials.gov | NCT04035486 |
2. Clinical Question
The central randomized question was whether adding chemotherapy to first-line osimertinib changes progression-free survival compared with osimertinib alone in patients with mutated epidermal growth factor receptor non-small cell lung cancer.
Population
Patients with mutated epidermal growth factor receptor non-small cell lung cancer receiving first-line treatment.
Intervention
Osimertinib with chemotherapy. The registry identifies pemetrexed/carboplatin and pemetrexed/cisplatin as chemotherapy interventions.
Comparator
Osimertinib alone in the randomized component.
Primary question
Does osimertinib with chemotherapy improve progression-free survival relative to osimertinib alone?
3. Trial Design
Osimertinib + chemotherapy
- Osimertinib
- Pemetrexed/carboplatin or pemetrexed/cisplatin
- Randomized component analyzed against osimertinib alone
Osimertinib
- Osimertinib alone
- Randomized component analyzed against osimertinib plus chemotherapy
4. Endpoints
The registry lists three primary endpoints. Two are formal time-to-event efficacy analyses in the randomized component; the third is a safety endpoint for the safety run-in treatment arms.
| Registered primary endpoint | Time frame | Endpoint type |
|---|---|---|
| Adverse Events Graded by Common Terminology Criteria for Adverse Event v5 (Safety Run-In Treatment Arms Only) | From first dose date to 28 days following last dose, up to 45 months | Other / unclear |
| Progression-free Survival (PFS) (Randomized Component) | Up to approximately 33 months after the first patient is randomized (maximum follow up of 33.3 months) | Time-to-event |
| Sensitivity Analysis for Progression-free Survival (PFS) by Blinded Independent Central Review (BICR) Assessment (Randomized Component) | Up to approximately 33 months after the first patient is randomized (maximum follow up of 33.2 months) | Time-to-event |
Progression-free survival definition
The registry defines progression-free survival using Investigator assessment as defined by RECIST 1.1. PFS is the time from randomization until the date of objective disease progression or death by any cause in the absence of progression, regardless of whether the patient withdraws from randomized treatment.
BICR sensitivity analysis
The second efficacy primary endpoint is a sensitivity analysis for PFS using blinded independent central review assessment. The registry again identifies Kaplan-Meier calculation of median PFS and defines the endpoint as the time from randomization to objective disease progression or death in the absence of progression.
5. Analysis Populations and Stratification
The primary randomized efficacy analyses used the Full Analysis Set, which the registry defines as including all randomized participants. This is important because the treatment comparison remains anchored to randomization rather than to whether participants completed treatment.
| Analysis population | Definition / role |
|---|---|
| Full Analysis Set | Includes all randomized participants; used for the randomized-component efficacy analyses. |
| Safety Analysis Set | Includes all participants in the safety run-in treatment arms who received at least 1 dose of study treatment. |
| Subgroup of Full Analysis Set | Used for the reported EGFR mutation subgroup analyses; includes randomized participants meeting the specified mutation subgroup definition. |
Stratified analysis
The randomized efficacy analyses used a log-rank test stratified by three factors: race (Chinese/Asian vs. Non-Chinese/Asian vs. Non-Asian), WHO performance status (0 vs. 1), and method used for tissue testing (central vs. local). These same factors were described in the analysis notes for the reported randomized efficacy comparisons.
6. Statistical Methodology
Kaplan-Meier estimation and time-to-event analysis
The registry specifies Kaplan-Meier estimation for median PFS. This is appropriate for a time-to-event endpoint because participants can be followed for different lengths of time and some participants may not experience progression or death during the available observation period.
At each observed event time, the estimated survival probability is updated according to the number of events and the number of participants at risk immediately before that time.
Log-rank test
The primary PFS analysis and its BICR sensitivity analysis used a stratified log-rank test. The same method was used for the reported randomized-component analyses of overall survival, PFS2, time to first subsequent therapy or death, and time to second subsequent therapy or death.
Cox proportional-hazards model
The registry identifies the Cox proportional-hazards model for the reported EGFR mutation subgroup PFS analyses. The main randomized primary analyses report hazard ratios from the log-rank framework and describe the stratification factors in the analysis notes.
