A Trial of Standard Chemotherapy With Metformin (vs Placebo) in Women With Metastatic Breast Cancer: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for A Trial of Standard Chemotherapy With Metformin (vs Placebo) in Women With Metastatic Breast Cancer (NCT01310231), a randomized phase 2 study. The primary outcome is Progression Free Survival..

2. Clinical Question

The trial asks whether the addition of metformin to chemotherapy is associated with a difference in the time participants remain alive without documented disease progression. It is a comparison of two treatment strategies, not a comparison of metformin with no cancer treatment: both groups were assigned standard chemotherapy, with metformin or placebo added.

The posted primary analysis reports a hazard ratio of 1.2, with a 95% confidence interval from 0.63 to 2.31 and p = 0.71. The interval includes values on both sides of 1, and the reported p-value does not indicate clear evidence of a difference under the sponsor’s analysis. The registry’s arm summaries also allow a separate, unadjusted comparison of reported mean progression-free survival. That recalculation gives a mean difference of -0.9 months and p = 0.047. Because the recalculation conflicts with the sponsor’s reported analysis, the sponsor’s analysis takes precedence for interpreting the trial result.

The registry record does not report the original sample-size justification or enough detail to infer how the trial was designed to detect an effect. The results should therefore be read as evidence from a small phase II trial, with attention to the uncertainty in the primary analysis and to the difference between the sponsor’s time-to-event model and the unadjusted summary-statistic recalculation.

3. Trial Design

This was a randomized phase II trial with a parallel-group design. The registry reports quadruple masking, meaning that the record identifies the study as masked at four levels but does not name those roles. The planned enrollment was 40 participants. The posted primary outcome data include 22 participants in the metformin group and 18 in the placebo group.

The study compared metformin plus standard chemotherapy with placebo plus standard chemotherapy. Chemotherapy options listed in both arm descriptions include anthracyclines, platinum, taxanes or capecitabine. The official title describes women receiving first to fourth line chemotherapy; the individual arm descriptions refer to first or second line. The registry record does not clarify this difference, so the eligibility or treatment-line description should not be silently narrowed to one account.

The study is marked completed. The registry gives a start date of 2011-08-22 and a primary completion date of 2016-07-14. These record dates describe the study timeline; they do not establish how long each participant was followed. The primary outcome timeframe is from randomization to documented progression or death, assessed up to 3 years. Scans were planned every 9 weeks, and local survival follow-up was to continue until all participants had died or for a maximum total follow-up of 3 years, whichever occurred first.

The registry says the two arms would be compared in an intent-to-treat fashion using Cox proportional hazards analysis, with stratification variables included in the model. The stratification variables are not reported in the registry record. Treatment discontinuation for toxicity or other reasons was to be considered an event according to the primary outcome description. The posted results identify a Cox regression analysis, but the registry record does not report further model diagnostics or details of how deviations from the stated analysis plan were handled.

Secondary outcomes include overall response rate, adverse events, quality of life, fasting glucose, fasting insulin, insulin resistance and exploratory immunohistochemical measures. The registry lists 9 secondary outcome results as posted. The detailed analysis and missing-data information for those outcomes is not reported 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

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 Progression Free Survival., measured over From date of randomization to first documented progression or death, which ever occurs first, assessed up to 3 years..

Progression-free survival is the time from randomization to the first documented disease progression or death, whichever occurs first. In this trial, the registry describes assessment up to 3 years. Scans were planned every 9 weeks, with local survival follow-up continuing until all participants had died or the maximum total follow-up of 3 years was reached, whichever occurred first.

This endpoint combines progression and death into a time-to-event outcome. It can capture how long participants remain alive without a documented progression, rather than only whether progression was present at a single visit. That makes it relevant to a comparison of treatment strategies when the timing of progression or death matters. A Cox proportional hazards model is commonly used for this type of outcome because it compares event rates over follow-up while accounting for the timing of events and censoring. This is a general explanation of the method, not a claim about unreported details of this trial’s data.

The registry’s primary outcome record also labels the measure as a mean with standard deviation and gives arm means in months. The posted primary analysis, however, is a Cox regression with a hazard ratio. These summaries answer different statistical questions: a difference between arithmetic means is not the same estimand as a hazard ratio from a time-to-event model. The registry record does not explain the relationship between these reported summaries. Interpretation should therefore distinguish the sponsor’s Cox analysis from the separate mean-based recalculation.

