← Clinical Trials
Breast Cancer Phase 3 Time-to-Event NCT00754845

MA.17: Complete Statistical Analysis of Letrozole in Breast Cancer

An independent statistical analysis of the randomized phase 3 MA.17 trial evaluating letrozole versus placebo in women with breast cancer who had received 5 years of aromatase inhibitor therapy, with disease-free survival as the primary endpoint.

Trial period: November 23, 2004  ·  Primary completion: December 21, 2015  ·  Enrollment: 1,918
Scope of this record

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

MA.17 was a randomized, parallel-group phase 3 trial evaluating letrozole versus placebo in breast cancer after 5 years of aromatase inhibitor therapy. The registered primary endpoint was disease-free survival (DFS), a time-to-event outcome analyzed using a stratified log-rank test with a hazard ratio as the effect measure.

1,918
Randomized
2 treatment arms
0.66
DFS HR
95% CI 0.48–0.91
0.01
DFS P-value
Two-sided
75 mo
Median Follow-Up
Primary endpoint timeframe
FeatureMA.17
PhasePhase 3
ConditionBreast Cancer
Brief titleLetrozole in Breast Cancer Who Have Received 5 Years of Aromatase Inhibitor Therapy
DesignRandomized, parallel-group
MaskingTriple
Primary purposeTreatment
AllocationRandomized
Enrollment1,918
Arms2
InterventionsLetrozole; placebo
Primary endpointDisease-free Survival (DFS)
Primary endpoint typeTime-to-event
Hypothesis typeSuperiority
Results postedYes
Lead sponsorCanadian Cancer Trials Group
Sponsor typeNetwork
ClinicalTrials.govNCT00754845

2. Clinical Question

The central statistical question was whether treatment with letrozole, compared with placebo, changed disease-free survival among women with breast cancer who had received 5 years of aromatase inhibitor therapy.

Population

Women with breast cancer who had received 5 years of aromatase inhibitor therapy.

Intervention

Letrozole.

Comparator

Placebo.

Primary question

Does letrozole produce a different disease-free survival experience from placebo under a superiority hypothesis?

3. Trial Design

01
Randomize 1,918 women
02
Assign Letrozole or placebo
03
Follow Time-to-event outcomes
04
Analyze Stratified log-rank
05
Estimate Hazard ratio and CI
Allocation
Randomized allocation to two parallel treatment groups.
Masking
Triple masking was registered for the study.
Primary purpose
Treatment.
Hypothesis
Superiority of letrozole versus placebo.

The trial started on November 23, 2004 and had a primary completion date of December 21, 2015. The registry reports an enrollment of 1,918 and two treatment arms.

ARM I

Letrozole

  • Intervention type: drug
  • Compared with placebo
  • Included in the primary DFS comparison
ARM II

Placebo

  • Intervention type: other
  • Comparator for letrozole
  • Included in the primary DFS comparison

4. Endpoints

EndpointRoleTime frameTypeReported analysis
Disease-free Survival (DFS) Primary Unitil the end of study with a median follow up of 75 months Time-to-event Stratified log-rank test; hazard ratio
Overall Survival (OS) Secondary Until the end of study with a median follow-up of 75 months Time-to-event Log-rank test; hazard ratio
Incidence of Contralateral Breast Cancer Secondary 10 years Time-to-event Log-rank test
Change From Baseline in Role Function- Physical Scale on SF(Short Form)-36 Health Survey Secondary 8 years Continuous Mixed-effects model

Primary endpoint: Disease-free Survival

The registry defines Disease-free Survival (DFS) as the months from the day of randomization to the earliest date when a recurrence of the primary disease was observed, including recurrence in the breast, chest wall and nodal sites or development of metastatic disease, or when a contralateral breast cancer was observed. Subjects who died without recurrence of the primary disease or development of the contralateral breast cancer were censored.

