Brain Function in Premenopausal Women Receiving Tamoxifen With or Without Ovarian Function Suppression for Early-Stage Breast Cancer on Clinical Trial IBCSG 24-02: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Brain Function in Premenopausal Women Receiving Tamoxifen With or Without Ovarian Function Suppression for Early-Stage Breast Cancer on Clinical Trial IBCSG 24-02 (NCT00659373), a randomized phase 3 study. The primary outcome is Change in Cognitive Function Over 1 Year in Premenopausal Breast Cancer Patients Who Receive Adjuvant Tamoxifen (T) Alone Against Those Receive Adjuvant Tamoxifen (T+OFS) or Exemestane (E+OFS) With Ovarian Function Suppression (OFS).

2. Clinical Question

3. Trial Design

The registry describes a randomized, parallel-group, Phase 3 study with no masking. The study is titled as an investigation of brain function in premenopausal women receiving tamoxifen, with or without ovarian function suppression, for early-stage breast cancer. Its official title places the cognitive assessment within selected centers participating in the SOFT trial. The registry lists enrollment of 86 and a completed status. The recorded start is 2007-12, and primary completion is 2014-12.

The study description names three treatment strategies: tamoxifen alone, tamoxifen with ovarian function suppression, and exemestane with ovarian function suppression. It describes tamoxifen at 20mg orally daily for 5 years; the tamoxifen-plus-suppression regimen adds suppression, and the exemestane-plus-suppression regimen uses exemestane 25mg orally daily for 5 years with suppression. The listed suppression approaches include triptorelin 3.75 mg by intramuscular injection every 28 days for 5 years, surgical oophorectomy, or ovarian irradiation.

The posted primary outcome, however, presents two result groups: Tamoxifen and Ovarian Function Suppression. It does not make clear whether the latter is a single combination regimen or a broader grouping, nor does it present separate cognitive results for all three strategies described in the study summary. The posted outcome includes 20 participants in the Tamoxifen group and 54 in the Ovarian Function Suppression group. The original allocation details, analysis population, and reasons that the outcome group labels differ from the three regimens in the study description are not reported in the registry record.

The cognitive assessment was planned at baseline and one year after randomization to the parent study. The posted primary endpoint specifies the one-year time point. The registry does not describe masking of participants or assessors beyond listing the study as having no masking, and does not report the original statistical design assumptions.

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.

Before assignment: Data were collected separately for the T+OFS and E+OFS participants in the parent study, IBCSG 24-02 (SOFT). The sample size for this Co-SOFT substudy was small, so the analysis plan was revised to pre-specify collective analysis for all patients receiving OFS.

Overall Study

Reasons for not completing this period:

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 Change in Cognitive Function Over 1 Year in Premenopausal Breast Cancer Patients Who Receive Adjuvant Tamoxifen (T) Alone Against Those Receive Adjuvant Tamoxifen (T+OFS) or Exemestane (E+OFS) With Ovarian Function Suppression (OFS), measured over 1 year after patient randomization to parent IBCSG 24-02 study.

The primary endpoint is change in cognitive function over one year, measured using CogState, a computerized battery of seven tasks. Tasks cover detection, identification, monitoring, memory, learning, and immediate and delayed performance on the International Shopping List Task. Speed is assessed for three tasks using reaction time for correct responses; accuracy is assessed for the other tasks using the proportion of correct responses. The registry states that lower transformed speed scores indicate better performance, while higher transformed accuracy scores indicate better performance.

The main outcome combines task scores after transformation and standardization by age-specific norms. It is expressed in standardized units. A positive composite score indicates performance better than average, and a negative score indicates below-average performance. A composite can summarize several task domains in one measure, rather than treating each task as a separate primary result. That can make an overall comparison easier to interpret, although the combined score does not show whether any particular task or domain changed differently from the others.