For the randomized comparisons in this record, the registry explicitly states that a hazard ratio less than 1 favors the osimertinib plus chemotherapy treatment arm. A hazard ratio is a relative time-to-event measure; it is not a percentage of patients cured, an absolute risk reduction, or a statement that every individual patient experiences the same proportional change.
Logistic regression
Objective response rate and randomized-component disease control rate were analyzed with logistic regression. The registry states that these analyses were stratified by race, WHO performance status, and method used for tissue testing.
ANCOVA
Depth of response was analyzed using analysis of covariance. The model included baseline tumor size and time from baseline scan to randomization as covariates, with race, WHO performance status, and method used for tissue testing as factors.
Exact binomial analysis
Disease control rate in the safety run-in treatment arms was analyzed using the exact Clopper-Pearson method. This is a binomial confidence-interval approach appropriate for a proportion when the sample is relatively small and an exact rather than normal-approximation interval is desired.
7. Primary Results: Progression-Free Survival
The registry reports a formal superiority analysis of investigator-assessed PFS in the randomized component. The Full Analysis Set included all randomized participants, and the comparison was osimertinib plus chemotherapy versus osimertinib alone.
Primary PFS hazard ratio
95% CI: 0.49–0.79 · P < 0.0001
Two-sided confidence interval; superiority hypothesis.
| Feature | Reported result |
|---|---|
| Endpoint | Progression-free Survival (PFS) (Randomized Component) |
| Analysis population | Full Analysis Set, includes all randomized participants |
| Groups compared | Randomized: AZD9291 + Chemo vs Randomized: AZD9291 |
| Method | Stratified log-rank test |
| Effect measure | Hazard ratio |
| Estimate | 0.62 |
| 95% CI | 0.49–0.79 |
| P-value | <0.0001 |
| Time frame | Up to approximately 33 months after the first patient is randomized; maximum follow up of 33.3 months |
The reported hazard ratio of 0.62 means that the estimated instantaneous rate of progression or death was 38% lower in the osimertinib plus chemotherapy group relative to osimertinib alone, because 1 − 0.62 = 0.38. This is a relative model-based interpretation, not a statement that 38% of patients avoided progression or that every patient experienced exactly a 38% reduction.
The 95% confidence interval of 0.49–0.79 describes statistical uncertainty around the estimated hazard ratio under the analysis framework. It does not describe the range of effects that individual patients experienced.
The P-value <0.0001 addresses the statistical evidence against the null hypothesis used for the superiority analysis. It does not measure the magnitude or clinical importance of the treatment effect. Effect size is described by the hazard ratio and its confidence interval, while absolute event-time information would provide a different perspective.
Because this is a time-to-event analysis, censoring and the proportional-hazards framework are relevant to interpretation. The registry does not provide enough information in the ClinicalTrials.gov record to independently assess the proportional-hazards assumption or reconstruct the Kaplan-Meier curve.
8. Primary Sensitivity Result: BICR-Assessment PFS
The second formal efficacy analysis was a sensitivity analysis of PFS using blinded independent central review assessment. It used the same randomized Full Analysis Set and the same stratification factors.
BICR PFS hazard ratio
95% CI: 0.48–0.80 · P = 0.0002
Two-sided confidence interval; superiority hypothesis.
| Feature | Reported result |
|---|---|
| Endpoint | Sensitivity Analysis for Progression-free Survival (PFS) by Blinded Independent Central Review (BICR) Assessment |
| Analysis population | Full Analysis Set, includes all randomized participants |
| Groups compared | Randomized: AZD9291 + Chemo vs Randomized: AZD9291 |
| Method | Stratified log-rank test |
| Effect measure | Hazard ratio |
| Estimate | 0.62 |
| 95% CI | 0.48–0.80 |
| P-value | 0.0002 |
| Time frame | Up to approximately 33 months after the first patient is randomized; maximum follow up of 33.2 months |
The BICR analysis gives the same point estimate, HR 0.62, as the investigator-assessed primary PFS analysis, while its confidence interval is 0.48–0.80. The two analyses therefore provide closely aligned estimates using different assessment approaches.
