S0300, Celecoxib in Preventing Breast Cancer in Premenopausal Women: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for S0300, Celecoxib in Preventing Breast Cancer in Premenopausal Women (NCT00088972), a randomized phase 2 study. The primary outcome is Mammographic Density.

2. Clinical Question

This is a question about a biomarker endpoint, not a direct comparison of breast cancer occurrence. The registry lists Ki-67 expression as a secondary outcome, but does not report an analyzed result for it. Because the study closed early due to poor accrual and had insufficient accrual to evaluate its endpoints, the posted record does not establish whether the planned biomarker comparison favored either arm. It also does not report a treatment-effect estimate or its precision.

3. Trial Design

The registry describes S0300 as a randomized, double-masked, parallel phase II trial of celecoxib versus placebo. Its official title identifies it as a randomized placebo-controlled biomarker-modulation trial in premenopausal women at high risk for breast cancer. SWOG Cancer Research Network is listed as the sponsor. The recorded study start is 2004-11, and primary completion is recorded as 2009-07. The current status is terminated.

The intended treatment period was 12 months. Arm I received oral celecoxib twice daily, and Arm II received oral placebo twice daily, subject to the stated stopping conditions of unacceptable toxicity or cancer diagnosis. The registry does not report the randomization method, allocation concealment procedures, or how masking was maintained. The terms randomized and double-masked describe the recorded design, but do not give those operational details.

Enrollment was 8. The registry states that the study closed early because of poor accrual and that there was insufficient accrual to evaluate study endpoints. It also says that the mammographic-density images submitted for the primary outcome were never analyzed. These details are central to interpreting the record: the planned design identifies a comparison, but the record does not provide an analyzed primary outcome to compare.

The listed secondary outcome is Ki-67 expression. The record lists 2 outcome measures and 0 analyses. It does not report a detailed statistical analysis plan, a sample-size justification, or whether any interim analyses were planned. The generic sample-size assumptions describe a planning illustration based on an observed effect; they do not establish that such an effect was observed for the primary endpoint or that the illustration formed part of the trial's original design.

4. Participant Flow and Accounting

The registry reports how participants moved through each stage of the study. All counts below are exactly as posted; missing counts are not inferred from other stages.

Overall Study

Analysis population for the primary outcome, as defined in the registry: Counts represent number of patients for by arm for whom a 1 year mammographic density image was submitted. However, since the study was closed early due to poor accrual, there was insufficient accrual to evaluate study endpoints, and these images were never analyzed.

5. Baseline Characteristics

The table reproduces the baseline characteristics posted in the registry, which describe the participants analysed at baseline and help show whether the randomized groups were comparable.

6. Primary Endpoint

The registry lists the primary outcome as Mammographic Density, measured over 1 year.

The primary endpoint was change in mammographic density over 1 year. Mammographic density is an imaging-based measure, and the registry describes the planned comparison in terms of the percent of pixels highlighted. The stated null hypothesis was no difference between the arms in change over the year. The stated alternative was a reduction of 10 points or more in the treatment arm relative to the change in the placebo arm.

As a general statistical rationale, a biomarker endpoint can be used to assess whether an intervention changes a biological measure of interest. A randomized comparison can help distinguish differences between assigned groups from changes that might occur over time in both groups. However, a biomarker result is not itself evidence about a clinical outcome unless the relationship between that measure and the clinical outcome is established and addressed in the analysis.

For this trial, the registry says that mammographic-density images were submitted for participants in each arm but were never analyzed because the study closed early and had insufficient accrual to evaluate endpoints. Thus, the record does not report the observed change, a between-arm estimate, or uncertainty around that estimate. It also does not describe the image-reading method, measurement reproducibility, or handling of images that were not submitted. The primary endpoint was specified, but its planned comparison cannot be interpreted from analyzed outcome data in the registry record.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The design specifies a randomized comparison between celecoxib and placebo, with parallel groups and double masking. In general, random assignment is intended to make the groups comparable at the start of a study, while masking can reduce the influence of treatment knowledge on participant behavior, care, or outcome assessment. These are design aims, not guarantees. The registry does not describe the randomization sequence, concealment procedures, or practical details of masking, so those aspects cannot be evaluated from the record.

