This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record. Numerical trial results on this page are restricted to the ClinicalTrials.gov record.
1. Trial at a Glance
SERENA-2 was a randomized, parallel, unmasked phase 2 treatment trial enrolling 240 post-menopausal women at least 18 years of age with advanced ER-positive HER2-negative breast cancer. The trial compared oral AZD9833 at three dose levels with intramuscular fulvestrant 500 mg, with progression-free survival as the single registered primary endpoint.
| Feature | SERENA-2 |
|---|---|
| Phase | Phase 2 |
| Condition | Advanced ER-Positive HER2-Negative Breast Cancer |
| Design | Randomized, parallel |
| Masking | None |
| Allocation | Randomized |
| Primary purpose | Treatment |
| Enrollment | 240 |
| Arms | 4 |
| Primary endpoint | Progression-free Survival (PFS) |
| Primary endpoint type | Time-to-event |
| Primary hypothesis | Superiority |
| Results posted | Yes |
| Statistical analyses posted | 6 |
| Lead sponsor | AstraZeneca |
| Trial status | Active, not recruiting |
| ClinicalTrials.gov | NCT04214288 |
2. Clinical Question
The central statistical question was whether oral AZD9833 produced a longer progression-free survival than intramuscular fulvestrant 500 mg in the trial population. The registered primary hypothesis type was superiority, and the formal primary analyses compared AZD9833 75mg and AZD9833 150mg separately with fulvestrant 500 mg.
Population
Post-menopausal women at least 18 years of age with advanced ER-positive HER2-negative breast cancer.
Intervention
Oral AZD9833, represented in the four-arm design by 75mg, 150mg, and 300mg dose groups.
Comparator
Intramuscular fulvestrant 500 mg.
Primary question
Does AZD9833 improve progression-free survival relative to fulvestrant 500 mg?
3. Trial Design
Oral AZD9833
- 75mg dose group
- Primary comparison: vs fulvestrant 500 mg
- Primary endpoint: progression-free survival
Oral AZD9833
- 150mg dose group
- Primary comparison: vs fulvestrant 500 mg
- Primary endpoint: progression-free survival
Oral AZD9833
- 300mg dose group
- Recruitment stopped early
- 20 patients randomized
Intramuscular fulvestrant
- 500 mg comparator group
- Primary comparator for AZD9833 75mg and 150mg
- Primary endpoint: progression-free survival
4. Trial Population and Analysis Populations
The trial enrolled 240 patients. The primary PFS analyses used the full analysis set (FAS), defined in the registry as all randomized patients, with treatment groups assigned according to randomization regardless of the actual treatment received.
| Population | Registry definition / role |
|---|---|
| Full analysis set (FAS) | All randomized patients, with treatment groups assigned in accordance with randomization, regardless of the actual treatment received. |
| ORR analysis population | A subset of the FAS with measurable disease. |
| CBR24 analysis population | FAS, with treatment groups assigned according to randomization regardless of actual treatment received. |
This distinction is important. The primary PFS analysis is anchored to randomized assignment rather than to whether patients actually received the assigned treatment. That preserves the treatment comparison created by randomization within the FAS.
5. Endpoints
| Endpoint | Registry definition / time frame | Type |
|---|---|---|
| Progression-free Survival (PFS) | From date of randomisation to date of objective disease progression (or last evaluable assessment in the absence of progression) or death (up to data cut-off of 29 months) | Time-to-event |
| Objective Response Rate (ORR) | From screening until disease progression (up to data cut-off of 29 months) | Binary |
| Clinical Benefit Rate at 24 Weeks (CBR24) | At Week 24 | Binary |
Primary endpoint: Progression-free Survival
The registered primary endpoint was PFS. The registry definition states that PFS was assessed by the Investigator according to RECIST version 1.1 and defined as time from randomisation until objective disease progression or death by any cause in the absence of progression, regardless of whether the patient withdrew from randomized therapy or received another anti-cancer therapy before progression.