The endpoint is relevant to the stated research question because it quantifies cognitive test performance at the specified follow-up. The registry says participants complete assessments at baseline and one year after randomization, but the posted group summaries do not provide task-level results, baseline scores, or details about how change was calculated for individuals. Those details are not reported in the registry record.

7. Results Posted in the Registry

• Ovarian Function Suppression: mean -0.21 versus -0.04 on control; difference -0.17 (95% CI -0.50 to 0.16), standardized effect d = -0.21; z-test p = 0.307.

8. Statistical Interpretation

9. Statistical Methods Explained

The registry reports a continuous outcome: a standardized composite cognitive score. For a continuous measure, group summaries commonly describe the average and the spread of observed scores. The posted means are -0.04 for Tamoxifen and -0.21 for Ovarian Function Suppression; the corresponding standard deviations are 0.49 and 0.92. The standard deviation describes variation among observations within a group. It is not the uncertainty around the group mean, and it is not the same as a confidence interval.

What does the mean difference describe?

The unadjusted recalculation compares the two posted averages by subtracting the Tamoxifen mean from the Ovarian Function Suppression mean. The resulting difference is -0.17. Under that direction of comparison, a negative value means the posted average is lower in the group labeled Ovarian Function Suppression. The difference retains the endpoint’s standardized-unit scale. It is a simple comparison of summary statistics and does not account for covariates or other features of the trial design.

General: a mean difference is useful when an outcome has a meaningful continuous scale and the group means are informative summaries. It gives both direction and magnitude in the units of the outcome. It can conceal important variation, however: the same mean difference can arise from different distributions, and a group average does not describe every participant’s score. Here, the reported standard deviations indicate that observations vary within each group, with greater spread in the group labeled Ovarian Function Suppression.

Why are there two different tests?

The registry-reported analysis uses the Wilcoxon (Mann-Whitney) method and reports p = 0.71. This is a rank-based comparison: observed scores are ordered, and the test assesses whether the ranks tend to differ between groups. It does not directly test the same quantity as a mean-difference calculation. Its interpretation as a comparison of medians requires additional assumptions about the shapes of the group distributions; the registry does not report distributional checks or state that interpretation.

The recalculated mean comparison instead uses the posted means and standard deviations to estimate an unadjusted difference, interval, and p-value. Its p-value is 0.307. Different methods can produce different p-values because they use different features of the data and can target different aspects of a distribution. The two p-values are therefore not interchangeable, and the recalculated test does not replace the registry-reported rank-based analysis.

General: a p-value measures how unusual the observed data, or data more extreme, would be under a specified null model and its assumptions. It is not the probability that the null hypothesis is true, and it does not quantify the size or practical importance of a difference. Neither p-value alone tells whether the result is precise. The estimate and its interval are also important.

What does the interval add?

The recalculated interval for the mean difference crosses zero. General: a confidence interval describes a range of values compatible with the data and the interval procedure, under its assumptions. An interval that includes zero indicates that the data do not rule out no mean difference at the corresponding interval level. An interval may also include differences in either direction. Its width reflects uncertainty; a wide interval allows a broader range of possible effects than a narrow one. The interval level is not specified in the registry record, so it should not be assigned a confidence percentage.

The recalculation is unadjusted and based on posted summary statistics. It cannot reproduce analyses that might use participant-level measurements, covariate adjustment, repeated measures, or other modeling choices. Whether the original analysis used any such methods beyond the reported Wilcoxon test is not reported in the registry record. The original handling of baseline performance and missing observations is also not reported.

Why can variation matter?

General: when data are more dispersed than a simple model assumes, a model that ignores that extra variation can estimate uncertainty poorly. In a continuous outcome, the standard deviations help describe observed spread; other designs, such as count outcomes, may require explicit allowance for over-dispersion when variability exceeds a model’s expectation. The present endpoint is continuous, and the registry does not report an over-dispersion analysis. The relevant general caution is to match the statistical model to the outcome distribution and design rather than assume that a single summary or test describes every feature of the data.

