S0221 Adjuvant Doxorubicin, Cyclophosphamide, and Paclitaxel in Treating Patients With Breast Cancer: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for S0221 Adjuvant Doxorubicin, Cyclophosphamide, and Paclitaxel in Treating Patients With Breast Cancer (NCT00070564), a randomized phase 3 study. The primary outcome is Disease-free Survival.

2. Clinical Question

The registry describes a randomized, open-label, parallel-group phase III trial with enrollment of 3294. Its stated purpose was to compare chemotherapy regimens after surgery for breast cancer. The treatments varied by the schedule of doxorubicin and cyclophosphamide and by the paclitaxel schedule. The registry lists doxorubicin, cyclophosphamide, paclitaxel, and growth-factor support among the interventions.

The central statistical question is whether the observed disease-free survival patterns differed among these schedules, and how precise those estimates are. The posted arm percentages provide a year-five summary, while the sponsor’s log-rank analyses use time-to-event comparisons. Those are related but not identical summaries: one presents the estimated endpoint at a specified time, while the other compares event-time patterns over follow-up. The registry record does not report the complete original statistical plan or the rationale for the chosen enrollment target.

3. Trial Design

This was a randomized, phase III, parallel-group trial without masking. The registry lists enrollment of 3294 and identifies SWOG Cancer Research Network as the sponsor. The record gives a start date of 2003-11, a primary completion date of 2015-01, and a status of active, not recruiting. Those record-level details describe the trial’s registration history and status, rather than the timing of an individual participant’s follow-up.

Six arms are included in the posted primary outcome results. Arms I through IV are marked as closed 11/10/10 in the arm descriptions; Arms V and VI are also listed. The available descriptions indicate different schedules of doxorubicin and cyclophosphamide, followed by paclitaxel on either a weekly or every-two-weeks schedule in the arms where that treatment is described. Growth-factor support is also listed. Because several arm descriptions end mid-sentence, some regimen details are not reported in the registry record.

The primary outcome population was limited to eligible and analyzable participants. The registry specifically states that two participants in Arm II and one in Arm IV with no follow-up were excluded. The arm-level primary results report denominators separately, and those denominators differ. The record does not establish whether other exclusions, missing observations, or analysis-population rules applied beyond the statement attached to the primary endpoint.

The registry lists nine outcome measures and 22 analyses. It does not report the original sample-size justification, the prespecified analysis plan, allocation concealment details, interim analyses, or a multiplicity strategy. Randomization supports a comparative design, but interpretation of the reported comparisons also depends on the analysis definitions and reporting choices, some of which are not available in the registry record.

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

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: Only eligible and analyzable patients were included in this analysis. Two patients in Arm II and one patient in Arm IV with no follow-up were excluded.

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.

Population: Only eligible and analyzable patients were included in the analysis.

Only the first 6 of 7 groups are shown.

6. Primary Endpoint

The registry lists the primary outcome as Disease-free Survival, measured over every 6 months (annually for mammograms) for 5 years.

The primary endpoint is disease-free survival, recorded as the percentage of participants disease-free at year five. The registry defines disease-free survival as the time from registration, described as randomization assignment, to the first recurrence, new breast primary tumor, or death from any cause. Follow-up for the measure is described as occurring every six months, with mammograms annually, for five years.

This endpoint combines several clinically important events into a time-to-event measure. General statistical context: time-to-event outcomes are useful when the timing of an event matters and when participants may have different lengths of follow-up. The endpoint definition also makes clear that a participant is counted as having an event after any one of the listed events, rather than only after recurrence of the original cancer.

The year-five percentage gives a single-time summary that is easier to read across arms, but it does not by itself describe when events occurred or how follow-up was handled for participants whose event or censoring time differs. The registry’s sponsor analyses use log-rank methods and report hazard ratios for selected comparisons, which draw on event-time information rather than only the year-five percentage.

The registry identifies the primary analysis population as eligible and analyzable participants and notes three exclusions for no follow-up in two arms. Further details about endpoint adjudication, censoring rules, or handling of additional missing data are not reported in the registry record.