The registry time-frame field posted on ClinicalTrials.gov for the endpoint is: “From date of randomisation to date of objective disease progression (or last evaluable assessment in the absence of progression) or death (up to data cut-off of 29 months)”.
Secondary endpoints
The posted statistical analyses include two binary secondary outcomes: ORR and CBR24. ORR was evaluated from screening until disease progression, with a data cut-off of 29 months. CBR24 was evaluated at Week 24.
6. Statistical Methodology
Primary PFS comparison: stratified log-rank and Cox analysis
The registry reports log-rank testing for both primary PFS comparisons. The accompanying analysis notes specify that the analysis was performed using a stratified Cox proportional-hazards model. The statistical method is therefore a survival-analysis framework combining a stratified comparison with a model-based hazard ratio.
The hazard ratio summarizes the relative event hazard estimated by the Cox model. A value below 1 was specified in the registry as favoring AZD9833 for longer progression-free survival than fulvestrant.
Stratified analysis
The primary analyses are identified as stratified. Stratification allows the time-to-event comparison to account for prespecified grouping factors when estimating the treatment effect. The ClinicalTrials.gov record does not identify the specific stratification variables, so this page does not name or infer them.
Logistic regression for binary endpoints
ORR and CBR24 were analyzed using logistic regression. This is appropriate for binary outcomes because the model describes the relationship between treatment and the probability of an event such as response or clinical benefit.
For the ORR analyses, the registry states that the logistic regression model adjusted for prior use of CDK4/6 inhibitors and presence of lung and/or liver metastasis.
The same covariate adjustment was reported for the CBR24 analyses. The effect measure was an odds ratio, with an odds ratio greater than 1 specified as favoring AZD9833.
Confidence intervals
The posted primary PFS analyses report 90% confidence intervals for the hazard ratios. The posted ORR analyses also report 90% confidence intervals, while the CBR24 75mg comparison explicitly identifies its confidence interval as two-sided.
A confidence interval should be read alongside the point estimate. It does not describe the range of outcomes for individual patients; it describes uncertainty around the estimated treatment effect under the relevant statistical framework.
Superiority hypothesis
The primary comparisons were registered as superiority analyses. This means the statistical question was whether AZD9833 produced evidence of a better outcome than fulvestrant, rather than whether the treatments were sufficiently similar to satisfy a non-inferiority margin.
7. Primary Results: Progression-Free Survival
Two primary PFS analyses were posted: AZD9833 75mg versus fulvestrant 500 mg and AZD9833 150mg versus fulvestrant 500 mg. Both used the FAS and a stratified survival-analysis framework.
AZD9833 75mg vs Fulvestrant 500 mg
Hazard ratio for progression-free survival
90% CI: 0.42–0.82 · P = 0.0167
Stratified log-rank comparison; stratified Cox proportional-hazards model.
The hazard ratio of 0.59 is below 1, consistent with the registry's specified direction favoring AZD9833 for longer progression-free survival than fulvestrant. Expressed as a simple relative interpretation, 0.59 corresponds to an estimated hazard that is about 41% lower under the fitted hazard-ratio framework.
The 0.59 hazard ratio is a relative time-to-event measure. It does not mean that 41% of patients avoided progression, that each patient experienced exactly a 41% reduction in risk, or that median PFS was 41% longer.
The 90% confidence interval of 0.42–0.82 communicates the precision of the estimated hazard ratio. It remains below 1 across the reported interval, so the interval is consistent with a treatment effect in the direction specified by the analysis.
The P-value of 0.0167 addresses evidence against the null hypothesis within the reported statistical framework; it does not measure the size or clinical importance of the treatment effect. Effect size is described by the hazard ratio and its confidence interval.
Because the effect was estimated with a Cox proportional-hazards model, interpretation of a single hazard ratio depends on the model's proportional-hazards framework. The ClinicalTrials.gov record does not report a separate assessment of that assumption.
AZD9833 150mg vs Fulvestrant 500 mg
Hazard ratio for progression-free survival
90% CI: 0.46–0.89 · P = 0.0090
Stratified log-rank comparison; stratified Cox proportional-hazards model.
