Metformin Hydrochloride in Preventing Breast Cancer in Patients With Atypical Hyperplasia or In Situ Breast Cancer: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Metformin Hydrochloride in Preventing Breast Cancer in Patients With Atypical Hyperplasia or In Situ Breast Cancer (NCT01905046), a randomized phase 3 study. The primary outcome is Test for the Presence or Absence of Cytological Atypia in Unilateral or Bilateral RPFNA Aspirates After 12 Months..

2. Clinical Question

3. Trial Design

This is a phase 3, randomized, double-masked, crossover trial listed as NCT01905046. The registry record gives an enrollment of 86 and identifies the sponsor as the Alliance for Clinical Trials in Oncology. The listed status is active, not recruiting. The recorded start date is 2015-11-23, and the primary completion date is 2023-11-10.

The trial evaluates metformin hydrochloride against placebo in patients with atypical hyperplasia or in situ breast cancer. Its official title specifies testing for atypia in RPFNA cytology after 12 months of metformin chemoprevention versus placebo control in premenopausal women. Randomization and double masking are reported, but details of the randomization procedure, allocation concealment, and who was masked are not reported in the registry record.

The primary comparison concerns the 12-month assessment. The metformin arm description includes treatment through months 13–24, while the placebo arm description includes placebo during months 0–12 and the option to cross over to metformin during months 13–24. The treatment schedules therefore matter when interpreting later assessments: after the initial placebo period, some participants may receive metformin. The registry record also describes a 24-month RPFNA for the metformin arm and follow-up visits at 36 and 48 months after treatment begins.

The record lists secondary assessments involving Masood scores, atypia or disappearance of atypia, reproducibility of duplicate RPPM determinations, and associations between baseline RPPM values and atypia at later time points. It does not report results for these assessments. Nor does it report the full protocol, detailed follow-up procedures, or a complete statistical analysis plan.

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.

Months 0-12

Months 13-24

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: Per protocol, evaluable patients are defined as having a baseline Masood score between 14 and 17 in at least one breast, and all evaluable breasts undergo satisfactory RPFNA at 12 months. At the time of the primary analysis, we had 66 evaluable patients out of 86 randomized patients (32 Placebo, 34 Metformin).

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 Test for the Presence or Absence of Cytological Atypia in Unilateral or Bilateral RPFNA Aspirates After 12 Months., measured over 12 months.

The primary endpoint is the presence or absence of cytological atypia in unilateral or bilateral RPFNA aspirates after 12 months. It is a binary participant-level classification, not a count of abnormal breasts. The registry record says that the classification is based on the average Masood score across all evaluable breasts for a patient. An average score of at least 13.5 classifies the patient as having atypia.

This endpoint is used to compare the two randomized arms at a defined time point. Turning the score into a binary outcome makes the primary comparison a comparison of the proportion classified with atypia, rather than a comparison of the full range of scores. That can make the result easier to interpret as a difference in classification, but it also means that score variation within each category is not represented in the primary endpoint. Participants near the classification threshold may be assigned to different categories despite having similar average scores.

The registry record describes an unadjusted chi-square test and a logistic regression analysis adjusted for group indicator, age, race, stratification factors, and baseline Masood score. It does not report the observed number or proportion classified with atypia in either arm, the estimated treatment contrast, or a confidence interval. The endpoint definition and planned analysis are reported, but the primary endpoint result is not reported in the registry record.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The registry record describes a binary primary endpoint: participants are classified as having atypia or not having atypia at 12 months. The classification is derived from the average Masood score across evaluable breasts, with an average of at least 13.5 defining atypia. The planned primary analysis compares that classification between randomized groups. No result from either analysis is reported in the registry record.

What does a chi-square test assess?

In general, a chi-square test can assess whether the distribution of a categorical outcome differs between groups. For a binary endpoint, the test compares the observed pattern of participants classified in each category with the pattern expected if group and outcome were unrelated. It is an unadjusted comparison: it does not account for participant characteristics or stratification factors. Its interpretation depends on the outcome counts and the test result, neither of which is reported for the primary endpoint in the registry record.