7. Results Posted in the Registry

• Arm II: 79.0% (538/681) versus 83.0% (551/664) on control; risk difference -4.0 percentage points (95% CI -8.2 to 0.2), risk ratio 0.95 (0.90–1.00); unadjusted two-proportion z-test p = 0.063.
• Arm III: 81.1% (552/681) versus 83.0% (551/664) on control; risk difference -1.9 percentage points (95% CI -6.0 to 2.2), risk ratio 0.98 (0.93–1.03); unadjusted two-proportion z-test p = 0.358.
• Arm IV: 81.0% (517/638) versus 83.0% (551/664) on control; risk difference -1.9 percentage points (95% CI -6.1 to 2.2), risk ratio 0.98 (0.93–1.03); unadjusted two-proportion z-test p = 0.360.
• Arm V: 84.1% (233/277) versus 83.0% (551/664) on control; risk difference 1.1 percentage points (95% CI -4.0 to 6.3), risk ratio 1.01 (0.95–1.08); unadjusted two-proportion z-test p = 0.671.

8. Secondary Efficacy Endpoints

The registry also posts results for secondary outcomes. For each outcome below, the endpoint type, the test applied to the posted summary statistics, the result and its interpretation are given in turn. Secondary endpoints are usually not adjusted for multiple testing, so they support rather than confirm the primary finding.

Disease-free Survival Comparison Between 2 Treatments in HR-positive, HER-2 Negative Group

• Endpoint type: Binary: the percentage of participants with the event (counts rebuilt from the posted percentages and group sizes).
• Test: Unadjusted two-proportion z-test, with a Wald 95% confidence interval for the risk difference.
• Result:
  • Arm II: 81.9% (303/370) versus 82.1% (308/375) on control; risk difference -0.2 percentage points (95% CI -5.8 to 5.3), risk ratio 1.00 (0.93–1.07); unadjusted two-proportion z-test p = 0.932.
  • Arm III: 86.1% (321/373) versus 82.1% (308/375) on control; risk difference 3.9 percentage points (95% CI -1.3 to 9.2), risk ratio 1.05 (0.98–1.12); unadjusted two-proportion z-test p = 0.142.
• Interpretation:
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between Arm II and the control. The interval gives the range of effects that remain compatible with the data.
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between Arm III and the control. The interval gives the range of effects that remain compatible with the data.

Overall Survival Comparison Between 2 Treatments in HR-positive, HER-2 Negative Group

• Endpoint type: Binary: the percentage of participants with the event (counts rebuilt from the posted percentages and group sizes).
• Test: Unadjusted two-proportion z-test, with a Wald 95% confidence interval for the risk difference.
• Result:
  • Arm II: 91.9% (340/370) versus 90.9% (341/375) on control; risk difference 1.0 percentage points (95% CI -3.1 to 5.0), risk ratio 1.01 (0.97–1.06); unadjusted two-proportion z-test p = 0.641.
  • Arm III: 90.9% (339/373) versus 90.9% (341/375) on control; risk difference -0.0 percentage points (95% CI -4.2 to 4.1), risk ratio 1.00 (0.96–1.05); unadjusted two-proportion z-test p = 0.982.
• Interpretation:
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between Arm II and the control. The interval gives the range of effects that remain compatible with the data.
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between Arm III and the control. The interval gives the range of effects that remain compatible with the data.

Disease-free Survival Comparison Between 2 Treatments in HR-negative, HER-2 Negative Group

• Endpoint type: Binary: the percentage of participants with the event (counts rebuilt from the posted percentages and group sizes).
• Test: Unadjusted two-proportion z-test, with a Wald 95% confidence interval for the risk difference.
• Result:
  • Arm II: 70.8% (131/185) versus 83.0% (127/153) on control; risk difference -12.2 percentage points (95% CI -21.0 to -3.3), risk ratio 0.85 (0.76–0.96); unadjusted two-proportion z-test p = 0.009.
  • Arm III: 72.2% (130/180) versus 83.0% (127/153) on control; risk difference -10.8 percentage points (95% CI -19.6 to -1.9), risk ratio 0.87 (0.78–0.98); unadjusted two-proportion z-test p = 0.019.
• Interpretation:
  • The 95% confidence interval excludes no difference, so the proportion with the event is statistically lower for Arm II than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.
  • The 95% confidence interval excludes no difference, so the proportion with the event is statistically lower for Arm III than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.