The outcome is a composite of seven task scores after transformation and standardization. This standardization places the combined result on a common scale according to the stated age-specific norms, but does not make the score a direct measure of every aspect of cognition. The registry does not report the formula used to combine task scores beyond describing an average after transformation and standardization, nor does it report task-specific estimates or reliability information.

10. Confidence Intervals and Precision

The posted summaries show a mean of -0.04 with standard deviation 0.49 for Tamoxifen, and a mean of -0.21 with standard deviation 0.92 for the group labeled Ovarian Function Suppression. The recalculated unadjusted mean difference is -0.17. Its confidence interval crosses zero. This means the recalculation does not isolate a single direction of mean difference with the reported precision; values on both sides of no difference remain compatible with the interval.

General: a confidence interval conveys uncertainty about an estimate, conditional on the method and assumptions used to construct it. The interval is more informative about precision than a p-value alone because it indicates which effect sizes remain plausible under that procedure. A narrow interval would support a more precise estimate; a wide interval would leave greater uncertainty. No numerical interval bounds are quoted because the registry’s recalculated bounds are recorded to more decimal places than the stated display limit. The registry does not specify the confidence level for this interval.

The posted group sizes are 20 and 54. These counts and the observed standard deviations are relevant context for precision, but they do not establish how representative the analyzed participants are or whether any observations were missing. The registry does not report the analysis population, missing-data handling, or a confidence interval for the sponsor-reported Wilcoxon analysis. Precision should therefore be understood in relation to the unadjusted mean comparison, not as a complete account of uncertainty in every analysis or treatment group described by the study.

11. Planning a Similar Trial: Sample Size

This is a planning illustration, not a re-evaluation of the original trial. It asks: if a new study wanted to detect a difference like the one observed, or a smaller one, how large would it need to be? It is deliberately not “observed power” of the finished trial, which is not a meaningful quantity.

The planning figures are participants per arm, not total event counts. They are illustrations based on the observed effect, rather than the original trial’s design assumptions. The stated planning assumptions are a two-sided significance level of 0.05, equal allocation between two arms, and a normal approximation. The effect to detect is set to the stated fraction of the observed effect. The original sample-size assumptions for the trial are not reported in the registry record.

The 80% power column gives the planned participants per arm for a design with an 80% chance of detecting the specified effect under the planning assumptions, if that effect and the assumptions are correct. The 90% power column gives the corresponding per-arm planning figure for a 90% chance. These are design calculations, not descriptions of the probability that the completed study found a result or of the probability that its conclusion is correct.

The planning illustration lists 373 participants per arm for 80% power and 500 for 90% power when planning for 100% of the observed effect. For 75% of the observed effect, the figures are 663 and 888 participants per arm. For 50% of the observed effect, the figures are 1,492 and 1,997 participants per arm. The pattern reflects the general principle that smaller effects require more participants to detect with the same planned power, all else held constant.

These figures are not observed power. They should not be read as evidence that the completed study was adequately or inadequately powered, nor as the trial’s original target. They are hypothetical planning illustrations tied to an observed unadjusted effect and stated statistical assumptions. A reader considering a future study would also need to assess whether the selected effect is scientifically meaningful and whether the planned design matches the intended treatment comparison and analysis.

12. How the Sample Size Is Derived

Power is the probability that a study of a given size rejects the null hypothesis when a specific true effect exists. It depends on four things: the effect to detect, the variability of the outcome, the significance level (here a two-sided α of 0.05, so z_1−α/2 = 1.960) and the desired power (z_1−β = 0.842 for 80% and 1.282 for 90%).

Formula

n is the number of participants in each arm, Δ is the difference in means to detect and σ is the common standard deviation.

Worked calculation for this trial

1. Observed difference Δ = -0.170; pooled standard deviation σ = 0.828.
2. n = 2 × (1.96 + 0.842)² × 0.828² / -0.170² = 372.8, rounded up to 373 participants per arm for 80% power.