The confidence interval indicates uncertainty around the BICR-based estimate; it does not indicate that individual patients have hazards confined to that numerical range. The P-value of 0.0002 is evidence for the prespecified superiority comparison, but it is not a measure of how large or clinically meaningful the effect is.
A sensitivity analysis is useful because it examines whether the principal result depends strongly on the method used to determine progression. Agreement between investigator assessment and BICR assessment can make the statistical interpretation less dependent on a single assessment source, although it does not eliminate all sources of uncertainty or bias.
9. Secondary Results
The registry also posts formal analyses for several secondary endpoints. These results should be distinguished from the two primary randomized PFS analyses.
Overall Survival
Overall survival hazard ratio
95% CI: 0.65–1.24 · P = 0.5238
Two-sided confidence interval; superiority hypothesis.
| Endpoint | Method | Estimate | 95% CI | P-value |
|---|---|---|---|---|
| Overall Survival (OS) | Stratified log-rank | HR 0.90 | 0.65–1.24 | 0.5238 |
| Progression Free Survival 2 (PFS2) | Stratified log-rank | HR 0.70 | 0.52–0.93 | 0.0132 |
| Time to First Subsequent Therapy or Death | Stratified log-rank | HR 0.73 | 0.56–0.94 | 0.0159 |
| Time to Second Subsequent Therapy or Death | Stratified log-rank | HR 0.69 | 0.51–0.93 | 0.0157 |
For all four endpoints, the registry reports the same stratification factors: race, WHO performance status, and method used for tissue testing. Each analysis used the Full Analysis Set and compared osimertinib plus chemotherapy with osimertinib alone.
Objective Response Rate
Objective response rate odds ratio
95% CI: 1.06–2.44 · P = 0.0261
Logistic regression; two-sided confidence interval.
The odds ratio greater than 1 favors the osimertinib plus chemotherapy group according to the registry analysis note. The estimate is an odds ratio rather than a risk ratio, so it should not be interpreted as saying that the probability of response was 61% higher.
Depth of Response
Least-square mean difference
95% CI: −7.44 to 0.72 · P = 0.1067
ANCOVA; two-sided confidence interval.
The outcome was percent change from baseline in tumor diameter. The ANCOVA incorporated baseline tumor size and time from baseline scan to randomization as covariates, together with the reported stratification factors. A difference in least-square means below 0 favors the osimertinib plus chemotherapy group according to the registry analysis note.
Disease Control Rate in the Randomized Component
Disease control rate odds ratio
95% CI: 0.63–2.81 · P = 0.4483
Logistic regression; two-sided confidence interval.
The confidence interval is wide and includes 1. The registry therefore provides substantial uncertainty around the estimated odds ratio for disease control in the randomized component.
Disease Control Rate in the Safety Run-In
| Endpoint | Estimate | 95% CI | Method |
|---|---|---|---|
| Disease Control Rate by Investigator, Safety Run-In Treatment Arms Only | 100.00% | 88.43%–100.00% | Exact Clopper-Pearson |
The analysis population consisted of participants in the safety run-in treatment arms who received at least 1 dose of study treatment. The registry states that the 95% confidence interval was calculated for the percentage of participants with a Response or Stable Disease using the exact Clopper-Pearson method.
10. EGFR Mutation Subgroup PFS Analyses
The registry reports two subgroup analyses of investigator-assessed PFS based on plasma epidermal growth factor receptor mutation status. These are subgroup analyses within the randomized Full Analysis Set rather than separate randomized comparisons.
| EGFR mutation subgroup | Method | HR | 95% CI |
|---|---|---|---|
| Exon 19 Deletion | Cox proportional-hazards model | 0.60 | 0.44–0.83 |
| L858R | Cox proportional-hazards model | 0.63 | 0.44–0.90 |
The subgroup Cox models included treatment, subgroup, and a treatment-by-subgroup interaction term according to the registry analysis notes. For both reported subgroup estimates, a hazard ratio below 1 favors osimertinib plus chemotherapy.