The hazard ratio of 0.64 is below 1 and therefore points in the registry-specified direction favoring AZD9833 for longer progression-free survival. A hazard ratio of 0.64 corresponds to an estimated hazard about 36% lower under the fitted model.
The 0.64 estimate is not an absolute probability and should not be translated into a statement that 36% of patients benefited or that every patient had a 36% lower personal risk of progression or death.
The 90% confidence interval of 0.46–0.89 gives the reported uncertainty around the estimated hazard ratio. Its upper limit remains below 1, while the width of the interval shows that the estimate is not known with perfect precision.
The P-value of 0.0090 quantifies evidence against the relevant null hypothesis under the reported analysis. It is not a measure of effect magnitude, probability that the treatment works, or probability that the null hypothesis is true.
As with the 75mg comparison, the estimate comes from a stratified Cox proportional-hazards model. The hazard-ratio interpretation is therefore model-based and should be distinguished from an absolute risk difference or a difference in median survival.
| Primary PFS comparison | Effect measure | 90% CI | P-value | Analysis population |
|---|---|---|---|---|
| AZD9833 75mg vs Fulvestrant 500 mg | HR 0.59 | 0.42–0.82 | 0.0167 | FAS |
| AZD9833 150mg vs Fulvestrant 500 mg | HR 0.64 | 0.46–0.89 | 0.0090 | FAS |
8. Secondary Results: Objective Response Rate
ORR was analyzed in a subset of the FAS with measurable disease. The registry reports logistic regression adjusted for prior use of CDK4/6 inhibitors and presence of lung and/or liver metastasis.
| Comparison | Odds ratio | 90% CI | P-value | Direction specified by registry |
|---|---|---|---|---|
| AZD9833 75mg vs Fulvestrant 500 mg | 1.42 | 0.63–3.31 | 0.4828 | OR > 1 favors AZD9833 |
| AZD9833 150mg vs Fulvestrant 500 mg | 1.57 | 0.69–3.67 | 0.3691 | OR > 1 favors AZD9833 |
AZD9833 75mg vs Fulvestrant 500 mg
The reported odds ratio was 1.42, with a 90% confidence interval of 0.63–3.31 and a P-value of 0.4828. Because the confidence interval spans 1, the reported interval includes both an odds ratio below 1 and an odds ratio above 1.
AZD9833 150mg vs Fulvestrant 500 mg
The reported odds ratio was 1.57, with a 90% confidence interval of 0.69–3.67 and a P-value of 0.3691. Again, the confidence interval spans 1, so the estimate is compatible with a range of treatment effects in either direction under the reported model.
9. Secondary Results: Clinical Benefit Rate at 24 Weeks
CBR24 was evaluated at Week 24 using logistic regression. The analysis population was the FAS, and the model adjusted for prior use of CDK4/6 inhibitors and presence of lung and/or liver metastasis.
| Comparison | Odds ratio | 90% CI | P-value |
|---|---|---|---|
| AZD9833 75mg vs Fulvestrant 500 mg | 1.48 | 0.84–2.64 | 0.2554 |
| AZD9833 150mg vs Fulvestrant 500 mg | 1.62 | 0.91–2.89 | 0.1658 |
The reported odds ratios were 1.48 for the 75mg comparison and 1.62 for the 150mg comparison. The corresponding 90% confidence intervals were 0.84–2.64 and 0.91–2.89, respectively. Both intervals include 1.
As with ORR, the odds ratios describe the modeled odds of the binary outcome rather than an absolute probability difference. The CBR24 endpoint is also tied to a fixed time point, Week 24, unlike PFS, which is a time-to-event endpoint.
10. Safety Results
The ClinicalTrials.gov record reports serious adverse events by arm as affected patients divided by patients at risk.
| Arm | Serious adverse events | Affected / at risk |
|---|---|---|
| AZD9833 75mg | Serious adverse events | 6 / 74 |
| AZD9833 150mg | Serious adverse events | 7 / 73 |
| AZD9833 300mg | Serious adverse events | 2 / 20 |
| Fulvestrant 500 mg | Serious adverse events | 4 / 73 |
The reported denominators show the at-risk populations for this safety measure. They sum to 240 across the four arms, consistent with the total trial enrollment.