The smaller-effect rows in the table above repeat the same calculation with the effect multiplied by 0.75 and 0.5 (for ratios, applied to the logarithm of the ratio).

R code

alpha <- 0.05
z_a   <- qnorm(1 - alpha / 2)          # two-sided
delta <- -0.17                      # observed difference in means
sigma <- 0.828495121691532                     # pooled standard deviation
n_per_arm <- function(power, frac = 1) {
  ceiling(round(2 * (z_a + qnorm(power))^2 * sigma^2 / (delta * frac)^2, 6))
}
for (f in c(1, 0.75, 0.5))
  cat(sprintf("%3.0f%% of observed effect: n per arm = %s (80%% power), %s (90%% power)\n", 100 * f,
              format(n_per_arm(0.80, f), big.mark = ","), format(n_per_arm(0.90, f), big.mark = ",")))

# Check against R's built-in function (t-based, so very slightly larger)
cat("power.t.test:", round(power.t.test(delta = delta, sd = sigma, power = 0.80)$n, 1), "\n")

# Simulation: exact two-sample t-test using simulated means and variances
set.seed(659373)
n    <- n_per_arm(0.80)
reps <- 20000
m_diff <- rnorm(reps, delta, sigma * sqrt(2 / n))             # difference in sample means
s2     <- sigma^2 * rchisq(reps, 2 * (n - 1)) / (2 * (n - 1))   # pooled sample variance
tstat  <- m_diff / sqrt(s2 * 2 / n)
cat(sprintf("Simulated power at n = %s per arm: %.3f (%d replicates)\n", format(n, big.mark = ","), mean(abs(tstat) > qt(0.975, 2 * (n - 1))), reps))

Output from running the code

100% of observed effect: n per arm = 373 (80% power), 500 (90% power)
 75% of observed effect: n per arm = 663 (80% power), 888 (90% power)
 50% of observed effect: n per arm = 1,492 (80% power), 1,997 (90% power)
power.t.test: 373.8 
Simulated power at n = 373 per arm: 0.799 (20000 replicates)

The simulation draws many sets of group means and variances at the planned sample size and counts how often the two-sample t-test rejects the null. It should land close to the target power; at very small required sample sizes it can fall below it, because the normal approximation is rough there.

13. Analyses Reported by the Sponsor

The registry-reported analysis compares the two groups using the Wilcoxon (Mann-Whitney) test and reports p = 0.71. The simple recalculation from the posted group summary statistics instead estimates an unadjusted mean difference of -0.17, with p = 0.307 and a confidence interval that crosses zero. These are not identical analyses: the sponsor-reported method is rank-based, while the recalculation concerns the difference between arithmetic means. Their p-values therefore need not match.

The sponsor-reported result should be read as the registry’s own stated analysis. The recalculation is a supplementary, unadjusted description derived from the posted summary values; it does not reproduce participant-level analysis or establish how the original investigators handled baseline scores, covariates, or missing data. The record does not provide the sponsor’s test statistic, interval estimate, or further interpretation of the rank-based result.

The two analyses both fail to provide strong evidence of a group difference under their respective tests, but they address different statistical features. Neither establishes equivalence, and neither should be treated as a complete account of uncertainty. The posted outcome labels also leave the exact composition of the ovarian-function-suppression group unclear, limiting interpretation of either analysis as a comparison of specific regimens.

14. Multiplicity, Missing Data and Analysis Populations

The registry lists one primary outcome measure and one reported analysis. It lists no posted secondary outcomes. That does not establish whether other outcomes, task-level results, or planned analyses were assessed but not entered in the registry; those details are not reported in the registry record. The record also does not state whether the study had a prespecified multiplicity plan covering multiple tasks, time points, comparisons, or outcomes.

