Assess the Efficacy of AZD8931 in Combination With Paclitaxel Versus Paclitaxel Alone in Patients With Gastric Cancer: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Assess the Efficacy of AZD8931 in Combination With Paclitaxel Versus Paclitaxel Alone in Patients With Gastric Cancer (NCT01579578), a randomized phase 2 study. The primary outcome is Change in Tumour Size at 8 Weeks Were Analyzed for Comparing Relative Efficacy of AZD8931 Plus Paclitaxel With Paclitaxel Alone.

2. Clinical Question

This is a question about a continuous tumour-size measure at a specified follow-up, rather than a question about survival or the proportion of participants meeting a response definition. The registry also lists progression-free survival and objective response rate as secondary outcomes. Those secondary outcomes do not replace the primary endpoint and are not the focus of the principal comparison.

3. Trial Design

NCT01579578 was a phase IIa, randomized, triple-masked, parallel-group study sponsored by AstraZeneca. Its official title describes a multicentre, double-blind, placebo-controlled comparison of AZD8931 plus paclitaxel with paclitaxel alone. Placebo was part of the control regimen, listed as placebo plus paclitaxel in the posted results. The masking designation is recorded as triple; which specific roles were masked is not reported in the registry record.

The intended population had metastatic gastric or gastro-oesophageal junction cancer, progression following first-line therapy, and ineligibility for trastuzumab based on HER2 status. The record lists enrollment of 39. The primary endpoint results include 11 participants in each treatment group, while the safety results list 13 participants at risk in the AZD8931 group and 12 in the placebo group. The record does not provide a participant flow that explains the different counts or the relationship between enrollment and each analysis population.

The study began in 2012-04, had primary completion in 2013-03, and is listed as terminated. The reason for termination is not reported in the registry record. The primary outcome was change in tumour size at 8 weeks, assessed from baseline and at 8 weeks, with data accessed up to the cut-off on 4 December 2012. Progression-free survival and objective response rate are listed as secondary outcomes. Details about randomization procedures, allocation concealment, the planned recruitment target, and the original sample-size justification are not reported in the registry record.

The recorded primary analysis used the Full Analysis Set, but the record does not explain how membership in that set was determined. The parallel-group design frames the estimate as a comparison between assigned regimens. The displayed outcome counts should therefore be read as the counts associated with the posted endpoint results, not as a complete account of enrollment, treatment receipt, or follow-up.

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: Due to the early stopping of this study as a result of the AZD8931 program being stopped, the full planned enrolment was not carried out. This study was stopped after 39 were enrolled, 25 of whom were treated, which is less than the number planned in the protocol.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: Full Analysis Set

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 Tumour Size at 8 Weeks Were Analyzed for Comparing Relative Efficacy of AZD8931 Plus Paclitaxel With Paclitaxel Alone, measured over Baseline and 8 weeks, accessed up to data cut off on 4 December 2012.

The primary endpoint is change in tumour size at 8 weeks, expressed as a percentage change from baseline. It is a continuous measure: each participant contributes a numerical value, and the group results are reported as least-squares means with standard errors. A negative percentage change represents a decrease in measured tumour size from baseline; a positive value would represent an increase. The registry record does not give further details about how tumour size was measured or how multiple lesions were combined.

Using a continuous measure can retain information about the amount of change rather than reducing the outcome to a yes-or-no response category. It can therefore support an estimate of the average difference between groups. The registry lists progression-free survival and objective response rate as secondary outcomes, which address other aspects of efficacy. The primary endpoint is specifically an early tumour-size comparison at 8 weeks, not a direct measure of survival or a summary of longer-term disease course.

The outcome timeframe is baseline and 8 weeks, with data accessed up to the stated cut-off on 4 December 2012. The Full Analysis Set is named as the analysis population, but its definition, endpoint measurement procedures, and handling of participants without an 8-week assessment are not reported in the registry record. Those details matter for understanding exactly which observations contribute to the estimated group means.

