Efficacy and Safety of Exenatide in Japanese Patients With Type 2 Diabetes Who Are Treated With Oral Antidiabetic(s): Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Efficacy and Safety of Exenatide in Japanese Patients With Type 2 Diabetes Who Are Treated With Oral Antidiabetic(s) (NCT00577824), a randomized phase 3 study. The primary outcome is Change in Glycosylated Hemoglobin (HbA1c) From Baseline to Week 24.

2. Clinical Question

3. Trial Design

The registry identifies this as a phase 3, randomized, double-masked, parallel-group trial with placebo control. The recorded enrollment was 181, and the trial status is completed. AstraZeneca is listed as the sponsor. The trial concerns Japanese patients with type 2 diabetes receiving oral antidiabetic medication who were not well controlled.

Three treatment groups are represented in the posted primary results: 5mcg exenatide twice daily, 10mcg exenatide twice daily, and placebo twice daily. The primary results include 71 participants in the 5mcg group, 72 in the 10mcg group, and 35 in the placebo group. The registry's intervention descriptions and arm descriptions do not provide additional regimen details. The randomization method, concealment procedures, and the planned allocation ratio are not reported in the registry record.

The recorded start was 2008-01, and primary completion was 2008-11. The main endpoint was assessed over 24 weeks, from baseline to week 24. The registry describes the trial as long term in its summary, but the primary endpoint is specifically the change through week 24. Information about treatment or follow-up beyond that endpoint is not reported in the registry record.

The primary analysis population is identified as the full analysis set, using last observation carried forward for missing values. The record does not give further details about which participants entered that population, the timing or extent of missing measurements, or how the full analysis set was defined. Those details matter for interpreting the analysis alongside the group counts attached to the endpoint.

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: Full analysis set; Last observation carried forward.

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 Glycosylated Hemoglobin (HbA1c) From Baseline to Week 24, measured over baseline, 24 weeks.

The primary endpoint was change in glycosylated hemoglobin (HbA1c) from baseline to week 24. It is expressed as percentage of hemoglobin, and the registry defines change as the week 24 value minus the week 0 value. A negative change therefore denotes a lower measured HbA1c at week 24 than at baseline; it is not a percentage-point difference between treatment groups.

The endpoint is reported as a least-squares mean, with a standard error for each group. In general, a continuous endpoint preserves the amount of change measured for each participant rather than reducing the result to whether a threshold was crossed. A group mean provides a compact estimate of average change, while its standard error describes the statistical uncertainty around that estimate. The endpoint can therefore be compared across the treatment groups and placebo on the same measurement scale.

The primary population is the full analysis set, with last observation carried forward. That method uses a participant's last recorded observation in place of a missing later observation, as indicated in the registry record. The record does not report the number of participants whose week 24 values were carried forward or the missing-data pattern. The endpoint describes change over the stated interval and does not, by itself, establish what happened after week 24.

7. Results Posted in the Registry

• 5mcg Exenatide BID: mean -1.34 versus -0.28 on control; difference -1.06 (95% CI -1.42 to -0.70), standardized effect d = -1.16; z-test p = <0.001.
• 10mcg Exenatide BID: mean -1.62 versus -0.28 on control; difference -1.34 (95% CI -1.70 to -0.98), standardized effect d = -1.46; z-test p = <0.001.

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.

Percentage of Patients Achieving HbA1c < 7.0%

• 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:
  • 5mcg Exenatide BID: 67.1% (47/70) versus 15.2% (5/33) on control; risk difference 52.0 percentage points (95% CI 35.5 to 68.4), risk ratio 4.43 (1.94–10.10); unadjusted two-proportion z-test p = <0.001.
  • 10mcg Exenatide BID: 71.0% (49/69) versus 15.2% (5/33) on control; risk difference 55.9 percentage points (95% CI 39.6 to 72.1), risk ratio 4.69 (2.06–10.66); unadjusted two-proportion z-test p = <0.001.
• Interpretation:
  • The 95% confidence interval excludes no difference, so the proportion with the event is statistically higher for 5mcg Exenatide BID 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 higher for 10mcg Exenatide BID than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.