The secondary outcomes cover disease-free survival and overall survival in biomarker-defined groups, in addition to an adverse-event outcome. The posted subgroup results show that the estimated patterns vary across outcomes and groups; the registry record does not report results for every secondary outcome listed. In particular, results for the listed HER2-positive subgroup comparisons are not reported in the registry record.

These secondary analyses add context but do not replace the primary disease-free survival result. They address different outcome definitions or restricted populations, and their sample sizes are smaller than those for the overall primary analysis. Subgroup estimates can therefore be less precise. The registry’s reported analyses include log-rank p-values for some groups, while the summary statistics also provide year-five percentages and recalculated contrasts. Those methods and summaries should not be conflated.

Unadjusted secondary results carry less weight because the record lists multiple outcomes and analyses and does not report a multiplicity strategy. General statistical context: when many hypotheses are tested, the chance of at least one apparently unusual result can increase even if no underlying differences exist. Adjustment methods and prespecified testing hierarchies can address this issue, but their use in the trial is not reported in the registry record.

Secondary findings are best viewed as exploratory context unless the analysis plan establishes otherwise. The registry does not report a hierarchy or adjustment that would establish confirmatory status for these subgroup tests. Their interpretation should also account for the fact that the subgroups were defined by biomarkers, that the results apply only to the analyzed subsets, and that missing-data and analysis-population details are limited.

Secondary outcomes registered for this trial, for reference (only those above were recalculated):

• Number of Patients With Gr 3 Through 5 Adverse Events That Are Related to Study Drugs
• Disease-free Survival Comparison Between 2 Treatments in HR-positive, HER-2 Negative Group
• Overall Survival Comparison Between 2 Treatments in HR-positive, HER-2 Negative Group
• Disease-free Survival Comparison Between 2 Treatments in HR-negative, HER-2 Negative Group
• Overall Survival Comparison Between 2 Treatments in HR-negative, HER-2 Negative Group
• Disease-free Survival Comparison Between 2 Treatments in HER2-positive Group
• Overall Survival Comparison Between 2 Treatments in HER-2 Positive Group.

9. Statistical Interpretation

10. Statistical Methods Explained

General statistical context: the endpoint is a time-to-event outcome, but the posted arm estimates express disease-free survival at year five as a percentage. The sponsor’s analyses use the log-rank method and, for selected comparisons, report hazard ratios. The recalculated comparisons use unadjusted differences in the posted year-five proportions. These approaches answer related questions, but their estimates are not interchangeable.

What does a year-five percentage describe?

A year-five disease-free survival percentage is an estimate of the proportion remaining free of the defined events at that time point. The registry lists each arm’s analysis denominator and estimate, along with a 95% confidence interval. The interval gives a range of values compatible with the data under the method used, subject to its assumptions. It is not a guarantee that the true value lies inside the interval, nor does overlap or non-overlap alone substitute for a formal comparison.

General statistical context: time-to-event methods use the timing of events and can account for participants whose observation ends before the endpoint time, according to specified censoring rules. The registry does not report the detailed censoring rules. A single year-five summary can conceal variation in when events occurred, while an event-time analysis can reflect more of the follow-up history.

What does the log-rank result represent?

The sponsor reports log-rank analyses for selected pairs of arms and for groups of arms. The log-rank test compares event-time patterns over follow-up under its assumptions. The reported hazard ratio describes a relative rate of events over time as modeled for the comparison; it is not the same quantity as a difference between year-five percentages. General statistical context: interpreting a hazard ratio as a constant relative contrast requires assumptions about how the relative event rate behaves over time. The registry does not provide additional diagnostics or model details.