7. Results Posted in the Registry

• AZD8931 + Paclitaxel: mean -7.70 versus -2.70 on control; difference -5.00 (95% CI -28.89 to 18.89), standardized effect d = -0.17; z-test p = 0.682.

8. Statistical Interpretation

9. Statistical Methods Explained

The posted sponsor analysis compares the treatment groups using ANCOVA, with the result described as a mean difference in final values. ANCOVA is a general statistical method that models a continuous outcome while accounting for other variables included in the model. Using baseline measurement as a covariate is a common general approach in trials with baseline and follow-up measurements, but the specific covariates in this study's model are not reported in the registry record. It would therefore be inappropriate to assume which adjustments were made.

What does a least-squares mean represent?

A least-squares mean is a model-based group mean. In general, it can account for covariates in the fitted analysis, so it may differ from a simple arithmetic average. The registry labels the primary group values as least-squares means and reports the comparison as an ANCOVA of final values. The specific model form, covariates, and any other adjustments are not reported in the registry record. The values -7.7 and -2.7 are the posted group estimates; the sponsor's comparison is -5.00.

What is the role of the standard error?

A standard error describes the estimated sampling variability of a statistic, such as a group mean. In general, a smaller standard error indicates greater precision for that estimate, while a larger standard error indicates more uncertainty. The posted group standard errors are 8.6 and 8.64, alongside endpoint counts of 11 in each group. Standard errors are not standard deviations: the standard deviation describes variation among individual observations, while the standard error describes uncertainty in an estimated quantity. The registry reports standard error as the dispersion measure for the primary group results.

Why distinguish the sponsor analysis from a recalculation?

The recalculation is an unadjusted comparison reconstructed from posted summary statistics. It is useful as a transparent check of the broad scale and direction of the result, but it cannot reproduce an adjusted model when only group summaries are available. The sponsor's ANCOVA result is the reported trial analysis; its estimate is -5.00, with a 95% confidence interval from -30.81 to 20.81 and p = 0.688. The recalculation gives -5.0, with a confidence interval from -28.89 to 18.89 and p = 0.682. The small differences between these results are consistent with the fact that the calculations use different methods; they should not be combined as if they came from one model.

What does a confidence interval communicate?

In general, a confidence interval expresses the uncertainty around an estimated effect under the assumptions of the analysis. A wider interval indicates less precision than a narrower interval on the same scale. The sponsor's 95% interval includes values on both sides of zero, so the posted data do not precisely locate the average difference. The interval should be read as a range of effect estimates compatible with the analysis, rather than as a range containing a specified proportion of individual participants' outcomes or as a probability statement about the true effect.

Why can variability matter?

In general, when observations vary substantially within groups, it can be harder to estimate a difference between group means precisely. For count outcomes, over-dispersion means that variation exceeds what a simple model assumes; it can make standard errors too small if ignored. The primary endpoint is continuous, and the registry does not report an over-dispersion analysis. That concept should not be treated as a finding from this trial. For this endpoint, the posted standard errors and wide confidence interval are the relevant visible indicators of uncertainty.

What can the posted p-value say?

A p-value evaluates how compatible the observed data are with a specified null model, under the assumptions of the test. It is not the probability that the null hypothesis is true, nor is it a measure of effect size or clinical importance. The sponsor's p-value is 0.688. The registry does not report the full test specification, including whether any multiplicity adjustment was applied. The result should therefore be interpreted alongside the estimated difference and its confidence interval rather than in isolation.

The outcome is measured at a fixed follow-up time, so this primary comparison is not a time-to-event analysis. The secondary outcomes include progression-free survival and objective response rate, but detailed estimates for those outcomes are not included in the primary result. The registry lists one primary analysis and two secondary outcomes; whether all planned analyses or all outcome results were posted is not reported in the registry record.

