Safety and Efficacy of Exenatide Once Weekly Injection Versus Metformin, Dipeptidyl Peptidase-4 Inhibitor, or Thiazolidinedione as Monotherapy in Drug-Naive Patients With Type 2 Diabetes (DURATION-4): Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Safety and Efficacy of Exenatide Once Weekly Injection Versus Metformin, Dipeptidyl Peptidase-4 Inhibitor, or Thiazolidinedione as Monotherapy in Drug-Naive Patients With Type 2 Diabetes (DURATION-4) (NCT00676338), a randomized phase 3 study. The primary outcome is Change in HbA1c From Baseline to Week 26.

2. Clinical Question

The trial asked how once-weekly exenatide compared with three active medicines when used as monotherapy in drug-naive patients with type 2 diabetes treated with diet and exercise. Its primary question concerned change in HbA1c from baseline to Week 26. The registry record also lists measures of fasting glucose, body weight, blood lipids, blood pressure, and hypoglycemic events.

This is a comparison among active treatments, not a comparison with no treatment or placebo. Accordingly, the reported differences address how the groups changed relative to one another within this trial. They do not by themselves establish the effect of any medicine relative to no treatment. The title and study summary describe the treatments as monotherapy, with patients otherwise treated with diet and exercise.

3. Trial Design

The registry record describes a randomized, double-masked, parallel-group phase 3 trial. It lists 820 participants enrolled and a completed study. The recorded start was 2008-11, and primary completion was 2010-07. The four treatment groups were exenatide once weekly, metformin, pioglitazone, and sitagliptin. The study summary describes exenatide at 2.0 mg once weekly and identifies the population as drug-naive patients with type 2 diabetes treated with diet and exercise.

Random allocation is intended to make treatment groups comparable on average at the start of a trial, while masking is intended to limit influences of treatment knowledge on care, reporting, or assessment. The registry record labels the study double-masked but does not report who was masked or how masking was maintained. A parallel design means participants were assigned to a treatment group rather than being described as switching among treatments.

The primary outcome was change in HbA1c from baseline to Week 26. The record identifies an intent-to-treat population comprising all randomized patients who had taken at least one dose of study drug. For that analysis, all scheduled post-baseline measurements were included, and unscheduled visit observations were carried forward to the next scheduled visits. The registry record does not report the original sample-size justification, treatment allocation ratio, detailed masking procedures, or the full statistical analysis plan. Those design details should not be inferred from the posted outcome summaries.

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: Intent to treat (ITT) population consisted of all randomized patients who had taken at least one dose of study drug. All scheduled post-baseline measurements were included in the analysis. Unscheduled visit observations were carried forward to the next scheduled visits.

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 HbA1c From Baseline to Week 26, measured over Baseline, Week 26.

The primary endpoint was change in HbA1c from baseline to Week 26, measured as a percentage of total hemoglobin. HbA1c is a measure used to assess blood glucose over a period of time; as general background, it is useful because it reflects more than a single glucose reading. The registry record reports least-squares mean changes, with standard errors as the measure of dispersion.

A change from baseline expresses the Week 26 outcome relative to each participant’s starting measurement, while a comparison of group means describes the difference between the treatment groups’ modeled average changes. Negative changes in the posted arm summaries indicate that the group’s mean HbA1c change was downward from baseline. The numerical outcome is not, by itself, a complete description of individual responses: a group mean does not show the full spread of participant-level changes.

The endpoint provides a common measure for comparing the four active-treatment groups at the specified follow-up. The registry record does not report a threshold for what would count as a clinically important difference, nor does it report the rationale for selecting Week 26 as the primary time point. Those matters should not be assumed from the outcome definition alone.

7. Results Posted in the Registry

• Metformin: mean -1.48 versus -1.53 on control; difference 0.05 (95% CI -0.14 to 0.24), standardized effect d = 0.05; z-test p = 0.614.
• Pioglitazone: mean -1.63 versus -1.53 on control; difference -0.10 (95% CI -0.31 to 0.11), standardized effect d = -0.10; z-test p = 0.347.
• Sitagliptin: mean -1.15 versus -1.53 on control; difference 0.38 (95% CI 0.17 to 0.59), standardized effect d = 0.38; 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.

Change in Fasting Serum Glucose (FSG) From Baseline to Week 26

• 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:
  • Metformin: mean -1.98 versus -2.25 on control; difference 0.27 (95% CI -0.12 to 0.66), standardized effect d = 0.14; z-test p = 0.173.
  • Pioglitazone: mean -2.57 versus -2.25 on control; difference -0.32 (95% CI -0.77 to 0.13), standardized effect d = -0.16; z-test p = 0.161.
• Interpretation:
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between Metformin and the control. The interval gives the range of effects that remain compatible with the data.
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between Pioglitazone and the control. The interval gives the range of effects that remain compatible with the data.