The sponsor’s reported analyses include confidence intervals and p-values for selected hazard ratios, as well as p-values for multi-arm tests without an accompanying effect estimate. A p-value measures how compatible the data are with a specified null hypothesis under the analysis assumptions. It does not state the size or clinical importance of a difference. An interval gives information about both the estimated contrast and its uncertainty; a broad interval allows a wider range of possible effects than a narrow one.

How should the recalculations be read?

The recalculated risk differences compare the posted year-five proportions for each arm with Arm I. They are unadjusted and use the posted summary statistics. A negative estimate indicates a lower observed year-five proportion in the named arm than in Arm I; a positive estimate indicates a higher one. The recalculated intervals for these comparisons include zero, so these summaries do not resolve the direction of a difference. The registry flags a discrepancy between the recalculations and sponsor analysis. In that circumstance, the sponsor’s own analysis takes precedence.

General statistical context: a confidence interval for a difference that includes zero indicates that the data are compatible with no difference under that calculation, while also allowing other values within the interval. It should not be interpreted as proof of equivalence. Establishing equivalence or noninferiority requires a prespecified margin and an appropriate design and analysis; such details are not reported in the registry record.

Why can over-dispersion matter?

General statistical context: over-dispersion occurs when observed variability exceeds what a model assumes. If a model understates variability, its standard errors may be too small and its confidence intervals too narrow. The registry does not report an over-dispersion adjustment for these outcomes, and no claim about over-dispersion in this trial can be made from the posted record. This concept is relevant when evaluating statistical methods generally, but it should not be treated as a specific finding about this trial.

The posted recalculations are not adjusted for covariates or for the multiple comparisons represented in the registry. The sponsor analyses use the log-rank method, while the recalculations use the year-five summaries. Since the registry identifies a discrepancy, the recalculations are best regarded as a separate, simple descriptive check. They do not supersede the sponsor’s analysis or establish what the original protocol specified.

Finally, statistical significance and precision are different ideas. A p-value summarizes evidence relative to a null hypothesis; a confidence interval displays a range of plausible estimates under the method. Both depend on the data and assumptions. Neither alone establishes clinical importance, and neither removes the need to consider the endpoint definition, analysis population, missing data, multiplicity, and the distinction between primary and secondary outcomes.

11. Confidence Intervals and Precision

The year-five disease-free survival estimates have 95% confidence intervals that vary in width across arms. Arm I is reported as 83.0, with an interval from 80.0 to 86.0. Arm II is 79.0, with an interval from 76.0 to 82.0. Arm III and Arm IV are each 81.0, with intervals from 78.0 to 84.0. Arm V is 84.0, with an interval from 79.0 to 88.0, and Arm VI is 85.0, with an interval from 80.0 to 89.0. The estimates for Arms V and VI have smaller analysis denominators than the first four arms, and their intervals are wider.

General statistical context: a confidence interval communicates uncertainty around an estimate. Wider intervals indicate less precision, while narrower intervals indicate more precision under the analysis method. Interval width depends on factors including the amount of information available and the variability in the data. The arm-level intervals are not tests of pairwise differences; a direct comparison requires an appropriate contrast and its own uncertainty estimate.

The recalculated risk-difference intervals also indicate uncertainty in the arm comparisons. Each interval includes zero. The registry flags a discrepancy between these recalculations and the sponsor’s analysis, so the sponsor’s reported time-to-event results take precedence. Those reported hazard-ratio intervals also differ in precision: for example, the Arm I versus Arm II interval is 1.04 to 1.68, while the Arm V versus Arm VI interval is 0.61 to 1.41. Their widths reflect uncertainty in those particular event-time comparisons, not certainty about a treatment effect.

Confidence intervals do not capture every source of uncertainty. They do not by themselves address missing-data mechanisms, selective reporting, multiplicity, or whether the analysis assumptions are appropriate. The registry does not report enough detail to evaluate all of those issues.

12. 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 sample-size illustrations are planning calculations based on the observed effect, not the trial’s original design. They show the participants per arm associated with 80% power and 90% power under the stated assumptions. For the time-to-event comparisons, the corresponding planning quantity is total events rather than participants per arm; the displayed disease-free survival comparisons are binary percentage outcomes, so their figures are participants per arm.