10. Confidence Intervals and Precision

The sponsor's primary estimate is a mean difference of -5.00 percentage points, with a 95% confidence interval from -30.81 to 20.81. The interval is wide and crosses zero. In practical statistical terms, the point estimate is not pinned down closely: the data are compatible with negative and positive average differences under the reported analysis. This uncertainty is important when interpreting the direction of the estimate. The point estimate alone would suggest a larger average decrease in the AZD8931 group, but the interval includes estimates in the opposite direction as well.

A confidence interval is a measure of uncertainty in an estimated effect under the analysis assumptions. It is not a range of individual participant outcomes, and it does not mean there is a 95% probability that the true value is inside this particular interval. The interval's breadth reflects limited precision in the posted estimate. The endpoint analysis includes 11 participants in each group, and the group standard errors are 8.6 and 8.64.

The unadjusted recalculation has a 95% confidence interval from -28.89 to 18.89, with an estimate of -5.0. This interval is also wide and includes zero. The sponsor's interval should guide interpretation of the reported trial analysis because it comes from the posted ANCOVA, whereas the recalculation is based on summary statistics and does not reproduce that model. The registry does not report a clinically meaningful difference threshold, so the interval cannot be divided into clinically important and unimportant values using the posted record alone.

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 illustrations based on the observed effect, not the trial's original design calculations. The assumptions listed for these figures are a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation; the effect to detect is the stated fraction of the observed effect. The figures are participants per arm, not total participants and not total events, because this is a continuous outcome rather than a time-to-event planning calculation.

The 80% power column gives the number of participants per arm estimated to provide an 80% chance of detecting the specified effect under those planning assumptions, if that effect were the true effect. The 90% power column uses a higher target chance of detecting it. For each effect scenario, the figures are planning requirements rather than observed trial counts. The entries include 514 and 688 participants per arm for planning for 100% of the observed effect; 913 and 1,222 for 75% of the observed effect; and 2,053 and 2,749 for 50% of the observed effect, for the 80% and 90% targets respectively.

Smaller effects require more participants to distinguish them from random variation at a given significance level and power target. Thus the planning figures rise as the assumed fraction of the observed effect falls. These figures are not observed power, and they do not establish what sample size the trial originally planned. The trial's own sample-size justification and original design assumptions are not reported in the registry record.

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 = -5.000; pooled standard deviation σ = 28.589.
2. n = 2 × (1.96 + 0.842)² × 28.589² / -5.000² = 513.2, rounded up to 514 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 <- -5                      # observed difference in means
sigma <- 28.5893826446113                     # 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(1579578)
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 = 514 (80% power), 688 (90% power)
 75% of observed effect: n per arm = 913 (80% power), 1,222 (90% power)
 50% of observed effect: n per arm = 2,053 (80% power), 2,749 (90% power)
power.t.test: 514.2 
Simulated power at n = 514 per arm: 0.804 (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 analysis is an ANCOVA comparing final values. It reports a mean difference of -5.00, a 95% confidence interval from -30.81 to 20.81, and p = 0.688. The recalculated result uses an unadjusted comparison derived from the posted group summaries, giving a mean difference of -5.0, a confidence interval from -28.89 to 18.89, and an unadjusted p-value of 0.682.

The results are close in estimated direction and magnitude, but the intervals and p-values are not identical. The recalculation cannot recreate the sponsor's adjusted analysis because it is based on summary statistics rather than the fitted ANCOVA model. The sponsor's posted analysis is therefore the appropriate primary result to cite. The registry does not flag a discrepancy between the reported and recalculated results; the distinction is that they are based on different statistical approaches, not that one replaces the other.

Neither result provides a precise estimate: both intervals include zero and extend into negative and positive values. The recalculation is a supplementary, unadjusted check, not an alternative sponsor analysis or a reason to disregard the posted ANCOVA.

14. Multiplicity, Missing Data and Analysis Populations

The registry lists one primary outcome and two secondary outcomes: progression-free survival and objective response rate. It also lists one primary analysis. The analysis result reported for the primary endpoint is an ANCOVA comparison of final values. The registry record does not report whether there was a prespecified hierarchy among outcomes, whether the secondary outcomes were confirmatory or exploratory, or whether multiplicity adjustments were planned or used.