Percentage of Patients Achieving HbA1c < 6.5%

• 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:
  • 5mcg Exenatide BID: 36.6% (26/71) versus 8.6% (3/35) on control; risk difference 28.0 percentage points (95% CI 13.5 to 42.6), risk ratio 4.27 (1.39–13.15); unadjusted two-proportion z-test p = 0.002.
  • 10mcg Exenatide BID: 47.2% (34/72) versus 8.6% (3/35) on control; risk difference 38.7 percentage points (95% CI 23.9 to 53.4), risk ratio 5.51 (1.82–16.70); unadjusted two-proportion z-test p = <0.001.
• Interpretation:
  • The 95% confidence interval excludes no difference, so the proportion with the event is statistically higher for 5mcg Exenatide BID 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 higher for 10mcg Exenatide BID than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.

Change in Fasting Blood Glucose

• Endpoint type: Continuous: a mean value per participant, compared as a difference in means.
• Test: Unadjusted two-sample z-test for the difference in means, using the posted standard errors.
• Result:
  • 5mcg Exenatide BID: mean -25.10 versus -7.60 on control; difference -17.50 (95% CI -30.57 to -4.43), standardized effect d = -0.54; z-test p = 0.009.
  • 10mcg Exenatide BID: mean -29.00 versus -7.60 on control; difference -21.40 (95% CI -34.45 to -8.35), standardized effect d = -0.66; z-test p = 0.001.
• Interpretation:
  • The 95% confidence interval excludes no difference, so the mean is statistically lower for 5mcg Exenatide BID 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 mean is statistically lower for 10mcg Exenatide BID than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.

The posted secondary results cover several kinds of outcomes, including HbA1c thresholds, fasting blood glucose, body weight, and lipid measures. The numeric results available for the threshold and fasting-glucose outcomes show a broad pattern consistent with the primary HbA1c comparison: the exenatide groups have larger decreases or higher threshold-attainment proportions than placebo. This is an overall description of direction, not a claim that every secondary measure has the same strength of evidence or interpretation.

The two HbA1c threshold outcomes are reported as percentages at week 24. The registry records a Cochran-Armitage analysis for each, with p < 0.001, and the recalculated unadjusted pairwise comparisons have their own p-values. Fasting blood glucose is reported as a continuous change, with least-squares means, standard errors, and a reported ANCOVA comparison for 10mcg exenatide versus placebo. These methods reflect different outcome formats and contrasts. The registry record does not report results for all listed secondary outcome topics in the detailed endpoint information, so a complete account of every secondary measure cannot be made from the posted results.

Secondary endpoints can help describe whether the primary pattern is accompanied by changes in related measures, but they generally carry less confirmatory weight when many outcomes are examined. In general, testing more outcomes creates more opportunities for a small p-value to occur by chance, even if no corresponding underlying effect exists. A prespecified hierarchy or multiplicity adjustment can address that issue, but such a plan is not reported in the registry record. The posted secondary p-values should therefore be interpreted as endpoint-specific findings rather than as a single multiplicity-controlled conclusion.

The registry records 17 outcome measures and 15 analyses. The information available does not establish which outcomes were prespecified as confirmatory, whether analyses were selected after reviewing the data, or whether the sponsor adjusted for multiple comparisons. These gaps limit the weight that can be placed on secondary statistical significance relative to the primary endpoint.

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

• Percentage of Patients Achieving HbA1c < 7.0%
• Percentage of Patients Achieving HbA1c < 6.5%
• Change in Fasting Blood Glucose
• Change in Body Weight
• Change in Total Cholesterol
• Change in Low Density Lipoprotein Cholesterol (LDL-C)
• Change in High Density Lipoprotein Cholesterol (HDL-C)
• Change in Triglycerides

9. Statistical Interpretation

10. Statistical Methods Explained

The primary endpoint is a continuous measure, and the posted group summaries are least-squares means with standard errors. The comparison values are unadjusted recalculations from the posted summary statistics. The registry also identifies an ANCOVA analysis for the 10mcg group versus placebo. These methods answer related but not identical statistical questions.

What does a least-squares mean represent?