The registered time frame is "Unitil the end of study with a median follow up of 75 months". The reported outcome unit is the probability of DFS at 5 years.

5. Statistical Methodology

Primary DFS analysis

The primary DFS analysis used a stratified log-rank test. The reported effect measure was a hazard ratio, with a two-sided 95% confidence interval and a superiority hypothesis.

Primary analysis structure
Randomized treatment groups → time-to-event follow-up → stratified log-rank comparison → hazard ratio + 95% CI + P-value

The analysis population included all women randomized, analyzed according to the treatment arm to which they were randomized.

Stratification

The registry's analysis notes state that the primary DFS analysis was stratified by the stratification factors at randomization. Stratification preserves the role of prespecified randomization factors when comparing time-to-event outcomes.

A stratified log-rank test does not simply compare two unadjusted event-time distributions. Instead, it incorporates the prespecified strata into the comparison, allowing the treatment contrast to be evaluated while respecting the trial's stratified randomization structure.

Analysis population

For the primary DFS analysis, all women randomized were included in the analysis based on treatment arm they were randomized. This is an intention-to-treat-style analysis principle: the efficacy comparison remains tied to randomized assignment rather than being redefined according to treatment actually received or later treatment changes.

Secondary analyses

Overall survival and incidence of contralateral breast cancer were also analyzed with the log-rank test. The OS analysis reported a hazard ratio, while the contralateral breast cancer analysis reported a P-value without a hazard-ratio estimate in the ClinicalTrials.gov record.

The longitudinal SF-36 physical role-function endpoint was analyzed with a mixed-effects model, reflecting a fundamentally different statistical structure from the time-to-event endpoints.

6. Results

Primary endpoint: Disease-free Survival

Hazard ratio for disease-free survival

0.66

95% CI: 0.48–0.91   ·   P = 0.01   ·   Two-sided

Comparison: Letrozole vs Placebo

Primary DFS resultReported value
Outcome unitProbability of DFS at 5 years
Analysis populationAll women randomized, analyzed according to randomized treatment arm
Groups comparedLetrozole vs Placebo
MethodStratified log-rank test
Effect measureHazard ratio
Estimate0.66
95% CI0.48–0.91
P-value0.01
HypothesisSuperiority
StratificationStratified by the stratification factors at randomization
Clinical Biostats interpretation

The estimated hazard ratio of 0.66 means that, under the time-to-event comparison represented by the reported hazard ratio, the estimated instantaneous rate of a DFS event in the letrozole group was approximately 66% of that in the placebo group. Expressed as a relative complement, this corresponds to an estimated 34% lower hazard for the letrozole group.

The HR does not mean that 34% of patients avoided recurrence, that 34% more patients were disease-free at 5 years, or that every patient experienced the same proportional reduction in event risk. A hazard ratio is a relative time-to-event measure, not an absolute probability.

The 95% confidence interval of 0.48–0.91 describes uncertainty around the estimated hazard ratio under the statistical model and sampling framework. It does not describe the range of effects that individual patients experienced.

The P-value of 0.01 addresses evidence against the null hypothesis in the specified superiority comparison. It does not measure the size of the treatment effect. Effect magnitude and statistical evidence should therefore be read separately: the HR describes the estimated relative effect, the CI describes its precision, and the P-value describes evidence against the null under the specified testing framework.

Because DFS is a censored time-to-event endpoint, the analysis also depends on how follow-up and censoring are handled. The hazard-ratio interpretation additionally relies on the appropriateness of representing the treatment contrast with a proportional-hazards-type summary over the analyzed follow-up. The ClinicalTrials.gov record does not provide a diagnostic assessment of that assumption.