The endpoint definition specifies treatment discontinuation for toxicity or other reasons as an event. It does not report further details about event adjudication, the handling of missed scans, or censoring rules beyond the follow-up description. Those details are not reported in the registry record.

7. Results Posted in the Registry

• Metformin: mean 5.40 versus 6.30 on control; difference -0.90 (95% CI -1.79 to -0.01), standardized effect d = -0.66; z-test p = 0.047.

8. Statistical Interpretation

9. Statistical Methods Explained

The primary analysis reported in the registry is a Cox regression. This is a time-to-event method: it uses information about when progression or death occurred, as well as information about participants whose event was not observed during follow-up. In general, Cox regression estimates a hazard ratio comparing event rates between groups over time. A hazard ratio is not a ratio of mean survival times, and it should not be interpreted as a direct difference in months.

What does the hazard ratio describe?

In general, a hazard ratio compares the instantaneous event rates between groups among participants who have not yet had the event. Its interpretation depends on which group is the reference and how the event is coded. The registry reports a hazard ratio of 1.2, but the posted results do not specify the coding and reference details needed to give the direction a more detailed verbal interpretation. The 95% confidence interval is 0.63 to 2.31, and the p-value is 0.71. The interval includes 1, so the estimate is compatible with no difference as well as with effects on either side of that value.

Why use a time-to-event model?

In general, participants may have different lengths of observation, and some may not have experienced progression or death by the end of their observed follow-up. Time-to-event methods can use partial follow-up rather than treating every participant as though their event status were observed for the same duration. A Cox model does this while estimating a relative event-rate measure. The registry states that stratification variables were included in the model, but it does not identify those variables or report model diagnostics. The proportional hazards assumption is a general feature of the Cox model; whether and how it was assessed in this trial is not reported in the registry record.

How does the mean-based recalculation differ?

The posted arm summaries report mean progression-free survival of 5.4 months for metformin and 6.3 months for placebo. A recalculation from those summaries gives an unadjusted mean difference of -0.9 months, with p = 0.047. This comparison uses the reported means and their standard deviations; it does not reproduce the Cox regression or account for the stated stratification variables. It is therefore a distinct analysis, not a second presentation of the same estimand.

A p-value measures how unusual data at least as incompatible with a specified null model would be if that model were true, under the assumptions of the analysis. It is not the probability that the null hypothesis is true, and it does not measure the size or clinical importance of an effect. The sponsor’s p-value of 0.71 and the recalculated p-value of 0.047 arise from different analyses and should not be compared as if they came from one common model. Because the registry identifies a discrepancy, the sponsor’s own analysis takes precedence.

What does a confidence interval contribute?

In general, a confidence interval describes the range of values reasonably compatible with the data under a particular model and its assumptions. A wider interval signals less precision than a narrower interval. The sponsor’s interval runs from 0.63 to 2.31 around the reported hazard ratio of 1.2. It includes the no-difference value of 1 and also allows values on either side, so the result is not precise enough to establish a clear direction or magnitude of association.

The recalculated mean comparison also has a confidence interval in the statistical output, but the record identifies the discrepancy with the sponsor analysis. The reported upper bound has more than three decimal places, so it is not reproduced. The recalculation remains an unadjusted check on the posted summaries and does not supersede the registry’s Cox analysis.

What does over-dispersion mean in general?

Over-dispersion occurs when observed variability is greater than a statistical model assumes. In general, failing to account for over-dispersion can make standard errors too small and intervals too narrow. The primary endpoint is time-to-event, and the registry reports a Cox analysis; it does not report an over-dispersion assessment or a model that would establish whether this issue arose. This general concept should not be read as a claim about the trial’s analysis.

The registry’s primary outcome definition includes treatment discontinuation for toxicity or other reasons as an event, while the outcome itself is defined as time to documented progression or death. The record does not explain how discontinuation was incorporated into the event data or how participants without an observed event were censored. Those specifics are not reported in the registry record and should not be inferred from the general description of Cox regression.