In general, a least-squares mean is a model-based estimate of a group mean. It can account for variables included in the model, whereas an ordinary arithmetic mean is calculated directly from the observed values. The registry identifies the primary group results as least-squares means, but the full model specification and the covariates used are not reported in the registry record. The estimates should therefore be read as the reported model-based group summaries, without assuming details about adjustment that are not documented.

What is the role of the standard error?

A standard error describes the sampling uncertainty in an estimated mean. It is not the spread of individual participant values. The primary standard errors are 0.11 for each exenatide group and 0.15 for placebo. In general, a smaller standard error indicates a more precise estimate under the statistical model and assumptions used. It does not describe the range of outcomes among participants. The registry does not report standard deviations for the primary endpoint.

How should the unadjusted comparisons be read?

The recalculated mean difference contrasts each exenatide group's posted mean change with the placebo group's posted mean change. For 5mcg exenatide the estimate is -1.06; for 10mcg exenatide it is -1.34. These recalculations use the posted group summaries and are not covariate-adjusted. The pooled standard deviations used for planning illustrations are approximate, inferred from the standard errors and group sizes; they are not directly reported standard deviations. The recalculated results thus have a different basis from the sponsor's ANCOVA analysis.

A p-value describes how compatible the observed data are with a specified null hypothesis under a statistical model. In general, a small p-value is evidence against that null model; it does not measure the probability that the null hypothesis is true, the probability that the result will replicate, or the clinical importance of an estimate. The posted p-values for the primary unadjusted recalculations are each p < 0.001. The sponsor's reported ANCOVA comparison also has p < 0.001. The reported ANCOVA estimate and interval bounds are not reported in the registry record.

What does ANCOVA contribute?

In general, analysis of covariance compares outcomes between groups while adjusting for specified covariates. Adjustment can improve precision or account for baseline differences when the model is appropriately specified. The registry reports ANCOVA for the 10mcg exenatide and placebo groups, identifies the reported confidence level as 95%, and gives p < 0.001. The covariates, model details, adjusted estimate, and interval bounds are not reported in the registry record. The reported p-value should be understood as the result of that sponsor analysis, not as a p-value calculated from the unadjusted summary-statistic comparison.

Why do the binary outcomes use a different analysis?

Some secondary outcomes report the percentage of participants below an HbA1c threshold. These are binary outcomes: each participant either meets the stated threshold or does not. The registry reports Cochran-Armitage analyses for the two threshold outcomes, with p < 0.001 for each. In general, this procedure assesses a trend across ordered groups. The record does not report additional pairwise adjusted tests for these outcomes. The recalculated risk differences compare each exenatide group with placebo, but those unadjusted calculations are distinct from the reported trend analysis.

For continuous secondary outcomes such as fasting blood glucose, the registry reports least-squares means and standard errors, and an ANCOVA result for the 10mcg group versus placebo. A continuous mean difference and a binary risk difference have different interpretations and scales; neither should be substituted for the other. Statistical methods should be matched to the outcome type and the contrast being asked.

Why can over-dispersion matter?

In general, over-dispersion means that the variation in observed data is greater than a statistical model assumes. If a method assumes less variation than is present, uncertainty can be understated and p-values can appear more decisive than warranted. Conversely, an appropriate model for the outcome and its variation can better represent uncertainty. The registry record does not report an assessment of over-dispersion, so no conclusion about it can be made for this trial.

The registry records 17 outcome measures and 15 analyses. The statistical plan, including any prespecified hierarchy among outcomes or adjustment for multiple testing, is not reported in the registry record. This limits how confidently the family of reported p-values can be read as a single confirmatory set of tests.

11. Confidence Intervals and Precision

Precision concerns how narrowly an estimate is located, not whether it is clinically important. In general, a confidence interval describes a range of values compatible with the data and statistical procedure, subject to the model assumptions. A wider interval indicates greater uncertainty about the size of the underlying difference; an interval that excludes zero is inconsistent with no difference under the procedure used, but it does not resolve questions about clinical importance or bias.