The bar display is intentionally based on the affected-patient counts reported in the ClinicalTrials.gov record rather than converting them into newly calculated percentages. The 300mg arm remains particularly difficult to interpret because only 20 patients were randomized and the registry explicitly cautions that its data are highly variable.
11. Statistical Methods Explained
Why was a log-rank test used for PFS?
PFS is a time-to-event endpoint, so the timing of progression or death matters rather than simply whether an event eventually occurred. A log-rank test compares the observed pattern of events over follow-up between treatment groups while accounting for censored observations. In SERENA-2, the registry identifies the primary PFS method as a stratified log-rank analysis.
What does a hazard ratio of 0.59 mean?
A hazard ratio of 0.59 means that the estimated hazard under the fitted Cox model was 0.59 times the comparator hazard. In the registry's interpretation, a hazard ratio below 1 favors AZD9833 for longer PFS. It does not mean that the treatment group had a 59% probability of remaining progression-free or that every patient's risk fell by exactly 41%.
Why was a stratified Cox model used?
The registry explicitly states that the primary PFS analyses were performed using a stratified Cox proportional-hazards model. Stratification permits the treatment-effect estimation to account for the prespecified strata rather than treating all patients as if they belonged to one homogeneous risk set. The ClinicalTrials.gov record does not identify the specific stratification variables.
Why was logistic regression used for ORR and CBR24?
ORR and CBR24 are binary endpoints. Logistic regression models the odds of the binary outcome and can incorporate adjustment variables. For SERENA-2, the reported models adjusted for prior CDK4/6 inhibitor use and presence of lung and/or liver metastasis.
What does an odds ratio of 1.57 mean?
An odds ratio of 1.57 means that the modeled odds of the binary outcome were estimated to be 1.57 times the comparator odds, under the specified logistic regression model. It does not mean that the probability of the outcome was 57% higher. The distinction matters because odds and probabilities are not interchangeable.
Why does the confidence interval matter more than the point estimate alone?
A point estimate such as HR 0.64 is only one estimate of the underlying treatment effect. The accompanying 90% confidence interval of 0.46–0.89 shows the statistical uncertainty reported around that estimate. A wider interval indicates less precision than a narrower interval, while the location of the interval relative to the null value provides information about the direction and strength of evidence under the specified model.
Why should the 300mg arm be interpreted cautiously?
The registry states that recruitment to the 300mg arm was stopped early at 20 patients randomized and that the resulting 300mg data are highly variable. A small sample provides less information for estimating an effect and can produce unstable estimates. This is a design and precision issue rather than evidence by itself that the dose is effective or ineffective.
12. Reading the Primary PFS Results Together
| Feature | AZD9833 75mg vs fulvestrant | AZD9833 150mg vs fulvestrant |
|---|---|---|
| Hazard ratio | 0.59 | 0.64 |
| 90% confidence interval | 0.42–0.82 | 0.46–0.89 |
| P-value | 0.0167 | 0.0090 |
| Direction specified by registry | HR < 1 favors AZD9833 | HR < 1 favors AZD9833 |
| Model | Stratified Cox proportional-hazards | Stratified Cox proportional-hazards |
| Primary population | FAS | FAS |
Both primary estimates are below 1. The 75mg comparison has an estimated hazard ratio of 0.59, while the 150mg comparison has an estimated hazard ratio of 0.64. The confidence intervals for both comparisons remain below 1, and both reported P-values are below 0.05.
That pattern describes the statistical results; it does not establish that one dose is superior to the other. The ClinicalTrials.gov record does not provide a direct 75mg-versus-150mg comparison, and the two estimates have overlapping uncertainty intervals. A cross-dose conclusion therefore should not be inferred from their separate comparisons with fulvestrant.