A test result alone would not describe the size or precision of a difference. A treatment contrast, such as a difference in proportions or an odds ratio, describes an aspect of the association between randomized group and the binary outcome. A confidence interval, in general, expresses the range of effect values reasonably compatible with the data and the method used. The registry record does not report a primary effect estimate or confidence interval, so neither magnitude nor precision can be assessed.

Why use logistic regression?

In general, logistic regression models the probability of a binary outcome in relation to one or more explanatory variables. The record says the primary multivariable analysis regresses the outcome on group indicator, age, race, stratification factors, and baseline Masood score. Adjustment can account statistically for differences in measured covariates included in the model. It does not automatically correct for missing outcomes, measurement limitations, unmeasured factors, or an unsuitable model. The record does not report the fitted coefficients, model diagnostics, or adjusted treatment estimate.

An adjusted association and an unadjusted comparison answer related but distinct statistical questions. The chi-square test assesses the group-outcome relationship without covariate adjustment; the logistic regression evaluates the relationship conditional on the listed model variables. If the estimates differ, possible explanations can include covariate imbalance or the effect of the model specification. The registry record does not provide results that would allow those comparisons. It also does not report whether the adjusted estimate was expressed as an odds ratio, a model-based probability, or another effect measure.

What does the endpoint threshold imply?

Converting a score-based measure into a binary classification discards detail within each category. Two participants with different average scores can receive the same classification, while scores on opposite sides of the threshold can receive different classifications even when they are close. In general, dichotomization can simplify a primary comparison but may reduce information about gradations in the measured characteristic. The record does not report sensitivity analyses using the continuous Masood score or alternative thresholds.

The primary measure averages scores across evaluable breasts for each participant. This makes the participant, rather than each breast, the unit of the primary outcome. The record does not provide the underlying breast-level data or explain in further detail how measurements from multiple breasts were aggregated when only one or more breasts were evaluable. It does state that all evaluable breasts must undergo satisfactory RPFNA at 12 months for a patient to meet the evaluability definition.

How should the planning calculations be read?

In general, a sample-size calculation estimates the amount of information a future study may need under specified assumptions about an effect and analysis. The registry record describes planning illustrations based on a stated fraction of an observed effect, a two-sided significance level of 0.05, equal allocation, and a normal approximation. Such illustrations are planning exercises, not observed power and not evidence about the trial's result. The primary outcome is not reduced to a single comparison in the available calculations, and no planning illustration is reported as an outcome of this trial. The record does not report the trial's own sample-size justification or the assumptions used to design it.

In general, statistical significance and clinical importance are not interchangeable. A small estimated effect can be statistically distinguishable from no difference when its uncertainty is sufficiently limited, while an estimate that is not statistically distinguishable from no difference can still be compatible with effects that matter. Interpreting either situation requires the effect estimate and its uncertainty, alongside context for the endpoint. Those primary result statistics are not reported in the registry record.

The trial is described as randomized and double-masked, features that define its design. The record does not provide enough detail to evaluate implementation of randomization or masking. It also does not report the complete analysis plan, handling of multiplicity, missing-data methods, or model checks. These are important components of statistical interpretation, but their absence should not be filled in by assuming conventional choices.

10. Multiplicity, Missing Data and Analysis Populations

The registry record lists a primary endpoint and several secondary assessments. The secondary outcomes include Masood score and atypia assessments at specified time points, comparison of scores between breasts, reproducibility of duplicate determinations, and correlations of baseline values with later atypia. When a study assesses multiple outcomes, time points, or analyses, the chance of obtaining at least one apparently notable result can increase. In general, multiplicity procedures can address that issue, but the record does not report whether any were planned or used. A correction or hierarchy should not be assumed.