The unadjusted recalculated mean difference for 5mcg exenatide versus placebo is -1.06, with interval bounds of -1.425 and -0.695. For 10mcg exenatide versus placebo, the recalculated mean difference is -1.34, with bounds of -1.705 and -0.975. Both intervals lie on the negative side of zero. These estimates are based on posted summary statistics and are not the adjusted ANCOVA estimates.

The registry identifies the sponsor's ANCOVA analysis as having a 95% confidence level, but its estimate and interval bounds are not reported in the registry record. The confidence level for the recalculated intervals is not separately stated in the comparison details. The group standard errors—0.11 for each exenatide group and 0.15 for placebo—give information about uncertainty in the group means, but they do not replace an interval for a between-group difference.

Precision also depends on the number of participants and the variability of the measured outcome. In general, more information can narrow uncertainty, while substantial variability can widen it. The endpoint group counts are 71, 72, and 35, respectively. The imbalance in those counts and the use of last observation carried forward are relevant context, but the registry does not report enough detail to quantify their effects on precision.

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 planning illustrations report participants per arm, not a total across both groups and not total events. They are based on the observed effect in each comparison, with scenarios for detecting 100% of that effect, 75% of the observed effect, or 50% of the observed effect. The 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 the stated fraction of the observed effect.

For the primary 5mcg comparison, planning for 100% of the observed effect gives 12 participants per arm for 80% power and 16 participants per arm for 90% power. Planning for 75% gives 21 participants per arm for 80% power and 28 participants per arm for 90% power. Planning for 50% gives 47 participants per arm for 80% power and 63 participants per arm for 90% power. For the primary 10mcg comparison, the corresponding figures are 8 and 10 for 100% of the observed effect, 14 and 18 for 75%, and 30 and 40 for 50%.

In plain terms, 80% power is the probability of detecting an effect of the specified size under the assumed model and design if that effect is truly present. The 90% column uses a higher probability of detection under those same assumptions. Smaller effects are harder to detect, so the planning illustrations require more participants per arm when the assumed effect is a smaller fraction of the observed effect. These figures are planning illustrations, not observed power, and they are not evidence about the trial's original design rationale. The original sample-size assumptions and justification 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, Δ is the difference in means to detect and σ is the common standard deviation.

Worked calculation for this trial

1. Observed difference Δ = -1.060; pooled standard deviation σ = 0.914.
2. n = 2 × (1.96 + 0.842)² × 0.914² / -1.060² = 11.7, rounded up to 12 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 <- -1.06                      # observed difference in means
sigma <- 0.914162079552804                     # 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(577824)
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 = 12 (80% power), 16 (90% power)
 75% of observed effect: n per arm = 21 (80% power), 28 (90% power)
 50% of observed effect: n per arm = 47 (80% power), 63 (90% power)
power.t.test: 12.7 
Simulated power at n = 12 per arm: 0.770 (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.

14. Analyses Reported by the Sponsor

The registry's reported primary sponsor analysis is an ANCOVA comparison of 10mcg exenatide with placebo. It reports p < 0.001 and identifies a 95% confidence level, but the adjusted estimate and confidence interval bounds are not reported in the registry record. The other primary comparison, for 5mcg exenatide versus placebo, is not included among the reported ANCOVA analyses in the registry record.

The recalculated comparisons use posted group means and standard errors to form simple, unadjusted estimates. They give mean differences of -1.06 for 5mcg exenatide versus placebo and -1.34 for 10mcg exenatide versus placebo, each with p < 0.001. Because the recalculations are not adjusted and the ANCOVA estimate is not posted, the two sets of estimates cannot be treated as numerically equivalent. Their p-values are both below 0.001 for the 10mcg comparison, but this agreement does not make their methods identical.

The registry discrepancy flag is false. The reported sponsor test and the recalculation therefore do not trigger a discrepancy qualification. For interpretation of the trial's own analysis, the sponsor's reported ANCOVA is the relevant registered result; the recalculation is a transparent, unadjusted comparison based on summary statistics and should be read as supplementary.

15. Multiplicity, Missing Data and Analysis Populations

Multiplicity arises when a study evaluates several outcomes, groups, or contrasts. In general, the more tests conducted, the greater the chance that at least one result will appear statistically unusual even when the corresponding null hypotheses are true. This does not make secondary results uninformative, but it affects how strongly they support a confirmatory conclusion.