The reported hazard ratios provide relative measures of the PFS treatment effect. The ClinicalTrials.gov record does not provide median PFS, time-specific PFS probabilities, or an absolute difference in PFS.
13. Why the P-Values Do Not Measure Effect Size
SERENA-2 illustrates an important distinction between statistical evidence and effect magnitude. The P-values for the primary PFS comparisons are 0.0167 and 0.0090, but neither number describes how large the treatment effect is.
Effect size
The hazard ratios, 0.59 and 0.64, quantify the estimated relative treatment effect under the Cox model.
Precision
The 90% confidence intervals, 0.42–0.82 and 0.46–0.89, describe uncertainty around the corresponding hazard-ratio estimates.
Evidence against the null
The P-values, 0.0167 and 0.0090, quantify evidence against the relevant null hypothesis within the reported analysis.
Clinical magnitude
A P-value alone cannot establish clinical importance. Absolute outcome measures would provide a different perspective, but the ClinicalTrials.gov record does not report them.
14. Interpreting Odds Ratios in the Secondary Analyses
The secondary analyses use odds ratios rather than hazard ratios because ORR and CBR24 are binary outcomes. The direction is explicitly defined in the registry: an odds ratio greater than 1 favors AZD9833.
An OR above 1 indicates higher modeled odds of the binary outcome with AZD9833; an OR below 1 indicates lower modeled odds. The null value is 1.
For ORR, the estimates were 1.42 and 1.57 for the 75mg and 150mg comparisons, respectively. For CBR24, the estimates were 1.48 and 1.62. However, all four reported confidence intervals include 1, so the point estimates should not be read without their uncertainty intervals.
The regression adjustment is also important. These are not simple unadjusted ratios of observed response frequencies. They are model-based odds ratios from logistic regression adjusted for prior CDK4/6 inhibitor use and presence of lung and/or liver metastasis.
15. Relationship Between PFS, ORR, and CBR24
The three reported efficacy outcomes answer different statistical questions.
| Endpoint | Question answered | Statistical scale |
|---|---|---|
| PFS | How does the time until objective progression or death compare between randomized groups? | Time-to-event; hazard ratio |
| ORR | How do the odds of objective response compare among patients with measurable disease? | Binary; odds ratio |
| CBR24 | How do the odds of clinical benefit at Week 24 compare? | Binary; odds ratio |
A treatment can have different results across these endpoints because they measure different features of disease control. PFS incorporates the timing of progression or death and censoring, whereas ORR focuses on a binary response classification and CBR24 focuses on a fixed Week 24 assessment.
Consequently, the primary PFS results should not be numerically combined with the ORR or CBR24 odds ratios. They are different estimands on different statistical scales.
16. Limitations and Interpretation Issues
- 300mg arm size: the 300mg AZD9833 arm stopped recruitment early at 20 patients randomized, and the registry states that its data are highly variable.
- Incomplete registry time-frame text:This page does not reconstruct the missing registry wording.
- No reported absolute PFS measures in the ClinicalTrials.gov record: the registry extract provides hazard ratios and confidence intervals but does not provide median PFS or time-specific PFS probabilities. These should not be derived from the hazard ratios alone.
- Hazard-ratio interpretation: the PFS estimates come from a stratified Cox proportional-hazards model. A hazard ratio is a model-based relative measure and is not equivalent to a risk ratio or absolute risk difference.
- Proportional-hazards assumption: the ClinicalTrials.gov record identifies the Cox model but does not report a separate diagnostic assessment of proportional hazards. The assumption is therefore a model consideration rather than something that can be independently confirmed from the registry extract.
- Secondary endpoint population: ORR was analyzed in a subset of the FAS with measurable disease, so its analysis population differs from the FAS used for the primary PFS comparisons.
- Adjusted secondary analyses: ORR and CBR24 odds ratios were obtained from logistic regression adjusted for prior CDK4/6 inhibitor use and presence of lung and/or liver metastasis. They should not be interpreted as simple unadjusted ratios.