10. Confidence Intervals and Precision

The sponsor’s hazard ratio estimate is 1.2, with a 95% confidence interval from 0.63 to 2.31. The interval includes 1 and extends to values on either side of it. This indicates that the estimate is not precise enough to distinguish clearly between a lower hazard, no difference and a higher hazard under the reported model. The p-value of 0.71 is consistent with the interval’s failure to exclude 1.

A confidence interval is a range generated by a statistical procedure that, under its assumptions, would cover the target parameter at the stated rate across repeated applications. It is not a range containing a known proportion of individual participants’ outcomes, nor is it the probability that a particular treatment effect lies inside this one interval. The meaning depends on the model and the parameter being estimated. Here, the sponsor’s interval pertains to the hazard ratio, not a difference in mean months.

The unadjusted recalculation from the arm summaries gives a mean difference of -0.9 months and p = 0.047. Its uncertainty should not be substituted for the uncertainty in the Cox estimate: the two analyses summarize the outcome differently. The registry marks a discrepancy, so the sponsor’s analysis takes precedence. The registry record does not provide an explanation for the difference or enough information to reconstruct the full Cox analysis independently.

The arm summaries are based on 22 participants receiving metformin and 18 receiving placebo. The record does not report the total number of progression or death events. Since event information is central to a time-to-event estimate, that omission limits assessment of how much information supports the reported hazard ratio. The confidence interval is the clearest posted measure of precision for the sponsor’s primary analysis, and its breadth warrants caution in describing the size or direction of the estimated effect.

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 display gives illustrative participants-per-arm figures for detecting specified fractions of the observed mean difference. Because the displayed endpoint is continuous and the planning assumptions specify equal allocation, the entries are participants per arm, not total event counts. The assumptions are a two-sided significance level of 0.05, equal allocation between the two arms and a normal approximation. The target effect is the stated fraction of the observed effect.

For planning at 100% of the observed effect, the display gives 37 participants per arm for 80% power and 49 participants per arm for 90% power. Planning for 75% of the observed effect gives 65 participants per arm for 80% power and 86 participants per arm for 90% power. Planning for 50% of the observed effect gives 145 participants per arm for 80% power and 194 participants per arm for 90% power.

In plain language, the 80% power column describes the illustrative group size associated with an 80% chance of detecting the assumed effect under the stated planning assumptions, if that effect and the assumptions were correct. The 90% power column uses a higher target chance of detection and therefore gives larger group sizes. Smaller effects are harder to distinguish from random variation, so planning to detect a smaller fraction of the observed effect requires more participants.

These figures are planning illustrations based on the observed effect, not the original design calculation. The trial’s own design assumptions and sample-size justification are not reported in the registry record. This is not observed power.

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 = -0.900; pooled standard deviation σ = 1.364.
2. n = 2 × (1.96 + 0.842)² × 1.364² / -0.900² = 36.1, rounded up to 37 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 <- -0.899999999999999                      # observed difference in means
sigma <- 1.36395709146894                     # 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(1310231)
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 = 37 (80% power), 49 (90% power)
 75% of observed effect: n per arm = 65 (80% power), 86 (90% power)
 50% of observed effect: n per arm = 145 (80% power), 194 (90% power)
power.t.test: 37 
Simulated power at n = 37 per arm: 0.805 (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’s sponsor-reported primary result is the Cox regression estimate: hazard ratio 1.2, 95% confidence interval 0.63 to 2.31, and p = 0.71. The separate recalculation based on the posted arm means gives a mean difference of -0.9 months and p = 0.047. The recalculation is unadjusted and compares means; the sponsor’s result is a time-to-event model. They are not interchangeable tests of the same quantity.

The registry flags a discrepancy between the sponsor analysis and the recalculation. The sponsor’s own analysis therefore takes precedence when describing the primary result. The recalculated finding should not be used to override the reported Cox analysis or to select the more favorable p-value. It is a check based on summary statistics, and it cannot reproduce the stated model’s stratification or use the underlying event and censoring information.

The record does not explain why the results differ. It also does not provide enough details to reproduce the Cox model independently, including the stratification variables and full event data. The appropriate reading is to report the sponsor’s result with its confidence interval and p-value, while transparently noting that a simple mean-based recalculation yields a different result.