The registry lists 17 outcome measures and 15 analyses. It reports a primary endpoint and several secondary outcome topics, including two HbA1c thresholds and fasting blood glucose. It also reports statistical tests for some of those secondary outcomes. The statistical analysis plan, any hierarchy among outcomes, and any adjustment for multiple testing are not reported in the registry record. The posted p-values therefore cannot be assumed to have been adjusted, and unadjusted secondary p-values should be treated cautiously.

Missing-data handling is specified for the primary analysis: the full analysis set used last observation carried forward. In general, this method carries a participant's last recorded measurement forward to represent a later missing measurement. It can make assumptions about what would have happened after the last observation, and those assumptions may not fit every participant's course. The registry record does not report how many primary endpoint values were imputed, why measurements were missing, or whether sensitivity analyses assessed alternative assumptions.

The analysis population is not described in enough detail to determine all inclusion rules. The primary group counts are 71, 72, and 35. Counts for secondary endpoints vary, and the reasons for those differences are not reported in the registry record. The definitions of the analysis sets for secondary outcomes, as well as the handling of missing values for each, are not reported in the registry record. These details affect interpretation of both the estimated effects and their precision.

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: gastric cancer (2.9%), gastric mucosal lesion (2.9%), duodenitis (2.9%), coronary artery stenosis (2.9%), sleep apnea syndrome (2.9%).

The serious adverse event summaries give counts of participants at risk and counts affected. The 5mcg exenatide group had 72 participants at risk, with 2 affected; the 10mcg group had 72 at risk, with 0 affected; and the placebo group had 35 at risk, with 4 affected. These are participant counts, not rates adjusted for duration of exposure. The registry record does not provide additional exposure-time information in these summaries.

The summaries for other adverse events list 72 participants at risk and 63 affected in each exenatide group, and 35 at risk and 22 affected in the placebo group. Those figures are distinct from the serious adverse event counts. The registry also lists serious-event terms with percentages, including gastric cancer, gastric mucosal lesion, duodenitis, coronary artery stenosis, and sleep apnea syndrome at 2.9%, and prostate cancer at 1.4%. The record does not give enough detail alongside this list to attribute every term to a particular group or to interpret these percentages as comparative risks.

Deaths are not reported in the registry record. The summaries do not establish whether an event was caused by treatment, nor do they provide a full account of event severity, timing, or duration. These figures describe the safety data posted in the registry and should not be interpreted as a complete comparative safety assessment.

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

The primary results show that the posted mean HbA1c changes from baseline to week 24 were more negative in both exenatide groups than in the placebo group. The unadjusted recalculated comparisons also place both mean differences below zero, with p < 0.001 for each. The sponsor's reported ANCOVA comparison of 10mcg exenatide with placebo has p < 0.001. Together, these statements describe the trial's posted statistical results; they do not establish how the findings should guide an individual's care.

A group mean is not a prediction for every participant. Participants may have had changes that differed from the group average, and the registry does not provide individual-level results. An estimate and its interval describe uncertainty under the method used, but they cannot account for every possible source of bias or model misspecification. The reported adjusted estimate and interval bounds are not reported in the registry record, so the sponsor analysis cannot be fully compared numerically with the recalculated result.

The endpoint covers baseline to week 24. It does not show whether the measured changes continued, diminished, or changed after that point. The registry summary describes the trial as long term, but longer-term outcome details beyond the primary timeframe are not reported in the registry record. Nor does the posted primary result establish effects on outcomes that were not measured or reported.

Interpretation also depends on the trial population, analysis population, missing-data approach, and statistical plan. The population described is Japanese patients with type 2 diabetes who were receiving oral antidiabetic medication but were not well controlled. The full analysis set used last observation carried forward, but the amount and pattern of missingness are not reported. The original sample-size rationale, full analysis plan, and multiplicity procedures are also not reported in the registry record.

The appropriate scope of the result is therefore narrow: it describes the recorded comparison of two twice-daily exenatide regimens with placebo for HbA1c change through week 24 in this trial. Statistical evidence about that endpoint should not be extended automatically to all secondary outcomes, other populations, longer follow-up, or individual outcomes.