The stated planning assumptions are a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation. The effect to detect is set to the stated fraction of the observed effect. The registry’s original sample-size assumptions and justification are not reported in the registry record.

In plain language, the 80% power column gives the illustrative number of participants per arm for a design with an 80% chance of detecting the specified effect under the calculation assumptions, if that effect is truly present. The 90% power column gives the corresponding planning number for a higher chance of detection. These are design illustrations, not observed power. Smaller effects require more participants to distinguish them from random variation, which is why the illustrations increase when planning for 75% or 50% of the observed effect.

These figures should not be read as evidence that the trial did or did not have adequate power. They use observed effects in a retrospective planning exercise, and an observed effect may be an unstable basis for designing a future study. They also do not recover the trial’s original assumptions, which are not reported in the registry record.

13. 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, p₀ and p₁ are the control and treatment proportions, and p̄ is their average.

Worked calculation for this trial

1. Control proportion p₀ = 0.830; treatment proportion p₁ = 0.790; average p̄ = 0.810.
2. First term: z_1−α/2 × √[2p̄(1−p̄)] = 1.96 × 0.5549 = 1.0876.
3. Second term: z_1−β × √[p₀(1−p₀) + p₁(1−p₁)] = 0.842 × 0.5542 = 0.4664.
4. n = (1.0876 + 0.4664)² / (-0.0398)² = 1,524.1, rounded up to 1,525 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
p0 <- 0.829819277108434                          # control proportion
p1 <- 0.790014684287812                          # treatment proportion
n_per_arm <- function(power, frac = 1) {
  q1 <- p0 + (p1 - p0) * frac               # smaller effect = fraction of the observed difference
  pbar <- (p0 + q1) / 2
  z_b <- qnorm(power)
  ceiling(round((z_a * sqrt(2 * pbar * (1 - pbar)) + z_b * sqrt(p0 * (1 - p0) + q1 * (1 - q1)))^2 / (q1 - p0)^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 (unrounded n per arm)
cat("power.prop.test:", round(power.prop.test(p1 = p0, p2 = p1, power = 0.80)$n, 1), "\n")

# Simulation: how often does the z-test reject at the 80%-power sample size?
set.seed(70564)
n    <- n_per_arm(0.80)
reps <- 20000
x0 <- rbinom(reps, n, p0); x1 <- rbinom(reps, n, p1)
ph <- (x0 + x1) / (2 * n)
z  <- ((x1 - x0) / n) / sqrt(ph * (1 - ph) * 2 / n)
cat(sprintf("Simulated power at n = %s per arm: %.3f (%d replicates)\n", format(n, big.mark = ","), mean(abs(z) > z_a, na.rm = TRUE), reps))

Output from running the code

100% of observed effect: n per arm = 1,525 (80% power), 2,040 (90% power)
 75% of observed effect: n per arm = 2,656 (80% power), 3,555 (90% power)
 50% of observed effect: n per arm = 5,852 (80% power), 7,834 (90% power)
power.prop.test: 1524.1 
Simulated power at n = 1,525 per arm: 0.798 (20000 replicates)

The simulation draws many pairs of event counts at the planned sample size and counts how often the two-proportion z-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.

14. Analyses Reported by the Sponsor

The sponsor’s own primary-outcome analyses are reported as log-rank comparisons, including hazard ratios for selected pairs and p-values for multi-arm comparisons. These are time-to-event analyses. The simple recalculations compare year-five percentages using unadjusted risk differences. Because the registry flags a discrepancy between the sponsor analysis and the recalculations, the sponsor’s own analysis takes precedence.

The recalculations can still help readers see how the posted year-five summaries compare on a straightforward proportion scale. They are not adjusted analyses, do not reproduce the sponsor’s time-to-event method, and should not be substituted for the sponsor’s reported results. Their confidence intervals include zero for all posted primary pairwise comparisons, but that observation does not resolve the discrepancy or invalidate the sponsor’s analysis.