Change in Body Weight From Baseline to Week 26

• 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:
  • Metformin: mean -2.00 versus -2.04 on control; difference 0.04 (95% CI -0.54 to 0.62), standardized effect d = 0.01; z-test p = 0.893.
  • Pioglitazone: mean 1.52 versus -2.04 on control; difference 3.56 (95% CI 2.90 to 4.22), standardized effect d = 1.17; z-test p = <0.001.
• Interpretation:
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between Metformin and the control. The interval gives the range of effects that remain compatible with the data.
  • The 95% confidence interval excludes no difference, so the mean is statistically higher for Pioglitazone than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.

Change in Fasting Total Cholesterol (TC) From Baseline to Week 26

• 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:
  • Metformin: mean -0.22 versus -0.24 on control; difference 0.02 (95% CI -0.15 to 0.19), standardized effect d = 0.02; z-test p = 0.814.
  • Pioglitazone: mean 0.09 versus -0.24 on control; difference 0.33 (95% CI 0.13 to 0.53), standardized effect d = 0.38; z-test p = <0.001.
• Interpretation:
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between Metformin and the control. The interval gives the range of effects that remain compatible with the data.
  • The 95% confidence interval excludes no difference, so the mean is statistically higher for Pioglitazone than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.

The listed secondary outcomes cover several domains, including fasting serum glucose, body weight, cholesterol measures, hypoglycemic events, and systolic blood pressure. Posted numerical summaries include fasting glucose, body weight, and total cholesterol. Across those measures, the directions of change vary by outcome and treatment group; the results do not amount to one uniform pattern across all secondary measures. The registry record lists further outcomes, but their numerical results are not reported in the registry record.

These secondary outcomes add context to the primary HbA1c comparison because they describe other measured changes over the same general follow-up period. They do not replace the primary endpoint or turn it into a broader assessment of every benefit or risk. The posted secondary findings should be read as separate outcomes with their own estimates, uncertainty, and analysis details. A pattern in a secondary measure does not establish the same pattern for HbA1c, and the primary outcome cannot settle questions about outcomes it did not measure.

Unadjusted secondary results carry less weight when many outcomes and comparisons are considered. General explanation: testing many hypotheses creates more opportunities for apparently notable results to occur by chance, even if each test is interpreted at a conventional threshold. The registry record identifies 11 outcome measures and 27 analyses, but a multiplicity adjustment or hierarchy for interpreting the secondary outcomes is not reported. It is therefore appropriate to treat unadjusted secondary comparisons as exploratory rather than as independent confirmatory evidence. The registry record also does not report all analysis details for every secondary measure, so the evidence should not be generalized beyond the comparisons and results actually posted.

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

• Change in Fasting Serum Glucose (FSG) From Baseline to Week 26
• Change in Body Weight From Baseline to Week 26
• Change in Fasting Total Cholesterol (TC) From Baseline to Week 26
• Change in Fasting High-Density Lipoprotein (HDL) From Baseline to Week 26
• Ratio of Fasting Triglycerides at Week 26 to Baseline
• Assessment on Event Rate of Treatment-emergent Major Hypoglycemic Events
• Assessment on Event Rate of Treatment-Emergent Minor Hypoglycemic Events
• Change in Systolic Blood Pressure From Baseline to Week 26.

9. Statistical Interpretation

10. Statistical Methods Explained

The primary outcome is continuous: HbA1c is measured on a numeric scale, and the registry reports a mean change for each treatment group. A least-squares mean is a model-based estimate of a group mean. General explanation: such estimates can account for features specified in a statistical model, but the registry record does not describe the full model specification, its assumptions, or all included terms. The reported standard error describes the estimated uncertainty in each least-squares mean; it is not the same as the standard deviation of individual participant measurements.

What does a group contrast represent?

A mean difference compares the estimated changes in two groups. The registry’s sponsor results label the contrast as a least-squares mean difference and report results from a mixed models analysis. The direction of a contrast depends on which group is subtracted from which. In the sponsor results, contrasts are presented as exenatide compared with the named active treatment. The unadjusted re-calculations use the reverse direction, comparator minus exenatide. A reversed direction changes the sign of an estimate and its interval but does not change the underlying separation between the groups.

A mixed model is a general statistical approach for analyzing repeated measurements, such as measurements taken at multiple visits. It can model the relationship among measurements from the same participant. The registry record names a mixed models analysis for the sponsor comparisons, but does not report details such as the covariance structure, model terms, handling of treatment-by-visit effects, or degrees-of-freedom method. Those details cannot be reconstructed from the reported estimates alone.