- Multiple comparisons: the ClinicalTrials.gov record reports two primary PFS comparisons and multiple secondary analyses but do not provide an interim-analysis or multiplicity-adjustment framework. No additional multiplicity procedure should therefore be inferred.
- Unmasked design: the trial is registered as having no masking. The ClinicalTrials.gov record does not provide enough detail to quantify how this feature affected any particular endpoint.
- Generalizability: the study population is specifically defined as post-menopausal women at least 18 years of age with advanced ER-positive HER2-negative breast cancer. Applicability outside that population is a separate question from statistical significance within the trial.
17. Why This Trial Matters Statistically
SERENA-2 is a useful teaching case because it connects a randomized time-to-event primary endpoint with adjusted binary secondary outcomes. The trial therefore illustrates how the same clinical question can require different statistical estimands and models depending on the endpoint.
| Statistical concept | How it appears in SERENA-2 |
|---|---|
| Randomization | Randomized parallel-group phase 2 design with 240 enrolled patients. |
| Full analysis set | Primary PFS analyses used all randomized patients according to randomized treatment assignment. |
| Time-to-event analysis | PFS was the registered primary endpoint. |
| Log-rank test | Reported for both primary PFS comparisons. |
| Hazard ratio | Used to quantify the primary PFS treatment effect. |
| Stratified Cox model | Used for the primary PFS analyses. |
| Confidence interval | 90% intervals were reported for the primary hazard ratios and secondary odds ratios. |
| Logistic regression | Used for ORR and CBR24. |
| Covariate adjustment | ORR and CBR24 models adjusted for prior CDK4/6 inhibitor use and lung and/or liver metastasis. |
| Odds ratio | Used for the binary secondary outcomes, with OR > 1 favoring AZD9833. |
| Small-arm interpretation | The 300mg arm stopped early at 20 randomized patients and was specifically flagged as highly variable. |
18. Trial Timeline
Trial start
The registered trial start date was April 22, 2020.
Primary completion
The registered primary completion date was August 30, 2022.
Secondary ORR assessment window
The posted ORR analyses used a time frame from screening until disease progression, up to a data cut-off of 29 months.
Active, not recruiting
the ClinicalTrials.gov record lists the study status as ACTIVE_NOT_RECRUITING.
19. What the Primary Hazard Ratios Do — and Do Not — Mean
The reported HR 0.59 means that the estimated hazard under the stratified Cox model was 0.59 times the comparator hazard. The registry specifies that an HR below 1 favors AZD9833 for longer progression-free survival.
It does not mean that 41% of patients avoided progression, that PFS was 41% longer for every patient, or that the absolute probability of progression was reduced by 41%.
The reported HR 0.64 means that the estimated hazard under the stratified Cox model was 0.64 times the comparator hazard. Under the registry's stated interpretation, this favors AZD9833 for longer progression-free survival.
It does not mean that 36% of patients avoided progression or that every patient's individual risk fell by 36%.
The 90% confidence intervals of 0.42–0.82 and 0.46–0.89 describe statistical uncertainty around the two estimated hazard ratios. They do not describe the range of treatment effects experienced by individual patients.
Hazard ratios are relative measures. An absolute interpretation would require quantities such as event-free probabilities at a specified time or median PFS. Those measures are not reported in the ClinicalTrials.gov record and therefore are not reproduced here.
20. Dose Comparisons: What Can and Cannot Be Concluded
The two formal primary comparisons were each against the same comparator: fulvestrant 500 mg. The reported hazard ratios were 0.59 for AZD9833 75mg and 0.64 for AZD9833 150mg.
It would be incorrect to interpret the smaller numerical hazard ratio as proof that 75mg was statistically superior to 150mg. The ClinicalTrials.gov record contains no direct 75mg-versus-150mg statistical comparison, and comparing two estimates that each have uncertainty is not the same as testing their difference.
The 300mg dose requires a separate caution. The registry states that recruitment to that arm was stopped early at 20 patients randomized and that the resulting data are highly variable. The statistical analyses posted on ClinicalTrials.gov do not include a formal primary PFS comparison for 300mg versus fulvestrant, so no primary hazard ratio for that dose is presented.
21. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The two reported primary PFS comparisons produced hazard ratios below 1, with 90% confidence intervals that remained below 1 and reported P-values of 0.0167 and 0.0090. The analyses used a stratified log-rank framework and stratified Cox proportional-hazards models in the FAS.
Endpoint interpretation
PFS, ORR, and CBR24 measure different outcomes. The primary endpoint is time-to-event, while the reported secondary endpoints are binary. Their effect measures therefore cannot be interpreted as interchangeable quantities.
Secondary analyses
ORR and CBR24 were analyzed with adjusted logistic regression. The odds-ratio confidence intervals include 1 for all four reported comparisons, so their point estimates should be interpreted together with their uncertainty.
Safety and design
Serious adverse events were reported separately by arm. The 300mg arm had only 20 patients at risk and was specifically identified by the registry as highly variable after early recruitment stopping.
22. A Practical Statistical Reading of SERENA-2
A disciplined reading of this trial starts with the estimand rather than the P-value. The primary estimand is a relative time-to-event comparison for PFS, represented by a hazard ratio. The two reported estimates, 0.59 and 0.64, are therefore the central measures of treatment effect for the primary endpoint.
The next question is precision. The corresponding 90% confidence intervals are 0.42–0.82 and 0.46–0.89. These intervals provide the uncertainty surrounding the estimates and show that the reported intervals remain below the null hazard ratio of 1.
The third question is statistical evidence. The reported P-values are 0.0167 and 0.0090. They provide evidence against the relevant null hypotheses under the reported analyses, but they do not tell us how large the treatment effect is or whether an effect is clinically important.
The fourth question is endpoint consistency. The secondary ORR and CBR24 analyses use logistic regression and odds ratios. Their estimates are directionally above 1, but the registry-reported confidence intervals include 1. That means these secondary results require a different interpretation from the primary PFS results.
Finally, the analysis must be placed in the context of trial design. The study is randomized and parallel, but unmasked. The 300mg arm stopped early at 20 randomized patients, and the registry explicitly warns that its data are highly variable. These design details matter because statistical estimates are only as informative as the amount and quality of information available to estimate them.
23. Related Statistical Tutorials
Learn more about the methods used in this trial:
24. Related Statistical Calculators
25. Sources
- ClinicalTrials.gov: SERENA-2, NCT04214288.
- PubMed: PubMed record for PMID 39481395.
- PubMed: PubMed record for PMID 38729567.
Continue through the Clinical Biostats statistical library
Explore tutorials and statistical calculators covering the survival-analysis and regression concepts illustrated by SERENA-2.
26. Record Summary
SERENA-2 provides a compact example of how randomized clinical-trial evidence can be analyzed across different endpoint types. Its primary endpoint, progression-free survival, was analyzed using a stratified log-rank framework and stratified Cox proportional-hazards models, producing hazard ratios of 0.59 for AZD9833 75mg versus fulvestrant 500 mg and 0.64 for AZD9833 150mg versus fulvestrant 500 mg. The corresponding 90% confidence intervals were 0.42–0.82 and 0.46–0.89, with P-values of 0.0167 and 0.0090.
The secondary endpoints demonstrate a different modeling approach. ORR and CBR24 were binary outcomes analyzed using logistic regression adjusted for prior CDK4/6 inhibitor use and presence of lung and/or liver metastasis. Their reported odds ratios ranged from 1.42 to 1.62, while all four reported confidence intervals included 1.
The most important statistical lesson is that these measures should not be collapsed into a single number. The hazard ratio describes a relative time-to-event effect; the odds ratio describes a modeled binary-outcome association; the confidence interval describes uncertainty; and the P-value describes evidence against a null hypothesis under the specified analysis. The 300mg arm also illustrates why sample size and trial conduct matter: recruitment stopped early at 20 randomized patients, and the registry explicitly cautions that the resulting data are highly variable.