General: multiplicity arises when several statistical tests are conducted. Even if each test uses a conventional threshold, conducting many tests can increase the chance of at least one apparently unusual result. Prespecified outcome hierarchies or adjustment procedures can help define how multiple results should be interpreted. The registry does not report such procedures, so the extent to which multiplicity was addressed cannot be assessed from this record.

Missing-data methods are also not reported in the registry record. The outcome description says participants were assessed at baseline and one year after randomization, but it does not state how many lacked a one-year assessment, how missing scores were handled, or whether any sensitivity analyses were conducted. The posted group counts for this outcome are 20 and 54; the record does not explain how those groups relate to the registry’s enrollment figure or to the randomized analysis population. No inference about retention or exclusions should be made from those counts alone.

The analysis population is not specified. The registry does not state whether the posted result is based on all randomized participants, participants with completed assessments, or another defined set. It also does not report whether group assignment was retained for analysis regardless of treatment received. These details matter because exclusions, nonadherence, and missing outcome data can affect how a randomized comparison is interpreted.

Finally, the study description names three treatment strategies while the posted endpoint result has two labels. The record does not explain the grouping decision or provide separate results for each named strategy. Without that information, it is not possible to evaluate whether the reported comparison corresponds to a prespecified contrast or how multiplicity across the treatment strategies was handled.

15. 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.

The registry reports zero serious adverse events among the 20 participants listed as at risk in the Tamoxifen group and zero among the 54 listed as at risk in the Ovarian Function Suppression group. It also reports zero other adverse events in each of those groups. These entries describe what was recorded in this registry outcome record; they do not establish that no adverse events occurred outside the reported categories or that a treatment has no risks.

The deaths field is blank for both groups. Death counts are therefore not reported in the registry record, rather than being stated as zero. The registry does not provide event definitions, follow-up duration for adverse-event collection, severity breakdowns beyond the serious-event category, or information about attribution to treatment. It also does not list serious adverse-event terms.

General: safety counts depend on how events are defined, collected, and reported, as well as the population and follow-up represented. A zero recorded count in a group is not evidence that the true event risk is zero. These figures should be understood as registry-reported counts within the stated at-risk groups, not as a comprehensive safety assessment or a basis for medical decisions.

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

The posted result describes one-year cognitive test scores in a comparison between Tamoxifen and a group labeled Ovarian Function Suppression. The group means are -0.04 and -0.21, respectively, on the standardized composite scale. The recalculated mean difference is -0.17 in the direction of the ovarian-function-suppression label relative to Tamoxifen. The interval crosses zero, and the sponsor-reported rank-based test gives p = 0.71. Together, these results do not identify a clear difference in the posted cognitive outcome, while leaving uncertainty about the size and direction of any underlying difference.

This is not proof that the groups are equivalent. Equivalence requires a meaningful margin defined in advance and an analysis designed to determine whether the uncertainty falls within that margin. The registry does not report an equivalence margin or an equivalence analysis. Nor does a non-significant test establish that there is no difference.

The endpoint is a standardized composite built from seven computerized tasks. It can summarize overall performance across the included tasks, but does not reveal which cognitive domains drive the group averages. The registry does not post task-specific results or baseline group summaries, and does not describe the handling of missing assessments. It is therefore not possible to determine from this record how the composite’s individual components behaved or how missing data may affect interpretation.

The treatment comparison is also not fully resolved by the outcome labels. The study description names tamoxifen alone, tamoxifen with ovarian function suppression, and exemestane with ovarian function suppression, while the result presents Tamoxifen and Ovarian Function Suppression. The record does not clarify which regimen or regimens make up the latter group. The result should not be attributed to a specific combination regimen without that clarification.

Finally, this finding concerns the stated cognitive measure at one year after randomization. It does not describe all aspects of cognition, longer-term outcomes, cancer outcomes, or individual experience. It is a statistical summary of the registry-posted comparison, not a clinical recommendation or a conclusion about the overall benefits and risks of treatment.