What would a between-arm estimate represent?

For the primary endpoint, an analysis would need to summarize mammographic-density change in each arm and estimate the difference between those changes. The registry defines the target as a comparison over 1 year and expresses the clinical hypothesis in points of percent of pixels highlighted. An estimate would describe the observed contrast in the analyzed participants under the chosen analysis method. Its interpretation would depend on how images and measurements were handled, which the record does not specify.

Randomization does not remove the need to define the analysis population and outcome procedure. For example, an analysis might use participants according to assigned group, or a narrower population with completed measurements; those choices answer somewhat different questions. This is a general methodological distinction. The registry does not report which population was intended for the primary analysis, and the images were never analyzed, so no such choice can be inferred from posted results.

How should uncertainty be understood?

In general, a confidence interval gives a range of values compatible with the observed data under the assumptions of the analysis. A narrow interval indicates greater precision than a wide interval, while an interval that spans effects in different directions may leave the direction uncertain. A p-value, in general, describes how unusual data at least as inconsistent with a null hypothesis would be under that hypothesis and the model. Neither measure is reported for the primary endpoint here, because the registry says the images were not analyzed. There is therefore no basis for interpreting statistical significance or precision for mammographic-density change.

The stated null hypothesis was no difference between arms in change over 1 year. The alternative described a reduction in the treatment arm of 10 points or more relative to placebo. This is a prespecified target contrast as described in the registry, not an observed result. The record does not report a test statistic, confidence interval, or p-value testing that contrast. It also does not explain whether the target was intended as a threshold for a clinically meaningful effect or solely as the alternative for a statistical test.

What do the planning assumptions say?

As a general concept, a sample-size calculation uses assumptions about the effect to detect, variability, significance level, power, and allocation to estimate the amount of information a planned study would need. The registry's generic description states a two-sided significance level of 0.05, equal allocation between arms, and a normal approximation. It says the effect to detect is a stated fraction of the observed effect. It also distinguishes participants per arm for most outcomes from total events for time-to-event outcomes.

These assumptions describe a planning illustration, not the original trial's design rationale. The registry does not report the trial's sample-size justification or a numerical planning estimate in the registry record for this analysis. More importantly, the primary endpoint was never analyzed, so an observed primary effect is not reported. Any planning illustration should not be read as observed power, evidence of a treatment effect, or a retrospective demonstration that the study could or could not have found a particular result.

Power is a design property calculated under assumed conditions; it is not a measure of whether a completed study has demonstrated an effect. General explanations of power can help readers understand a planning exercise, but they cannot supply missing outcome data. With no analyzed primary estimate, the central limitation is absence of an endpoint result, not a numerical power value that can be interpreted as evidence.

Why does over-dispersion matter in general?

For outcomes analyzed with count models, over-dispersion means that variability exceeds what a simpler model assumes. If that extra variability is ignored, standard errors can be too small and uncertainty can be understated. This is a general statistical issue; the registry does not report a count-model analysis or an over-dispersion assessment for the primary outcome. The primary endpoint is described as mammographic-density change, and no analysis was performed.

Other analysis choices can also affect interpretation. These include adjustment for baseline values, missing outcome handling, transformations, and sensitivity analyses. The registry record does not report those procedures. They should not be inferred from the randomized design or from the stated endpoint description. Without an analysis plan and analyzed outcome data, the record supports a description of the intended comparison and the reason it was not evaluated, but not a reconstruction of an unreported statistical analysis.

10. Multiplicity, Missing Data and Analysis Populations

The registry lists mammographic density as the primary outcome and Ki-67 expression as a secondary outcome. It also records 2 outcome measures and 0 analyses. The primary images were never analyzed, and the registry says that insufficient accrual prevented evaluation of study endpoints. The record therefore does not provide estimates or tests for either outcome in the material.