The primary analysis population is defined by baseline and follow-up criteria. Evaluable patients must have a baseline Masood score between 14 and 17 in at least one breast and satisfactory RPFNA at 12 months for all evaluable breasts. At the time of the primary analysis, 66 of 86 randomized patients were reported as evaluable. The registry record does not report the reasons for non-evaluability in sufficient detail, the amount or pattern of missing endpoint data by arm, or the method used to handle missing observations.

In general, an analysis restricted to participants with evaluable outcomes can be affected if the availability of measurements is related to treatment assignment or outcome. Randomization supports comparison of assigned groups, but exclusions after randomization can change which participants contribute to an analysis. The record does not say whether an intention-to-treat analysis was also conducted, whether imputation was used, or whether sensitivity analyses assessed missing-data assumptions. These details are not reported in the registry record, so the impact of missingness on the primary result cannot be evaluated.

The record describes adjusted and unadjusted primary analyses, but does not report which analysis was designated as the main inferential basis if their findings differed. It also does not report a complete analysis plan or whether the listed secondary assessments were adjusted for multiple testing.

11. Safety Results

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

Deaths

Other (non-serious) adverse events

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

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: resp, thoracic, mediastinal - oth spec (2.4%), thromboembolic event (2.4%), abdominal distension (1.8%), diarrhea (1.8%), nausea (1.8%).

The registry record gives serious adverse event counts separately for Arm I, the placebo period in Arm II, and the later metformin period in Arm II. The reported serious counts are 1 among 42 at risk in Arm I, 0 among 44 at risk during the placebo period, and 2 among 57 at risk during the later metformin period. The record lists 0 deaths in each of these groups. These are descriptive figures for the reported group-period definitions; they are not a randomized comparison of identical exposure periods.

The record also lists serious adverse event terms with percentages: “Resp, thoracic, mediastinal - Oth spec” and “Thromboembolic event” are each listed as 2.4%; “Abdominal distension,” “Diarrhea,” and “Nausea” are each listed as 1.8%. The registry record does not identify an arm or period for each listed term in this summary, so those percentages should not be assigned to a particular treatment group.

In general, adverse-event counts describe events recorded under the study's reporting definitions; they do not by themselves establish causation. The registry record does not report a formal comparative safety analysis, event timing, severity breakdown beyond the serious-event designation, or an attribution assessment. The counts should therefore be interpreted descriptively and in the context of the different treatment periods and denominators.

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

The record establishes the clinical and statistical question: a randomized comparison of metformin hydrochloride and placebo, using a 12-month binary cytological atypia endpoint based on average Masood scores. It also specifies a planned unadjusted chi-square comparison and adjusted logistic regression. These details explain what the study set out to measure and how the primary comparison was intended to be made.

The registry record does not report the primary endpoint result. It gives no arm-specific atypia counts or proportions, no estimated difference or odds ratio, no confidence interval, and no primary test result. Therefore, it is not possible to determine from this record whether the observed groups differed, the direction or magnitude of any difference, or how precisely a difference was estimated. No conclusion about the primary outcome should be inferred from the study title, stated purpose, or treatment schedule.

In general, a trial's question and design are different from its findings. Randomization describes how treatment groups were assigned; it does not itself show the outcome pattern. A planned analysis describes how data may be compared; it is not evidence that a comparison favored either group. Likewise, safety counts are not a substitute for the missing primary efficacy results, and the reported adverse-event figures do not establish that treatment caused an event.

The primary analysis concerns evaluable patients meeting stated baseline and RPFNA criteria. Because the record does not fully describe missing-data handling or the relationship between the evaluable group and all randomized participants, the extent to which the analysis represents the randomized population cannot be fully assessed. The possible crossover after the placebo period is also relevant to later assessments, though the primary comparison is specified at 12 months.

The appropriate reading is narrow: the registry record documents a planned phase 3 randomized, double-masked comparison and an endpoint definition, but not the primary result. Any claim about an observed benefit, lack of benefit, or treatment effect would require results not reported in the registry record.