17. Why This Trial Matters Statistically

This trial raises several statistical issues that are common in clinical research. First, the main outcome is a composite assembled from different cognitive tasks. Combining standardized task scores can create a concise overall measure, but the resulting average may conceal variation among components. Without task-level results, readers cannot tell whether the overall pattern is consistent across the tasks or is influenced more by particular domains.

Second, the registry includes both a rank-based sponsor analysis and a recalculated comparison of means. The methods do not ask exactly the same statistical question. This illustrates why a p-value needs to be interpreted alongside the outcome definition, test method, effect estimate, and uncertainty interval. A single threshold-based label would not capture all of those features.

Third, the study description and posted outcome use different levels of treatment detail. Three treatment strategies are described, but the posted result has two groups and does not explain the composition of the ovarian-function-suppression group. Clear correspondence between randomized arms and analyzed outcome groups is essential for interpreting a treatment contrast and assessing whether it answers the intended question.

Finally, the record does not report the original sample-size assumptions, analysis population, or missing-data approach. Those details affect interpretation of precision and of the relationship between randomization and the posted result. The planning illustration can help explain how assumed effect size affects future sample-size needs, but it cannot substitute for the trial’s original design rationale. Taken together, these issues show why statistical interpretation depends not only on the observed means and p-values, but also on transparent reporting of outcomes, groups, and analysis choices.

18. 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 leaves important details unresolved. The study description names three endocrine-treatment strategies, but the primary outcome result contains two groups, with the second labeled Ovarian Function Suppression. The record does not specify whether that label refers to tamoxifen with suppression, exemestane with suppression, or a pooled group. Consequently, the posted contrast cannot be assigned confidently to a particular regimen.

The outcome record gives group sizes, means, and standard deviations, but does not post baseline group summaries, individual change data, task-specific cognitive results, or a detailed account of how the composite was calculated beyond averaging transformed and standardized task scores. The registry does not report missing-data handling, the analysis population, or whether the one-year result includes all randomized participants. These omissions limit evaluation of the change measure and the representativeness of the analyzed groups.

The sponsor-reported result includes a p-value for a Wilcoxon (Mann-Whitney) analysis but no test statistic or interval estimate. The unadjusted recalculation uses summary statistics and cannot incorporate any participant-level modeling or adjustment that may have been used in an original analysis. The registry does not report whether such additional methods were used.

Multiplicity procedures, the original sample-size rationale, and any interim analyses are not reported in the registry record. The posted safety fields likewise do not explain event definitions or adverse-event follow-up, and death counts are not reported. These reporting gaps do not show that the relevant procedures were absent; they mean they cannot be assessed from this record.

19. Related Tutorials

20. Related Statistical Calculators

21. Sources

• ClinicalTrials.gov: Brain Function in Premenopausal Women Receiving Tamoxifen With or Without Ovarian Function Suppression for Early-Stage Breast Cancer on Clinical Trial IBCSG 24-02 — NCT00659373.
• Registry record last updated 2016-07-25; status last verified 2016-07.
• Results first posted 2016-07-25.
• 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.

22. Record Summary

This completed randomized, parallel-group, unmasked Phase 3 study concerns cognitive function in premenopausal women with early-stage breast cancer receiving adjuvant endocrine therapy. The registry describes three treatment strategies, while the posted primary cognitive outcome compares a Tamoxifen group with a group labeled Ovarian Function Suppression.

The primary measure is a standardized composite from seven CogState tasks, assessed at one year after randomization. Posted group means are -0.04 and -0.21. The sponsor-reported Wilcoxon (Mann-Whitney) analysis gives p = 0.71; the unadjusted recalculation of the mean difference is -0.17, with an interval crossing zero and p = 0.307.

The exact composition of the ovarian-function-suppression result group, the analysis population, missing-data handling, and original design assumptions are not reported in the registry record. The record reports zero serious and other adverse events in the listed at-risk groups; deaths are not reported.