Multiplicity concerns arise when several outcomes, time points, or comparisons are tested, because the chance of at least one apparently notable result can increase as testing expands. As a general approach, a protocol may identify a primary endpoint, specify a testing hierarchy, or adjust inference across multiple tests. The registry record does not report a multiplicity strategy, any planned hierarchy, or whether other analyses were contemplated. The stated primary endpoint and listed secondary outcome alone do not establish how multiplicity would have been handled.

Missing data are also consequential. The registry says images were submitted for participants by arm, but does not report an analysis of those images. It does not describe the number of missing images, reasons for missingness, or a method for addressing missing outcome measurements. Nor does it state how the primary analysis population would have been defined. Because the endpoint was not analyzed, there is no reported missing-data analysis or sensitivity analysis to assess.

These omissions should not be filled by assuming that all enrolled participants had complete outcomes, or that a particular missing-data method was used. The most direct interpretation is that the intended endpoint comparison is unavailable in the registry record. The record does not report an analysis population, a missing-data procedure, or a multiplicity adjustment.

11. Safety Results

The registry reports serious adverse events as affected participants divided by participants at risk.

Other (non-serious) adverse events

Participants with at least one non-serious event at or above the registry's 0% reporting threshold.

Most frequent non-serious adverse event terms

The registry reports adverse-event counts separately for Arm I and Arm II. For serious adverse events, Arm I records 0 affected among 3 at risk, and Arm II records 0 affected among 5 at risk. For other adverse events, Arm I records 2 affected among 3 at risk, and Arm II records 0 affected among 5 at risk. These are the posted counts; they do not establish that the underlying risks are the same or different.

The counts should be read descriptively, with the small recorded at-risk groups and early termination in mind. They are not estimates of comparative risk, and no rates or between-arm safety effect are reported. The registry does not list serious adverse-event terms. Deaths are not reported in the registry record. The record also does not supply detail in the registry record on event timing, duration, severity grading for other events, or attribution to study treatment.

Zero recorded serious events in these groups does not prove that serious harms cannot occur. Conversely, the other-event count in Arm I cannot by itself establish that celecoxib caused those events. A comparative safety conclusion would require more complete information and appropriate analysis. The posted data support only the arm-specific counts and the stated at-risk denominators.

12. What the Result Does, and Does Not, Mean

This trial record documents a randomized, double-masked comparison of celecoxib with placebo for a planned biomarker question in premenopausal women at high risk for breast cancer. The primary endpoint was change in mammographic density over 1 year, with a stated alternative involving a reduction of 10 points or more relative to placebo. The registry also lists Ki-67 expression as a secondary outcome.

The result does not answer whether celecoxib changed mammographic density relative to placebo. The study was terminated after poor accrual, the registry says there was insufficient accrual to evaluate endpoints, and the submitted mammographic-density images were never analyzed. No primary-effect estimate, confidence interval, or p-value is reported. It is therefore not appropriate to interpret the absent analysis as evidence of no difference, or as evidence that the intended effect occurred.

The record also does not establish whether the intervention prevented breast cancer. Its rationale and title refer to prevention, but the primary endpoint was a biomarker measure, and no analyzed breast cancer outcome is reported in the registry record. A biomarker comparison, even if it had been analyzed, would not automatically establish a clinical prevention effect.

The safety entries are counts from the registry and should be interpreted as descriptive rather than definitive comparisons. They do not provide a complete account of the nature or timing of events, and the record does not report deaths. The small enrollment and early termination further limit what can be concluded from these counts.

In short, the registry record describes the intended design, endpoint, and reason the study could not evaluate its planned endpoints. It does not provide evidence that supports a conclusion for or against a celecoxib effect on the primary endpoint or cancer prevention.

13. Why This Trial Matters Statistically

This trial illustrates an important distinction between a planned randomized comparison and an analyzable result. Randomization and masking can strengthen a study's design, but neither creates evidence when the endpoint is not analyzed. Here, the registry gives a clear primary question and treatment contrast, yet explicitly states that the submitted mammographic-density images were never analyzed.