Secondary endpoint: Overall Survival

Hazard ratio for overall survival

0.97

95% CI: 0.73–1.28   ·   P = 0.83   ·   Two-sided

Comparison: Arm I vs Arm II

OS resultReported value
Time frameUntil the end of study with a median follow-up of 75 months
Outcome unitProbability of OS at 5 years
Analysis populationAll women randomized
Groups comparedArm I vs Arm II
MethodLog-rank test
Effect measureHazard ratio
Estimate0.97
95% CI0.73–1.28
P-value0.83
HypothesisSuperiority
Clinical Biostats interpretation

The OS hazard ratio of 0.97 is close to 1. Under the reported model-based time-to-event comparison, the estimated instantaneous rate of death in Arm I was approximately 97% of that in Arm II.

This does not mean that overall survival was exactly the same in the two groups. The estimate is subject to sampling uncertainty, as reflected by the 95% confidence interval of 0.73–1.28.

The confidence interval spans 1.00, so the data are compatible with a range of relative hazard differences in either direction under the specified analysis. The P-value of 0.83 indicates little evidence against the null hypothesis in this superiority test; it is not a measure of how large or small the observed HR is.

The registry reports the analysis as a log-rank comparison with a hazard ratio. As with the primary DFS analysis, interpretation of a single HR requires attention to censoring and the appropriateness of the proportional-hazards summary over follow-up.

Secondary endpoint: Incidence of Contralateral Breast Cancer

ResultReported value
Time frame10 years
Outcome unitNumber of new case per 1000 person years
Analysis populationAll women randomized
Groups comparedLetrozole vs Placebo
MethodLog-rank test
P-value0.007
HypothesisSuperiority
Clinical Biostats interpretation

The registry reports a P-value of 0.007 for the superiority comparison of letrozole versus placebo over 10 years. The reported outcome unit is the number of new cases per 1000 person-years.

No hazard-ratio estimate or confidence interval is included in the registry-reported statistical analysis for this endpoint. Consequently, the result should not be converted into a hazard ratio or an estimated relative reduction without additional reported data.

The P-value indicates the statistical evidence reported for the comparison; it does not quantify the magnitude of the difference. For magnitude and precision, an effect estimate and confidence interval would provide additional information, but those values are not included in the registry analysis.

Secondary endpoint: SF-36 Role Function—Physical Scale

ResultReported value
EndpointChange From Baseline in Role Function- Physical Scale on SF(Short Form)-36 Health Survey
Time frame8 years
Outcome unitScore on a scale
Analysis populationAll randomized patients
Groups comparedLetrozole vs Placebo
MethodMixed Models Analysis / mixed-effects model
P-value<0.01
HypothesisSuperiority
Clinical Biostats interpretation

The SF-36 physical role-function endpoint was analyzed with a mixed-effects model, rather than a survival-analysis method. The registry reports a superiority-test P-value of <0.01 for the comparison of letrozole versus placebo.

Because the registry analysis does not report an estimated between-group difference or confidence interval for this endpoint, the P-value alone does not establish the magnitude or precision of the difference. A mixed-effects model is designed to account for the correlation among repeated observations from the same participant, but the exact model specification is not provided in the ClinicalTrials.gov record.

7. Statistical Methods Explained

Why was a log-rank test used for DFS?

DFS is a time-to-event endpoint. Participants can experience an event at different times, while others may remain event-free at the end of available follow-up and therefore contribute censored observations. A log-rank test compares the event-time experience of randomized groups while incorporating the timing of events and censoring rather than reducing the outcome to a simple yes/no proportion at one arbitrary date.

Why was the DFS analysis stratified?

The registry states that the primary DFS analysis was stratified by the stratification factors used at randomization. Stratification allows the time-to-event comparison to respect the trial's prespecified randomization structure. It is particularly useful when the randomization procedure intentionally balances treatment assignments within clinically or statistically important strata.

What does an HR of 0.66 mean?

An HR of 0.66 is a relative time-to-event estimate. It indicates an estimated instantaneous event rate in the letrozole group that is 66% of the corresponding rate in the placebo group under the fitted analysis. The complementary interpretation is an estimated 34% lower hazard.