11. Safety Results
The registry identifies serious adverse events separately for the safety run-in and randomized components. Because these groups arise from different parts of the trial design, the affected/at-risk counts should not be combined into a single two-arm randomized safety comparison.
| Trial component | Treatment group | Serious adverse events | At risk |
|---|---|---|---|
| Safety Run-In | AZD9291 + Carboplatin + P | 5 | 15 |
| Safety Run-In | AZD9291 + Cisplatin + Pem | 6 | 15 |
| Randomized | AZD9291 + Chemo | 104 | 276 |
| Randomized | AZD9291 | 53 | 275 |
The randomized-component counts correspond to the affected and at-risk numbers reported by the registry. This page does not calculate an additional comparative risk measure because the ClinicalTrials.gov record identifies the counts but does not provide a formal statistical analysis for this safety comparison.
12. Planned Analysis Versus Reported Analysis
FLAURA2 is unusual in this record because the registry contains both primary efficacy analyses and a primary safety-run-in endpoint without a formal statistical comparison posted for the safety endpoint.
| Endpoint | Registry status | Statistical interpretation |
|---|---|---|
| Safety-run-in adverse events | Results posted; no formal statistical analysis reported | Descriptive safety summaries are appropriate for this endpoint; the ClinicalTrials.gov record does not report a formal comparison. |
| Investigator-assessed PFS | Formal primary analysis posted | Stratified log-rank test with hazard ratio and 95% CI. |
| BICR PFS sensitivity analysis | Formal primary analysis posted | Stratified log-rank test with hazard ratio and 95% CI. |
For a time-to-event endpoint such as PFS, a Kaplan-Meier analysis describes the event-time distribution, while the log-rank test compares the groups and the hazard ratio summarizes the relative event rate under the associated model framework. The registry directly reports these methods for the two formal PFS analyses.
13. Interim Analysis and Alpha Spending
The registry identifies interim analysis / alpha spending as an analysis concept for the reported overall-survival analysis. The ClinicalTrials.gov record also notes that an adjusted confidence interval was computed for the OS analysis at the two-sided 99.84% level.
Why interim analysis matters
When accumulating trial data are examined before the end of follow-up, the statistical framework can account for repeated looks at the data rather than treating each analysis as an entirely new hypothesis test.
Why the OS CI is notable
The registry reports both a 95% CI for the OS hazard ratio and an adjusted confidence interval calculated at the two-sided 99.84% level. These intervals should not be silently substituted for one another.
The ClinicalTrials.gov record does not specify the full alpha-spending function, information fractions, or interim-analysis schedule. Those details are therefore not reproduced here.
14. Statistical Methods Explained
Why was a stratified log-rank test used for PFS?
PFS is a time-to-event endpoint with censoring, so a standard comparison of simple proportions would discard important information about follow-up time. The log-rank test compares the event-time experience between treatment groups while accounting for the timing of events and censoring. In FLAURA2, the analysis was stratified by race, WHO performance status, and tissue-testing method.
What does a hazard ratio of 0.62 mean?
A hazard ratio of 0.62 means that the estimated instantaneous rate of progression or death was 0.62 times that in the comparator group under the analysis framework. Equivalently, the estimated relative reduction in the instantaneous event rate is 38%. It does not mean that 38% of patients were prevented from progressing.
Why is the confidence interval important?
The estimate alone does not show how precisely the treatment effect was estimated. For investigator-assessed PFS, the 95% CI is 0.49–0.79. For the BICR sensitivity analysis it is 0.48–0.80. The intervals communicate uncertainty around their corresponding hazard-ratio estimates and should be considered alongside the point estimates.
Why doesn't the P-value measure effect size?
A P-value describes the statistical evidence against a null hypothesis under the specified analysis. It is affected by both the observed data and the amount of information available. A small P-value does not tell us whether an effect is large, clinically important, or precisely estimated. Those questions require effect estimates, confidence intervals, and appropriate absolute measures.
Why use logistic regression for objective response rate?
Objective response is a binary outcome: a participant either meets the response definition or does not. Logistic regression models the probability of that binary outcome and can produce an odds ratio. In FLAURA2, the registry describes stratification by race, WHO performance status, and tissue-testing method.