The registry does not explain the source of the discrepancy, nor does it report enough methodological detail to reconcile the estimates. It does not report the sponsor’s full statistical plan or all model details. The appropriate reading is therefore to keep the two analyses distinct: use the sponsor’s log-rank findings as the primary reported comparative analysis, and treat the recalculations as supplementary unadjusted summaries of the posted endpoint percentages.

15. Multiplicity, Missing Data and Analysis Populations

The registry lists nine outcome measures and 22 analyses. It does not report a multiplicity adjustment, a prespecified testing hierarchy, or another strategy for controlling the interpretation of multiple comparisons. Consequently, the family of comparisons behind the reported p-values is unclear. General statistical context: examining many outcomes and contrasts can increase the chance of observing a small p-value by chance, so unadjusted p-values should not be treated as if they came from one isolated test.

The primary endpoint analysis included eligible and analyzable participants. The registry specifically says that two participants in Arm II and one in Arm IV with no follow-up were excluded. It does not report the extent or handling of any other missing endpoint information, the reasons for missingness, or whether sensitivity analyses were performed. The detailed analysis-population definitions beyond the stated eligibility and analyzability criteria are not reported in the registry record.

Missing follow-up can matter because disease-free survival depends on the time to an event and the length of observation. General statistical context: time-to-event analyses commonly use censoring for participants whose event status is not observed through the full follow-up period, but the validity of that approach depends on assumptions about how follow-up ends. The registry does not provide the censoring rules or enough detail to assess those assumptions.

The record also does not report interim-analysis procedures, stopping rules, the original sample-size justification, or a complete prespecified analysis plan. These omissions do not establish that such procedures were absent; they mean the information is not reported in the registry record. Interpretation should therefore distinguish what is posted from details that cannot be assessed.

16. 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 5% reporting threshold.

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: neutrophils/granulocytes (anc/agc) (1.1%), febrile neutropenia (0.7%), secondary malignancy-poss rel to cancer tx (0.6%), left ventricular systolic dysfunction (0.4%), pericardial effusion (non-malignant) (0.3%).

The registry reports serious adverse events for each arm as counts among participants at risk, alongside counts for other adverse events. For serious adverse events, the reported figures are 10 among 654 in Arm I, 8 among 663 in Arm II, 17 among 674 in Arm III, 9 among 624 in Arm IV, 5 among 275 in Arm V, and 8 among 291 in Arm VI. The record also lists the most frequently reported serious-event terms and their percentages, including neutrophils/granulocytes (ANC/AGC) at 1.1% and febrile neutropenia at 0.7%.

These figures describe recorded safety events, not a causal comparison between regimens. The registry lists a secondary outcome concerning grade 3 through 5 adverse events related to study drugs, but the serious-event counts should not be assumed to be identical to that outcome. The deaths field has no value for the arms, and the registry does not report a complete narrative explaining event attribution or follow-up completeness.

Counts should be read alongside the number at risk and the event definitions. The registry’s listed serious-event terms include events described as related to cancer treatment, but the record does not provide enough detail to infer comparative safety or explain individual event circumstances.

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

The posted results describe five-year disease-free survival estimates across six randomized arms and sponsor-reported time-to-event comparisons for selected arms. They show variation in the observed estimates and differing degrees of precision. The sponsor reports a hazard ratio of 1.32 for the Arm I and Arm II comparison, with a 95% confidence interval from 1.04 to 1.68 and p = 0.022; other reported comparisons have intervals that include 1. The multi-arm log-rank comparison across the first four arms has p = 0.11.

These statistics do not, by themselves, establish that one schedule is clinically preferable. A hazard ratio and a p-value summarize particular statistical comparisons; they do not express the full clinical balance of benefits and harms. The registry record does not provide a complete analysis plan, multiplicity strategy, or full details needed to evaluate all assumptions. The recalculated year-five comparisons differ in method from the sponsor analysis, and the registry flags a discrepancy, so the sponsor’s analysis takes precedence.