That is different from saying that 34% of participants benefited or that the probability of an event was reduced by exactly 34% at every time point. Absolute event probabilities require separate estimates such as Kaplan-Meier survival probabilities.

Why is the confidence interval important?

The DFS 95% CI of 0.48–0.91 communicates the uncertainty surrounding the estimated HR of 0.66. The interval is more informative than the point estimate alone because it shows how precisely the treatment effect was estimated under the specified statistical framework.

A confidence interval is not a range containing the true treatment effect with a fixed probability after the data have been observed. It is a statement about the performance of the interval procedure under repeated sampling.

Why doesn't the P-value measure treatment effect?

The DFS P-value of 0.01 and OS P-value of 0.83 answer questions about evidence against their respective null hypotheses. They do not measure clinical importance, effect magnitude, or probability that the treatment works.

For that reason, a statistically interpretable trial result should be read using the full set of available quantities: the effect estimate, its confidence interval, the P-value, the endpoint definition, and the analysis population.

Why was a mixed-effects model used for the SF-36 endpoint?

The SF-36 endpoint is continuous rather than time-to-event and is measured longitudinally. Repeated measurements from the same participant are correlated, so treating every observation as independent can understate uncertainty. A mixed-effects model can represent this within-participant correlation while estimating treatment-related differences over repeated assessments.

The ClinicalTrials.gov record reports only that a mixed-effects model was used and give a P-value of <0.01. They do not provide the model's covariance structure, fixed-effect specification, random-effect structure, or an estimated treatment difference.

8. Kaplan-Meier Estimation and Censoring

Although the registry's normalized methods list identifies the primary method as a log-rank test, DFS and OS are time-to-event outcomes for which Kaplan-Meier estimation is the standard descriptive framework. Kaplan-Meier curves estimate the probability of remaining event-free or alive over time while retaining information from participants who are censored.

Conceptual Kaplan-Meier estimator
S(t) = ∏ti ≤ t (1 − di/ni)

Here, di represents events at time ti, while ni represents participants at risk immediately before that event time.

The important statistical distinction is that censoring does not mean that an individual experienced the event at the censoring time. Instead, that participant contributes information through the time at which follow-up ends or the participant otherwise becomes censored under the analysis rules.

Educational note: the ClinicalTrials.gov record does not contain individual event and censoring times, so this page does not reconstruct or fabricate a Kaplan-Meier curve.

9. Hazard Ratios in MA.17

EndpointComparisonHR95% CIP-value
Disease-free SurvivalLetrozole vs Placebo0.660.48–0.910.01
Overall SurvivalArm I vs Arm II0.970.73–1.280.83

The contrast between the two estimates illustrates why a trial should not be summarized by one number alone. The primary DFS analysis reports an HR of 0.66, whereas the secondary OS analysis reports an HR of 0.97. These are estimates for different endpoints and should be interpreted within their respective endpoint definitions and follow-up structures.

Relative versus absolute effects

A hazard ratio is inherently relative. It does not provide the absolute probability of remaining disease-free or alive at a particular time. The registry identifies the outcome unit for both DFS and OS as the probability at 5 years, but the statistical analyses posted on ClinicalTrials.gov do not provide the corresponding numerical 5-year probabilities. Therefore, the reported HRs should not be substituted for those absolute quantities.

10. Secondary Endpoint Structure

EndpointStatistical familyReported evidenceInterpretive focus
Disease-free Survival (DFS)Survival analysisHR 0.66; 95% CI 0.48–0.91; P = 0.01Primary time-to-event treatment comparison
Overall Survival (OS)Survival analysisHR 0.97; 95% CI 0.73–1.28; P = 0.83Secondary time-to-event comparison
Incidence of Contralateral Breast CancerSurvival analysisP = 0.00710-year incidence comparison
SF-36 Role Function—Physical ScaleLongitudinal / mixed modelsP < 0.01Repeated continuous outcome

These endpoints require different statistical interpretations. DFS and OS incorporate event timing and censoring. Contralateral breast cancer is also represented as a time-to-event endpoint in the registry. The SF-36 outcome is continuous and longitudinal, so its mixed-effects analysis addresses repeated measurements rather than event-time distributions.