Why was ANCOVA used for depth of response?
Depth of response is expressed as percent change from baseline in tumor diameter. ANCOVA permits comparison of adjusted group means while incorporating baseline tumor size and time from baseline scan to randomization as covariates. The registry reports a least-square mean difference of −3.36 with a 95% CI of −7.44 to 0.72.
Why is the safety-run-in DCR analysis different?
The safety-run-in disease-control analysis is a single-group proportion estimate rather than a randomized comparison. The registry therefore uses the exact Clopper-Pearson method to construct a 95% confidence interval around the observed percentage. Its estimate of 100.00% should not be interpreted as proof that the underlying population probability is exactly 100%.
15. Interpreting the Secondary Time-to-Event Endpoints
| Endpoint | HR | 95% CI | P-value | What the estimate describes |
|---|---|---|---|---|
| Overall Survival | 0.90 | 0.65–1.24 | 0.5238 | Relative hazard of death |
| PFS2 | 0.70 | 0.52–0.93 | 0.0132 | Relative hazard for the PFS2 event definition |
| TFST or Death | 0.73 | 0.56–0.94 | 0.0159 | Relative hazard for first subsequent therapy or death |
| TSST or Death | 0.69 | 0.51–0.93 | 0.0157 | Relative hazard for second subsequent therapy or death |
These endpoints describe different stages of the disease and treatment pathway. PFS focuses on the first progression or death event. PFS2 extends the concept farther into subsequent disease control. TFST and TSST incorporate subsequent therapy timing, making them different estimands from either PFS or OS.
The fact that several hazard ratios are below 1 does not make them interchangeable. Each endpoint has a different event definition and therefore answers a different statistical question. The OS analysis in particular has a wider confidence interval than the primary PFS analysis and a reported P-value of 0.5238.
16. Odds Ratios and Mean Differences: A Different Scale
Not every FLAURA2 result is a hazard ratio. The response and depth-of-response analyses illustrate why the effect measure must be interpreted on its own statistical scale.
ORR: odds ratio 1.61
An odds ratio above 1 favors the osimertinib plus chemotherapy group according to the registry. It is a ratio of odds, not a ratio of response probabilities.
DCR: odds ratio 1.33
The confidence interval of 0.63–2.81 spans 1, showing substantial uncertainty around the estimated randomized-component odds ratio.
Depth of response: −3.36
This is a least-square mean difference on the percent-change scale. The negative direction favors osimertinib plus chemotherapy according to the registry.
Run-in DCR: 100.00%
This is a single-group percentage with an exact 95% CI of 88.43%–100.00%, not a treatment-versus-control effect estimate.
17. What the Primary PFS Result Does — and Does Not — Mean
The investigator-assessed PFS hazard ratio of 0.62 corresponds to an estimated 38% lower instantaneous rate of progression or death for osimertinib plus chemotherapy relative to osimertinib alone under the reported analysis.
It does not mean that 38% of patients were cured, that 38% fewer patients experienced progression, or that each participant experienced the same proportional reduction in risk.
The 95% CI of 0.49–0.79 provides the uncertainty interval around the estimated hazard ratio. A narrower interval would generally indicate greater statistical precision, while a wider interval would indicate more uncertainty.
The reported P-value <0.0001 provides evidence against the null hypothesis specified for the superiority analysis. It should not be treated as a measure of effect size or as a substitute for the hazard ratio and confidence interval.
The BICR analysis produced the same point estimate, HR 0.62, with a 95% CI of 0.48–0.80 and P = 0.0002. This provides a second assessment of the PFS comparison using blinded independent central review.
18. Limitations
- Registry-level scope: this analysis is constrained to the ClinicalTrials.gov record. Details not included in that data are not reconstructed from external publications.
- No baseline table reported: the ClinicalTrials.gov record does not contain a full baseline demographic or disease-characteristic table, so no additional baseline comparisons are presented.
- No median PFS values reported: although the registry states that median PFS is calculated using Kaplan-Meier estimation, the ClinicalTrials.gov record does not provide the median PFS estimates themselves.