14. Multiplicity, Missing Data and Analysis Populations

The registry lists several secondary outcomes, including overall response rate; grade 1 or 2 and grade 3 or 4 adverse events; EORTC quality-of-life measures; fasting glucose; fasting insulin; insulin resistance measured using HOMA-IR; and immunohistochemical predictors and changes in those variables among participants undergoing serial biopsies. The registry reports 9 secondary outcome results as posted. The number of outcome measures in the record is 10, and the number of reported analyses is 2.

When a study examines multiple outcomes or performs multiple comparisons, the chance of obtaining at least one apparently unusual result by chance can increase. This is a general statistical issue known as multiplicity. Researchers may use a prespecified hierarchy, adjustment method or other strategy to control how multiple findings are interpreted. The registry record does not report a multiplicity adjustment or a hierarchy for these outcomes. The secondary findings should therefore be read as separate results rather than as a collectively controlled set of confirmatory tests.

The primary outcome description states that the arms would be compared in an intent-to-treat fashion. In general, intent-to-treat analysis aims to retain participants in their randomized groups, regardless of treatment changes after randomization. The record does not detail how missing outcome information, withdrawals, or deviations from assigned treatment were handled in the posted analysis. The stated approach to treatment discontinuation says it would be considered an event, but it does not replace a full description of missing-data handling.

The registry also states that stratification variables were to be included in the Cox model, but does not name them. It does not report whether the proportional hazards assumption was assessed, whether any sensitivity analyses were performed, or how missing secondary outcome data were treated. These details are not reported in the registry record, so no particular method should be assumed.

These gaps matter for interpretation because outcome selection, repeated testing, missingness and analysis-population choices can affect estimates and their uncertainty. The posted information supports describing the reported primary Cox result and the listed secondary outcomes, but it does not support reconstructing an unreported statistical plan or judging whether all analyses followed one.

15. 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 5% reporting threshold.

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: febrile neutropenia with urinary tract infection (5.9%), hypoxia (5.9%), ascites with hyponatraemia (5.9%), thromboembolism (5.9%), febrile neutropenia with respiratory infection (4.5%).

The registry reports serious adverse events for 3 of 22 participants at risk in the metformin group and 4 of 17 participants at risk in the placebo group. It also reports deaths of 19 participants in the metformin group and 15 in the placebo group. These are counts as posted, and the denominators for serious adverse events differ from the primary outcome group sizes.

The serious-event terms with the largest posted percentages include febrile neutropenia with urinary tract infection, hypoxia, ascites with hyponatraemia and thromboembolism, each listed at 5.9%. Febrile neutropenia with respiratory infection and urosepsis are each listed at 4.5%. These terms and percentages describe the registry’s reported event listings; they do not by themselves establish comparative risk or causation.

Adverse-event figures need context about definitions, observation time, event counting and exposure. The registry record does not provide enough detail to determine whether counts refer to participants with at least one event or to all occurrences for every listed term, or to compare exposure-adjusted rates. The group counts alone should not be treated as proof of a treatment effect on safety. The record’s serious adverse event figures are descriptive and should be interpreted alongside the incomplete reporting detail.

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

The main statistical result in the registry is the sponsor’s Cox estimate for progression-free survival: hazard ratio 1.2, 95% confidence interval 0.63 to 2.31, p = 0.71. The interval includes 1 and is broad, so the result does not clearly establish a difference between the assigned strategies. It also does not show that the strategies have the same effect. The data and interval leave substantial uncertainty about the direction and magnitude of the hazard ratio.

The posted arm means are 5.4 months for metformin plus chemotherapy and 6.3 months for placebo plus chemotherapy. A simple unadjusted recalculation gives a mean difference of -0.9 months with p = 0.047. This result does not match the sponsor’s Cox analysis, and the registry flags the discrepancy. The sponsor’s own analysis takes precedence; the recalculation should not be interpreted as a substitute primary result.

The trial was a phase II study with 40 planned participants and posted primary outcome groups of 22 and 18. The registry does not report the original sample-size justification or the number of progression or death events. The results should therefore be read with attention to limited group sizes and the precision of the reported estimate, rather than as a definitive answer to the broader treatment question.

The endpoint concerns progression-free survival, not overall survival alone. Although deaths are part of the primary endpoint definition, the reported hazard ratio should not be recast as an effect on overall survival. The registry also lists response, adverse events, quality of life and metabolic measures, but does not report a multiplicity strategy. Those secondary results require separate, cautious interpretation.