11. Safety Results

The ClinicalTrials.gov record reports serious adverse events by randomized arm using affected participants over participants at risk:

ArmSerious adverse eventsAffected / at risk
Arm ISerious adverse events15 / 959
Arm IISerious adverse events19 / 954
Clinical Biostats interpretation

The registry reports serious adverse events as 15/959 in Arm I and 19/954 in Arm II. These are descriptive safety counts rather than a reported inferential comparison in the statistical analyses posted on ClinicalTrials.gov.

The denominators are the reported numbers at risk for the safety measure. They should not be silently substituted with the overall enrollment of 1,918, because safety summaries can use populations defined by treatment exposure or evaluability.

No additional serious-adverse-event confidence interval, P-value, risk ratio, or hazard ratio is included in the ClinicalTrials.gov record. Accordingly, no formal comparative safety conclusion is added here.

12. Multiplicity and Multiple Endpoints

MA.17 has one registered primary endpoint and three posted secondary outcome analyses. The primary endpoint is DFS, while OS, incidence of contralateral breast cancer, and change from baseline in the SF-36 physical role-function scale are secondary endpoints.

Endpoint roleEndpointReported statistical result
PrimaryDisease-free Survival (DFS)HR 0.66; 95% CI 0.48–0.91; P = 0.01
SecondaryOverall Survival (OS)HR 0.97; 95% CI 0.73–1.28; P = 0.83
SecondaryIncidence of Contralateral Breast CancerP = 0.007
SecondaryChange From Baseline in Role Function- Physical Scale on SF(Short Form)-36 Health SurveyP < 0.01

The existence of several statistical analyses creates an important interpretive distinction: a nominal P-value for a secondary endpoint does not automatically have the same confirmatory interpretation as the prespecified primary endpoint. Whether and how multiplicity was controlled depends on the trial's prespecified statistical plan. The ClinicalTrials.gov record does not report an alpha-allocation or multiplicity-adjustment procedure, so none is inferred here.

13. Randomization and the Intention-to-Treat Principle

The primary DFS analysis included all women randomized according to the treatment arm to which they were randomized. This is important because the purpose of randomization is to create comparable treatment groups at baseline, subject to chance variation.

Preserves randomized comparison

Analyzing participants according to randomized assignment maintains the treatment comparison generated by the randomization process.

Avoids post-randomization selection

Reassigning participants according to treatment actually received can introduce selection related to adherence, discontinuation, or subsequent clinical events.

Relevant to DFS

The registry explicitly states that all randomized women were included in the primary analysis based on randomized treatment arm.

Different from safety analysis

Safety populations can be defined around treatment exposure. The registry-reported serious-AE results are presented using affected and at-risk counts rather than the overall randomized enrollment.

14. Time-to-Event Analysis: What the Data Actually Tell Us

Time-to-event analysis contains more information than a simple event proportion because it incorporates when events occur. Two participants can both eventually experience a recurrence, yet their contribution to a time-to-event analysis differs if one recurrence occurs much earlier than the other.

MA.17's primary DFS analysis therefore uses a method suited to the structure of the endpoint: the stratified log-rank test. The hazard ratio provides a relative summary of the treatment contrast over follow-up.

Conceptual interpretation
HR = instantaneous event rate under letrozole ÷ instantaneous event rate under placebo

For the reported DFS result, HR = 0.66. This is a model-based relative measure and should not be interpreted as an absolute risk ratio.

The ClinicalTrials.gov record does not provide the underlying event counts, individual follow-up times, Kaplan-Meier estimates at multiple time points, or median DFS. Those quantities are therefore not reconstructed on this page.