- Safety endpoint: the safety-run-in primary endpoint has results posted but no formal comparative statistical analysis is reported in the ClinicalTrials.gov record.
- Subgroup uncertainty: the reported EGFR mutation subgroup estimates come from smaller subgroups than the overall randomized population and therefore should be interpreted with corresponding uncertainty.
- Proportional-hazards assumption: hazard-ratio interpretation depends on the time-to-event model framework. The ClinicalTrials.gov record does not provide diagnostics for proportional hazards.
- Multiple analyses: FLAURA2 includes primary and secondary endpoints, a BICR sensitivity analysis, subgroup analyses, and an analysis incorporating interim analysis / alpha spending. Individual P-values should therefore be interpreted in the context of the prespecified statistical framework rather than treated as isolated tests.
- Safety run-in versus randomized population: safety-run-in participants and randomized participants represent different components of the trial and should not be combined indiscriminately.
- Open-label design: the registry identifies masking as none. This is relevant when considering outcomes that can be affected by knowledge of treatment assignment, although the BICR PFS sensitivity analysis provides an independent central assessment of progression.
19. Why This Trial Matters Statistically
FLAURA2 is a useful statistical teaching case because the ClinicalTrials.gov record brings together randomized time-to-event analysis, sensitivity analysis, stratification, logistic regression, ANCOVA, exact binomial inference, subgroup Cox modeling, and interim-analysis concepts within one phase 3 trial.
| Concept | How it appears in FLAURA2 |
|---|---|
| Randomization | Two parallel randomized treatment groups in the randomized component. |
| Full Analysis Set | Primary randomized efficacy analyses include all randomized participants. |
| Kaplan-Meier estimation | Registered PFS endpoint specifies median PFS calculated using the Kaplan-Meier method. |
| Log-rank test | Primary PFS and BICR sensitivity analyses use stratified log-rank testing. |
| Hazard ratio | Primary and secondary time-to-event analyses report hazard ratios. |
| Confidence intervals | Primary and secondary analyses provide two-sided 95% confidence intervals. |
| Stratified analysis | Race, WHO performance status, and tissue-testing method are incorporated into the randomized analyses. |
| Logistic regression | ORR and randomized-component DCR are analyzed using logistic regression. |
| Odds ratio | ORR and DCR treatment effects are reported as odds ratios. |
| ANCOVA | Depth of response uses baseline tumor size and time from baseline scan to randomization as covariates. |
| Exact binomial inference | Safety-run-in DCR uses the exact Clopper-Pearson method. |
| Subgroup analysis | Exon 19 deletion and L858R PFS subgroups use Cox proportional-hazards models. |
| Interim analysis / alpha spending | The reported OS analysis includes interim analysis / alpha spending as an analysis concept. |
20. Related Tutorials
Learn more about the methods used in this trial:
21. Related Calculators
22. Sources
- ClinicalTrials.gov: NCT04035486 — FLAURA2.
- Linked publication: PubMed 42728193.
- Linked publication: PubMed 41104938.
- Linked publication: PubMed 40637621.
- Linked publication: PubMed 40311309.
- Linked publication: PubMed 38042525.
Continue with the underlying statistical methods
Explore the survival-analysis, regression, confidence-interval, and clinical-trial methods represented in the FLAURA2 statistical record.
23. Record Summary
FLAURA2 provides a compact example of several major clinical-trial statistical methods. The randomized primary PFS analysis used a stratified log-rank test and reported a hazard ratio of 0.62 with a 95% CI of 0.49–0.79 and P < 0.0001. The BICR sensitivity analysis produced the same hazard-ratio estimate with a 95% CI of 0.48–0.80 and P = 0.0002. Secondary analyses extend the statistical story to overall survival, PFS2, subsequent-therapy endpoints, objective response, disease control, depth of response, and EGFR mutation subgroups.
The most important statistical distinction is between the different estimands represented in the record. Hazard ratios describe relative event rates over time; odds ratios describe relative odds for binary outcomes; ANCOVA produces an adjusted mean difference; and exact binomial inference describes uncertainty around a single proportion. Keeping these measures on their appropriate scales prevents the common mistake of treating every reported number as if it represented the same underlying treatment effect.