Overall, the posted evidence describes an uncertain primary time-to-event estimate and a conflicting mean-based recalculation. It does not justify selecting one result based on its p-value, inferring equivalence, or asserting a clinical benefit or lack of benefit. The record supports a careful account of the analysis and its limitations, not medical advice.

17. Why This Trial Matters Statistically

This trial illustrates why the outcome definition and statistical estimand matter. Progression-free survival is a time-to-event endpoint, and the registry describes a Cox analysis that uses event timing and follow-up information. The arm-level means offer a different summary. A simple test of those means does not answer precisely the same statistical question as a hazard ratio, even when both are based on progression-free survival.

The discrepancy is consequential: the sponsor’s Cox analysis reports p = 0.71, while the unadjusted recalculation from the posted means reports p = 0.047. Those values should not be treated as competing votes on one identical analysis. Recognizing the difference in methods prevents overemphasis on the result that appears more statistically decisive. The registry flags the discrepancy, so the sponsor’s analysis takes precedence.

The confidence interval for the hazard ratio ranges from 0.63 to 2.31. Its breadth conveys uncertainty that a single point estimate cannot. It is possible to describe the estimated association while also making clear that the data do not pin down its direction or magnitude with precision. This is a useful distinction between an estimate and the certainty of that estimate.

The sample-size illustrations also demonstrate how planning depends on the effect one aims to detect. Smaller target effects require larger groups under the stated assumptions. Those illustrations are based on the observed effect and are not a record of the original trial design or evidence of the trial’s achieved power.

Finally, the registry lists numerous secondary outcomes and provides limited detail about multiplicity, missing data and model diagnostics. Statistical interpretation depends not only on a headline p-value, but also on what was measured, how it was analyzed, and what uncertainty and reporting limitations remain.

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 posted record leaves important limits on interpretation. The primary analysis is a Cox regression, but the number of progression or death events is not reported. The registry does not identify the model’s stratification variables, provide diagnostics, or describe how the proportional hazards assumption was assessed. It also does not explain the discrepancy between the sponsor’s hazard-ratio result and the unadjusted mean-based recalculation.

The primary outcome description combines progression and death, and says treatment discontinuation for toxicity or other reasons would be considered an event. It does not explain how discontinuation was represented in the analysis alongside the stated progression-or-death endpoint. Details about event adjudication, censoring, missed scans and the handling of participants without observed events are not reported in the registry record.

The trial’s group sizes for the primary summaries are 22 and 18. The registry record does not report the original sample-size justification, the assumptions used for the design, or an interim analysis plan. Planning illustrations based on the observed effect cannot fill in those missing design details.

The participant description is not fully consistent across the record: the official title refers to first to fourth line chemotherapy, while the arm descriptions refer to first or second line. The record does not resolve the difference. The safety summaries also lack details about event-counting rules and exposure time, limiting interpretation of group comparisons.

Finally, secondary outcomes are numerous, but a multiplicity strategy and detailed missing-data methods are not reported. These limitations do not erase the posted results; they define what can be concluded from the registry record without making unsupported assumptions.

19. Related Tutorials

20. Related Statistical Calculators

21. Sources

• ClinicalTrials.gov: A Trial of Standard Chemotherapy With Metformin (vs Placebo) in Women With Metastatic Breast Cancer — NCT01310231.
• Registry record last updated 2021-04-14; status last verified 2021-03.
• Results first posted 2021-04-14.
• 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

NCT01310231 is a completed randomized phase II parallel-group trial sponsored by Ozmosis Research Inc. It compared metformin plus standard chemotherapy with placebo plus standard chemotherapy in women with metastatic breast cancer. The record reports quadruple masking and planned enrollment of 40 participants.

The primary outcome was progression-free survival from randomization to documented progression or death, assessed up to 3 years. The sponsor-reported Cox analysis gave a hazard ratio of 1.2, with a 95% confidence interval from 0.63 to 2.31 and p = 0.71. An unadjusted recalculation from the posted means gives a mean difference of -0.9 months and p = 0.047; the registry flags this discrepancy, and the sponsor’s analysis takes precedence.

The record lists secondary outcomes covering response, adverse events, quality of life, metabolic measures and exploratory biopsy markers. Important details about the original sample-size justification, multiplicity, missing data and model diagnostics are not reported in the registry record.