15. Longitudinal Analysis of SF-36

The SF-36 physical role-function outcome is structurally different from DFS and OS. It is a continuous score measured over time, with the registry defining the outcome as change from baseline over an 8-year time frame.

The registry reports a mixed-effects model for this analysis. Mixed-effects models are useful when repeated observations are obtained from the same participants because those observations are generally correlated.

General mixed-effects structure
Yij = fixed effects + random effects + residual error

The fixed component can describe systematic treatment and time effects, while random effects and the covariance structure account for dependence among observations from the same participant. The exact specification used in MA.17 is not provided in the ClinicalTrials.gov record.

Interpretation caution: the registry reports only a P-value of <0.01 for this endpoint. It does not provide an estimated between-group change, confidence interval, or detailed covariance/model specification in the ClinicalTrials.gov record. The statistical evidence can therefore be reported, but its magnitude cannot be reconstructed from the P-value alone.

16. Clinical Biostats Interpretation of the Primary Result

What the DFS result says

The primary DFS analysis reported a hazard ratio of 0.66 for letrozole versus placebo, with a two-sided 95% confidence interval of 0.48–0.91 and a P-value of 0.01. The analysis included all randomized women according to randomized treatment arm and used a stratified log-rank test.

What the DFS result does not say

The result does not provide a 5-year absolute DFS probability in the ClinicalTrials.gov record. It also does not imply that every participant experienced a 34% reduction in individual risk, nor does it establish the probability that any particular participant will remain disease-free.

Why the confidence interval matters

The interval from 0.48 to 0.91 indicates uncertainty around the estimated HR. It shows that the point estimate of 0.66 should not be treated as an exact treatment effect. The interval also communicates substantially more information about precision than the P-value alone.

Why the P-value is secondary to the effect estimate for magnitude

A P-value of 0.01 indicates statistical evidence under the specified superiority test. It does not tell us whether the HR is clinically large, clinically small, or precisely estimated. Those questions are addressed by the HR, confidence interval, and absolute outcome measures.

17. Limitations

18. Why This Trial Matters Statistically

MA.17 is a useful teaching example because its registry record connects several core clinical-trial methods in a single randomized study: time-to-event endpoints, stratified log-rank testing, hazard ratios, confidence intervals, intention-to-treat analysis, and longitudinal mixed-effects modeling.

Statistical conceptHow it appears in MA.17
RandomizationRandomized allocation to two parallel treatment arms
MaskingTriple masking
Superiority testingRegistered hypothesis type for the reported analyses
Time-to-event endpointDFS is the registered primary endpoint
Stratified analysisPrimary DFS analysis stratified by the factors used at randomization
Log-rank testUsed for DFS, OS, and contralateral breast cancer analyses
Hazard ratioReported for DFS and OS
Confidence interval95% two-sided intervals reported for DFS and OS HRs
Intention-to-treat principleAll women randomized included in the primary DFS analysis according to randomized arm
Mixed-effects modelUsed for change from baseline in the SF-36 physical role-function scale
Longitudinal outcomeSF-36 endpoint assessed over an 8-year time frame
Secondary endpointsOS, contralateral breast cancer incidence, and SF-36 physical role function

19. Statistical Methods Explained Through the Primary DFS Result

Randomization

Randomized assignment provides the basis for comparing letrozole and placebo without relying on investigators to choose which participants receive each treatment.

Stratification

The primary DFS analysis respects the stratification factors used at randomization rather than ignoring that design feature.

Log-rank test

The primary comparison evaluates time-to-event experience rather than simply counting whether each participant eventually experienced an event.

Hazard ratio

The HR of 0.66 summarizes the relative instantaneous event rate under the fitted time-to-event comparison.

20. Interpreting DFS and OS Together

The ClinicalTrials.gov record shows two different time-to-event estimates: a DFS HR of 0.66 and an OS HR of 0.97. These should not be treated as contradictory measurements of the same outcome. DFS captures the first occurrence of specified disease-related events or contralateral breast cancer, while OS concerns death from any cause.