18. Why This Trial Matters Statistically

This trial illustrates why the estimate, its uncertainty, and the analysis method need to be read together. The primary group means are least-squares estimates, the uncertainty summaries are standard errors, and the registry's own reported comparison uses ANCOVA. A recalculation from summary statistics can make the contrast more transparent, but it is unadjusted and cannot reconstruct the sponsor's full model when its specification and adjusted estimate are not reported.

The distinction is consequential for statistical interpretation. A p-value alone does not state the size of a difference, while a point estimate without an interval gives limited information about precision. The recalculated mean differences and their intervals add an estimate of the contrast and its uncertainty; the registry-reported ANCOVA p-value documents a separate adjusted test. Both should be understood in light of what each method does and does not show.

The trial also highlights the importance of analysis populations and missing-data assumptions. The primary endpoint used the full analysis set with last observation carried forward, but the extent of imputation and the missingness pattern are not reported. Those details affect the context for interpreting the estimates. Similarly, the registry lists multiple outcomes and analyses without reporting a multiplicity plan, which limits how confirmatory the secondary p-values can be considered.

Finally, the planning illustrations show how sample-size needs depend on the effect size one seeks to detect. They are based on the observed trial effects and specified planning assumptions; they do not explain the original design. Keeping planning calculations separate from observed trial results is essential when reading power and sample-size information.

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.

Several limitations follow from the information available for this trial. The study population is described as Japanese patients with type 2 diabetes treated with oral antidiabetic medication but not well controlled. The registry record does not report detailed eligibility criteria, participant characteristics, or the specific oral regimens used. The extent to which the results apply beyond the recorded population cannot be determined from these details alone.

The primary endpoint covers baseline to week 24. Although the registry summary calls the trial long term, outcome details beyond the primary 24-week timeframe are not reported in the registry record. The randomized allocation procedures, masking procedures, and any safeguards against unmasking are also not described in detail.

The primary analysis used last observation carried forward, but the number of values carried forward, reasons for missing measurements, and sensitivity analyses are not reported in the registry record. Counts also differ across some outcome summaries, and the reasons for those differences are not given. These omissions make it difficult to assess how missing data may have affected the estimates.

The sponsor's reported ANCOVA result gives p < 0.001 but not its adjusted estimate or interval bounds. Covariates and model specification are not reported. The unadjusted recalculations therefore provide useful contrast estimates but cannot stand in for the adjusted analysis. In addition, the statistical analysis plan and multiplicity handling are not reported, limiting assessment of the confirmatory status of secondary findings.

Safety information is also limited to posted counts and selected serious-event terms. Deaths are not reported in the registry record, and the summaries do not establish event relatedness, timing, or exposure-adjusted rates. These are limitations of the registry information and do not, on their own, establish that any particular bias or problem occurred.

20. Related Tutorials

21. Related Statistical Calculators

22. Sources

• ClinicalTrials.gov: Efficacy and Safety of Exenatide in Japanese Patients With Type 2 Diabetes Who Are Treated With Oral Antidiabetic(s) — NCT00577824.
• Registry record last updated 2015-04-09; status last verified 2015-03.
• Results first posted 2009-12-31.
• 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

NCT00577824 is a completed phase 3 randomized, double-masked, parallel-group, placebo-controlled trial sponsored by AstraZeneca. It enrolled 181 Japanese patients with type 2 diabetes who were receiving oral antidiabetic medication but were not well controlled. The recorded start was 2008-01, and primary completion was 2008-11.

The primary endpoint was change in HbA1c from baseline to week 24, analyzed in the full analysis set using last observation carried forward. Posted least-squares mean changes were -1.34 for 5mcg exenatide twice daily, -1.62 for 10mcg exenatide twice daily, and -0.28 for placebo twice daily. The registry reports an ANCOVA comparison of 10mcg exenatide with placebo with p < 0.001; adjusted estimate and interval bounds are not reported in the registry record. Recalculated unadjusted comparisons also report p < 0.001 for both exenatide groups versus placebo.