The record also shows why early closure should be reported alongside outcome data. Poor accrual and insufficient accrual to evaluate endpoints explain why the planned analysis is unavailable. Without that context, a reader might mistakenly treat the lack of a reported effect as a null result. The distinction matters: no estimate is not an estimate near zero, and no significance test is not a nonsignificant test.

Another statistical lesson is that biomarker outcomes and clinical outcomes answer different questions. Mammographic density was the primary endpoint, and Ki-67 expression was listed as secondary. The prevention rationale does not turn those measures into direct evidence about breast cancer incidence. The record does not report an analyzed cancer outcome, so the intended biomarker question and the broader prevention aim should remain separate in interpretation.

Finally, the adverse-event counts show the limits of descriptive safety data in a small, terminated study. Counts by arm are useful to report, but they do not on their own establish comparative risks or rule out uncommon harms. Transparent reporting of what was and was not analyzed helps readers avoid over-interpreting sparse observations and keeps the strength of the conclusion aligned with the available evidence.

14. Limitations

• The posted summary statistics are the only data these results rest on; patient-level data, full protocols and statistical analysis plans are not available.
• Recalculated results are simple, unadjusted calculations from posted summaries. They do not replace the sponsor's prespecified analysis.

The principal limitation is that the primary endpoint was not analyzed. The registry states that images submitted for the 1 year mammographic-density assessment were never analyzed because the study closed early due to poor accrual and had insufficient accrual to evaluate study endpoints. No treatment-effect estimate or measure of uncertainty is therefore available for the primary question.

Enrollment was 8, and the study status is terminated. The record does not report why accrual was poor, whether the intended enrollment changed, or what recruitment targets were set. It also does not report a sample-size justification, the original statistical analysis plan, or any interim analysis plan. Those details cannot be reconstructed from the generic planning assumptions.

There is limited information on outcome measurement and analysis. The record does not report image-reading procedures, baseline mammographic-density summaries, assessment reliability, the primary analysis population, or handling of missing images. It does not provide an analyzed result for the listed secondary outcome, Ki-67 expression, or a multiplicity strategy. As a result, there is no basis for evaluating modeling choices, covariate adjustment, or sensitivity analyses.

The safety record is also limited: it provides arm-specific counts but does not list serious-event terms, and deaths are not reported in the registry record. Event timing, detailed severity, and treatment attribution are not described in the registry record. The counts should not be treated as comprehensive comparative safety evidence. Overall, these limitations mean the registry is informative about the intended study and its early termination, but not about the magnitude or precision of the planned efficacy comparison.

15. Related Tutorials

16. Related Statistical Calculators

17. Sources

• ClinicalTrials.gov: S0300, Celecoxib in Preventing Breast Cancer in Premenopausal Women — NCT00088972.
• Registry record last updated 2018-08-10; status last verified 2018-07.
• Results first posted 2013-04-09.
• Data retrieved from ClinicalTrials.gov on 2026-11-09.
• Group sizes, means, event counts and sponsor-reported p-values are reproduced as posted. Effect estimates, confidence intervals and p-values not attributed to the sponsor are independently calculated from the posted summary statistics.

18. Record Summary

S0300 was a randomized, double-masked, parallel phase II trial sponsored by SWOG Cancer Research Network. It compared oral celecoxib with oral placebo, each taken twice daily for 12 months unless unacceptable toxicity or cancer diagnosis occurred first. The intended population was premenopausal women at high risk for breast cancer. Enrollment was 8, and the registry lists the study as terminated.

The primary endpoint was change in mammographic density over 1 year. The registry says the study closed early due to poor accrual, had insufficient accrual to evaluate endpoints, and the submitted primary-outcome images were never analyzed. Ki-67 expression was listed as a secondary outcome, and the record lists 0 analyses. Arm-specific adverse-event counts are posted, but they do not support a complete comparative safety assessment. The registry record does not report a primary treatment-effect estimate or its precision.