Multiplicity matters in general when several outcomes or comparisons are tested: the more tests performed, the greater the chance that at least one result will appear notable by chance if all null hypotheses are true. Adjustments or a prespecified testing sequence can address that issue, but neither the presence nor absence of such a plan should be inferred from the short analysis record. The reported p-value of 0.688 should be interpreted as the value attached to the posted primary analysis; the record does not explain its place in a wider testing strategy.

The primary analysis population is called the Full Analysis Set. Its definition, inclusion rules, and relationship to randomization or treatment received are not reported in the registry record. The record also does not report how missing baseline or 8-week tumour measurements were handled, whether participants with incomplete measurements were excluded, or whether any imputation or sensitivity analysis was used. These details are relevant because the observed endpoint counts are 11 in each group, while the registry reports overall enrollment of 39.

The primary outcome is assessed at baseline and 8 weeks, with data accessed up to the cut-off on 4 December 2012. The record does not describe visit windows, measurement procedures, or whether the analysis used all available measurements. It also does not report the covariates included in ANCOVA or diagnostic checks for the model. These omissions do not invalidate the posted result, but they limit independent assessment of the analysis population, missing-data assumptions, and the extent to which the reported model followed a prespecified plan.

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 5% reporting threshold.

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: diarrhoea (15.4%), decreased appetite (15.4%), febrile neutropenia (8.3%), pyrexia (8.3%), peritonitis (8.3%).

The safety results list 13 participants at risk in the AZD8931 40mg plus paclitaxel group, of whom 6 were affected by serious adverse events. In the placebo plus paclitaxel group, 12 participants were at risk and 3 were affected by serious adverse events. The record also lists other adverse events for 13 of 13 participants in the AZD8931 group and 12 of 12 in the placebo group. These are counts of participants affected, not event counts, and the registry entry does not give additional details about severity grading or attribution.

The listed serious adverse event terms and percentages are diarrhoea, 15.4%; decreased appetite, 15.4%; febrile neutropenia, 8.3%; pyrexia, 8.3%; peritonitis, 8.3%; and muscular weakness, 8.3%. The record does not assign these listed terms to a particular treatment group in the summary, so they should not be interpreted as group-specific rates. Death counts are not reported in the registry record.

These safety figures describe the participants and events recorded in the registry; they do not by themselves establish that a treatment caused an event or allow a detailed comparison of risk. The small at-risk counts also make the posted counts a limited basis for interpreting uncommon events.

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

The main statistical result is an estimated mean difference of -5.00 percentage points in tumour-size change at 8 weeks, comparing AZD8931 plus paclitaxel with placebo plus paclitaxel. The sponsor's 95% confidence interval, from -30.81 to 20.81, includes differences in both directions, and the reported p-value is 0.688. The posted analysis therefore leaves substantial uncertainty about the average difference. It does not establish that the groups have the same effect, and it does not precisely identify the size or direction of a difference.

The endpoint is a continuous tumour-size measure at a single follow-up time. It is not a survival outcome, and the result alone does not answer the separate questions addressed by progression-free survival or objective response rate, both of which are listed as secondary outcomes. Detailed results for those secondary outcomes are not part of the primary comparison. Nor does the primary endpoint establish a longer-term effect or a patient-centred outcome not measured in this comparison.

The estimate comes from the posted Full Analysis Set, with 11 endpoint observations in each group. Overall enrollment is listed as 39, and the record gives different at-risk counts for safety. The participant flow and analysis-set definition are not reported in the registry record. The result should consequently be understood in the context of the endpoint analysis counts, rather than assumed to represent every enrolled participant.

The trial is listed as terminated, but the reason for termination is not reported in the registry record. The record also does not report the original sample-size justification, detailed missing-data procedures, or the full statistical analysis plan. These gaps limit assessment of the intended precision and whether all planned analyses are represented. The data describe the reported estimate and its uncertainty; they do not support a definitive conclusion about the comparative effect of the regimens or an inference of equivalence.