Differences between endpoints are expected because they represent different clinical events and because events that occur before death can affect the DFS experience without necessarily producing the same relative difference in overall survival.

FeatureDFSOS
RolePrimarySecondary
TypeTime-to-eventTime-to-event
Reported effectHR 0.66HR 0.97
95% CI0.48–0.910.73–1.28
P-value0.010.83
AnalysisStratified log-rankLog-rank
Time frameEnd of study; median follow-up 75 monthsEnd of study; median follow-up 75 months

The appropriate interpretation is therefore endpoint-specific: the primary DFS analysis reports an estimated hazard ratio below 1 with its associated confidence interval and P-value, while the secondary OS analysis reports an HR close to 1 with a wider interval spanning 1.00.

21. The Difference Between Statistical Evidence and Effect Size

MA.17 illustrates why the statistical result should be read as a collection of quantities rather than as a single P-value.

Four complementary quantities
Effect estimate + confidence interval + P-value + endpoint definition

The estimate describes the observed relative effect, the confidence interval communicates uncertainty, the P-value evaluates evidence against a null hypothesis, and the endpoint definition establishes what clinical event is actually being analyzed.

For DFS, the four pieces are HR 0.66, 95% CI 0.48–0.91, P = 0.01, and a registered endpoint defined from randomization to recurrence, metastatic disease, or contralateral breast cancer, with specified censoring for participants who died without those events.

For OS, the corresponding values are HR 0.97, 95% CI 0.73–1.28, and P = 0.83, for the endpoint defined as overall survival in the registry analysis.

22. Important Statistical Interpretation Issues

23. Trial Timeline

November 23, 2004

Trial start

The registered MA.17 trial began.

Phase 3 design

Randomized comparison

The study used randomized allocation in a parallel design with triple masking and two treatment arms: letrozole and placebo.

Primary analysis

Disease-free survival

The primary DFS endpoint was analyzed using a stratified log-rank test with a hazard ratio of 0.66, 95% CI 0.48–0.91, and P = 0.01.

December 21, 2015

Primary completion

The registered primary completion date was December 21, 2015.

24. Related Tutorials

Learn more about the methods used in this trial:

25. Related Calculators

26. Sources

Continue through Clinical Biostats

Explore the statistical methods behind randomized clinical trials, survival analysis, longitudinal models, confidence intervals, and related clinical-trial calculations.

27. Record Summary

MA.17 provides a compact example of several central principles in clinical-trial statistics. The study is a randomized phase 3, parallel-group trial with triple masking and two interventions, letrozole and placebo. Its registered primary endpoint is disease-free survival, a time-to-event outcome analyzed with a stratified log-rank test and summarized using a hazard ratio.

The primary DFS result was HR 0.66, with a two-sided 95% confidence interval of 0.48–0.91 and P = 0.01. The analysis included all randomized women according to their randomized treatment arm and was stratified by the factors used at randomization.

The secondary analyses demonstrate why endpoint-specific interpretation matters. Overall survival produced an HR of 0.97 with a 95% CI of 0.73–1.28 and P = 0.83. The incidence of contralateral breast cancer was analyzed with a log-rank test and reported P = 0.007, while the longitudinal SF-36 physical role-function endpoint was analyzed using a mixed-effects model with P < 0.01.

The statistical lesson is broader than any single result. A rigorous interpretation combines the endpoint definition, randomized analysis population, statistical method, effect estimate, confidence interval, and P-value. It also distinguishes time-to-event methods from longitudinal continuous-outcome models and avoids converting unreported quantities into estimates that the registry does not provide.

Clinical Biostats methodology: This page separates reported registry results from statistical interpretation. Numerical values are reproduced from the ClinicalTrials.gov record without rounding, recomputation, or substitution of results from other sources.