13. Why This Trial Matters Statistically

This trial raises several statistical issues that matter for interpreting biomarker and cytology outcomes in a randomized setting. The primary measure is not a direct count of cancer diagnoses. It is a participant-level classification of atypia based on the average Masood score across evaluable breasts. Keeping the distinction between a cytological endpoint and a clinical cancer outcome clear is essential: the record identifies the endpoint it measures, and does not report a cancer-incidence result as the primary outcome.

The threshold of at least 13.5 creates a binary endpoint from a score. In general, this can simplify comparisons, but it removes information about score differences within each classification and makes results potentially sensitive to observations near the cutoff. Reporting both the classification and the underlying score distribution can help readers understand what the binary result represents. The registry record lists secondary Masood score assessments, but does not report their results.

The design also includes a possible crossover after the first 12 months. That feature means that later comparisons need to be interpreted in light of treatment received during earlier periods. The primary comparison is specified at 12 months, before the described crossover period, while later observations may reflect a different treatment sequence. The registry record does not report outcome results at either time point.

Finally, the primary analysis is described for an evaluable subset, with criteria based on baseline score and satisfactory follow-up RPFNA. Understanding how a subset is formed, and how missing measurements are handled, is important because the randomized groups may not be fully represented in an evaluable-only analysis. The record reports the evaluable count but not the complete missing-data analysis. These features make the absence of treatment-effect estimates and uncertainty especially consequential: design and methods alone cannot answer the comparative question.

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 registry record does not report the primary endpoint findings, including arm-specific outcome counts, an estimated treatment effect, a confidence interval, or a test result. As a result, neither the size nor the precision of the primary comparison can be assessed. The record also does not report results for the listed secondary outcomes.

The stated primary analysis population consists of evaluable patients meeting baseline Masood score and 12-month RPFNA requirements. Although the record reports 66 evaluable patients out of 86 randomized patients, it does not give a complete account of why participants were not evaluable, how missing measurements were handled, or whether results from an intention-to-treat analysis are available. This limits assessment of potential selection related to inclusion in the primary analysis.

The endpoint is a binary classification based on an average score threshold. The registry record does not report the distribution of underlying scores, the number of breasts contributing to individual averages, or sensitivity analyses using the score without dichotomization or a different threshold. Thus, the robustness of the classification-based result cannot be assessed.

The record describes a crossover option after the placebo period, but does not provide enough detail to evaluate adherence, treatment exposure, or the interpretation of later comparisons. It also does not report the full randomization and masking procedures, the complete statistical analysis plan, multiplicity handling, model diagnostics, or sample-size justification. These limitations concern information available in the registry record and do not establish that the study lacked such procedures.

The safety summaries are descriptive and divided by group and treatment period. The record does not provide a formal comparative safety analysis, event attribution, or a complete account of event timing. The listed serious event terms are not linked to a specific arm in the summary.

15. Related Tutorials

16. Related Statistical Calculators

17. Sources

• ClinicalTrials.gov: Metformin Hydrochloride in Preventing Breast Cancer in Patients With Atypical Hyperplasia or In Situ Breast Cancer — NCT01905046.
• Registry record last updated 2026-10-08; status last verified 2026-04.
• Results first posted 2025-04-13.
• 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

NCT01905046 is listed as a randomized, double-masked, phase 3 crossover trial sponsored by the Alliance for Clinical Trials in Oncology, with enrollment of 86. It evaluates metformin hydrochloride against placebo in premenopausal women with specified atypical hyperplasia or in situ breast conditions. The primary endpoint is presence or absence of cytological atypia at 12 months, classified using average Masood scores across evaluable breasts and a threshold of at least 13.5.

The record describes a chi-square comparison and an adjusted logistic regression analysis, and reports 66 evaluable patients at the time of the primary analysis. It lists serious adverse event counts by group and period, with 0 deaths in each listed group. The registry status is active, not recruiting. Primary outcome results, effect estimates, confidence intervals, missing-data methods, and multiplicity procedures are not reported in the registry record.