The record also does not establish equivalence among arms. A nonsignificant result is not proof that outcomes are the same, and an interval that includes a null value can still allow effects in either direction. Equivalence or noninferiority would require a prespecified margin and an analysis designed for that purpose; those details are not reported in the registry record.

The subgroup results are secondary and involve smaller groups and multiple comparisons. Their findings should not be treated as replacing the primary result. Likewise, safety counts are descriptive and do not by themselves establish comparative risk. The statistical record can describe estimates, uncertainty, and reported tests, but it does not support medical advice or a treatment recommendation.

18. Why This Trial Matters Statistically

This trial is statistically informative because it compares multiple randomized chemotherapy schedules and reports both a fixed-time disease-free survival summary and selected time-to-event analyses. That combination illustrates why different statistical summaries can appear to tell different parts of the story. A year-five percentage is a direct snapshot of the endpoint at a specified time, while a log-rank test and hazard ratio use event-time information across follow-up.

The posted arm denominators and confidence intervals also make precision visible. The analyses for Arms V and VI have smaller denominators and wider intervals than several of the earlier arms, so their estimates are less precise. This is a reminder that the size of an observed estimate and the certainty around it are separate considerations.

The discrepancy flag is another important statistical lesson: a simple recalculation from summary percentages may not reproduce a sponsor’s time-to-event analysis. The recalculation can be useful for understanding the posted proportions, but it cannot substitute for the sponsor’s analysis when the registry identifies a discrepancy. Differences in endpoint summaries and methods need to be kept explicit rather than blended into one conclusion.

Finally, the record shows why multiplicity and reporting completeness matter. Several outcomes and analyses are listed, while the adjustment strategy and full analysis plan are not reported. A statistically careful reading therefore considers the primary endpoint, the analysis method, uncertainty, secondary comparisons, and missing information together, rather than relying on one isolated estimate or p-value.

19. 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.
• Participant counts were reconstructed from posted percentages and arm sizes, so small rounding differences are possible.

Several limitations are specific to the information available in this registry record. The arm descriptions are truncated, so the complete treatment schedules cannot be reconstructed from the posted text. The record states that Arms I through IV were closed 11/10/10, but does not provide an explanation for that closure or a full account of how the later arms relate to the earlier arms.

The primary analysis population is described as eligible and analyzable, with two participants in Arm II and one in Arm IV excluded for no follow-up. More detailed information about missingness, censoring, endpoint assessment, and analysis populations is not reported in the registry record. These details limit assessment of how participants with incomplete follow-up contributed to the year-five and time-to-event results.

The sponsor reports selected log-rank comparisons, but the complete statistical plan, model details, and multiplicity approach are not reported. The registry lists nine outcomes and 22 analyses, making the absence of a reported strategy for multiple comparisons relevant. The record also does not explain the flagged discrepancy between sponsor analyses and simple recalculations.

Some secondary subgroup outcomes are listed without posted results, including the HER2-positive comparisons. The safety information is summarized as counts and selected event terms, without a full account of attribution or the deaths field. These are limits of the posted record and do not establish what analyses or procedures may have occurred outside it.

20. Related Tutorials

21. Related Statistical Calculators

22. Sources

• ClinicalTrials.gov: S0221 Adjuvant Doxorubicin, Cyclophosphamide, and Paclitaxel in Treating Patients With Breast Cancer — NCT00070564.
• Registry record last updated 2026-08-28; status last verified 2026-08.
• Results first posted 2017-04-17.
• 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.

23. Record Summary

NCT00070564 is a randomized, open-label, parallel-group phase III trial of adjuvant chemotherapy schedules for breast cancer. The registry lists enrollment of 3294 and identifies five-year disease-free survival as the primary outcome. The endpoint includes recurrence, a new breast primary tumor, or death from any cause.

Six arms have posted year-five estimates. Sponsor analyses use log-rank comparisons and report selected hazard ratios. Simple unadjusted recalculations from the year-five summaries are also available, but the registry flags a discrepancy; the sponsor’s own analysis takes precedence. The record lists secondary outcomes and adverse events, while the full original statistical plan, multiplicity strategy, and several analysis details are not reported in the registry record.