17. Why This Trial Matters Statistically

This trial illustrates why an effect estimate and its uncertainty are more informative together than a p-value alone. The point estimate is -5.00 percentage points, but the 95% confidence interval extends from -30.81 to 20.81. The estimate therefore has limited precision, and values on either side of zero remain compatible with the sponsor's analysis. A single p-value cannot convey that full range of uncertainty.

It also illustrates the distinction between a model-based reported analysis and an unadjusted recalculation from summary statistics. The sponsor's result comes from ANCOVA of final values. The recalculation is useful as a secondary check, but it cannot reproduce model adjustments that are not captured in the posted group summaries. Clear labeling of these approaches helps readers avoid treating slightly different estimates as conflicting answers to exactly the same calculation.

The endpoint analysis counts are small relative to the recorded enrollment, and the registry does not explain the participant flow or define the Full Analysis Set. That makes population and missing-data details important for statistical interpretation. The listed sample-size planning illustrations also show how required numbers change when planning around smaller effects, but they are not the study's original design calculations and should not be mistaken for evidence about the trial's achieved power.

Finally, this record shows the value of separating the primary endpoint from secondary outcomes and safety summaries. Tumour-size change at 8 weeks, progression-free survival, objective response rate, and serious adverse events are different kinds of outcomes. Their estimates and limitations should be interpreted according to their own definitions, populations, and analysis methods. The registry record leaves important aspects of those methods unreported, which is itself relevant when assessing how much weight to place on any single result.

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 primary endpoint analysis includes 11 participants in each group, whereas enrollment is listed as 39. The safety results use at-risk counts of 13 and 12. The registry record does not explain the participant flow or reconcile these populations, so the relationship between enrollment, endpoint availability, and safety follow-up cannot be assessed from the posted information alone.

The Full Analysis Set is named but not defined. The registry does not report the handling of missing 8-week measurements, the exact tumour measurement procedures, or the covariates included in the sponsor's ANCOVA. The outcome is based on baseline and 8-week measurements, but there is no detailed information about visit windows or how incomplete observations were treated. Those omissions limit assessment of reproducibility and sensitivity to analysis choices.

The study is listed as terminated, without a stated reason. The original target enrollment, sample-size justification, and design assumptions are not reported in the registry record. The planning figures associated with the observed effect are illustrations, not evidence of the original plan. The record also does not explain the masking roles, randomization procedure, or allocation concealment.

The primary result is imprecise, with a 95% confidence interval from -30.81 to 20.81. The record does not state a threshold for a clinically meaningful difference, so the importance of estimates within this interval cannot be classified using the registry information alone. Multiplicity plans for the primary and secondary outcomes are also not reported. Finally, the serious adverse event terms are listed without treatment-group assignment in the summary, and death counts are not reported in the registry record. These are limitations of the posted information, not evidence that unreported procedures or analyses were absent.

19. Related Tutorials

20. Related Statistical Calculators

21. Sources

• ClinicalTrials.gov: Assess the Efficacy of AZD8931 in Combination With Paclitaxel Versus Paclitaxel Alone in Patients With Gastric Cancer — NCT01579578.
• Registry record last updated 2014-09-12; status last verified 2014-09.
• Results first posted 2014-09-12.
• 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

NCT01579578 is a randomized, triple-masked, parallel-group phase IIa study of AZD8931 plus paclitaxel versus placebo plus paclitaxel in metastatic gastric or gastro-oesophageal junction cancer following progression after first-line therapy, among participants ineligible for trastuzumab based on HER2 status. The primary endpoint was percentage change in tumour size at 8 weeks in the Full Analysis Set.

The sponsor's ANCOVA reported a mean difference of -5.00, a 95% confidence interval from -30.81 to 20.81, and p = 0.688. The endpoint analysis included 11 participants per group. Enrollment was 39, and the study is listed as terminated; the reason is not reported in the registry record. Progression-free survival and objective response rate are listed as secondary outcomes.