What do the p-values say?

A p-value summarizes how compatible the observed data are with a specified statistical null hypothesis under the assumptions of the analysis. It is not the probability that a treatment effect is true, and it does not measure the size or clinical importance of a difference. The primary sponsor p-values were 0.620, 0.328, and <.001 for the comparisons with metformin, pioglitazone, and sitagliptin, respectively. The record classifies these as noninferiority or equivalence analyses. Without the margins and decision rules, a p-value should not be treated as a standalone determination of noninferiority or equivalence.

The recalculated p-values are unadjusted and come from simple comparisons of posted summary statistics. They are not the same as the sponsor’s model-based tests. The registry record flags a discrepancy, so the sponsor’s reported analysis is the primary reference for interpreting the trial’s formal comparison results.

Why can over-dispersion matter?

General explanation: for a continuous outcome, variability among individual measurements affects how precisely a mean or mean difference can be estimated. More variability generally means greater uncertainty for a fixed amount of information. The re-calculations estimate pooled standard deviations from the posted standard errors and sample sizes, rather than using observed participant-level standard deviations. That approximation is identified in the posted calculations. It may not reproduce the sponsor model’s handling of repeated measurements, covariates, or missing observations. Over-dispersion is especially a concern in count outcomes when the observed variability exceeds what a simpler count model assumes; the primary endpoint is continuous, so count-model over-dispersion is not the issue governing these HbA1c comparisons.

The primary analysis population included randomized patients who took at least one dose of study drug. The registry says that all scheduled post-baseline measurements were included, and that unscheduled observations were carried forward to the next scheduled visit. This describes the stated analysis population and one aspect of visit handling, but it does not fully explain how missing scheduled values were treated. No further missing-data method, sensitivity analysis, or interim-analysis plan is reported in the registry record.

11. Confidence Intervals and Precision

A confidence interval expresses a range of values compatible with the data and the specified analysis, under its assumptions. It reflects both the estimated contrast and its uncertainty; a narrow interval is more precise than a wide one. An interval that crosses zero includes no difference as one value compatible with the data, but that fact alone does not establish equivalence. For equivalence or noninferiority, interpretation also depends on a prespecified margin and analysis rule, which are not reported in the registry record.

For the primary sponsor comparisons, the reported intervals are 98.3% confidence intervals. The interval for exenatide versus metformin runs from -0.26 to 0.17, and the interval for exenatide versus pioglitazone runs from -0.15 to 0.35. Both include zero, indicating that the reported estimates do not sharply distinguish the groups under these analyses. The interval for exenatide versus sitagliptin runs from -0.62 to -0.13 and does not include zero. The corresponding p-value is <.001. These statements describe statistical uncertainty, not the clinical importance of the contrasts.

The unadjusted re-calculated intervals are based on summary statistics and differ somewhat from the sponsor’s model-based results. Because the record identifies a discrepancy, the sponsor analysis takes precedence. The recalculated intervals should be read as approximate, unadjusted contrasts rather than as replacements for the reported analysis. Across either kind of estimate, the interval endpoints help show the range of uncertainty; they do not guarantee that the true effect lies within the interval in any particular trial.

12. Planning a Similar Trial: Sample Size

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

The sample-size figures are planning illustrations based on observed effects, not the trial’s original design justification. For these continuous outcomes, the figures mean participants per arm, not a total across all arms and not total events. The planning assumptions are a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation. The effect to detect is set to the stated fraction of the observed effect. The original trial’s design assumptions and sample-size justification are not reported in the registry record.

The 80% power column gives the number of participants per arm estimated to yield an 80% chance of detecting the specified effect under those planning assumptions, if that effect and the assumptions are correct. The 90% power column gives the corresponding number for a 90% chance. These are prospective planning concepts: they do not describe the probability that the completed trial’s result is correct, nor do they measure the chance that the trial found a true effect.

Each planning line uses a different fraction of the observed effect: 100%, 75%, or 50%. Smaller target effects require more participants to distinguish the effect from random variation with the same planned power and significance level. The figures can become very large when the observed contrast is small relative to the estimated variation, as in some of the posted comparisons. Conversely, larger observed contrasts can produce smaller planning estimates. Such figures are sensitive to the observed effect and should not be treated as a recommended design without a clinically justified target effect and suitable assumptions.

This is not observed power. The illustrations are hypothetical sample-size calculations, not a retrospective score of the trial’s ability to detect its actual effect. They also do not replace the sponsor’s reported analyses, account for all features of the trial design, or establish clinical importance.

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 = 0.050; pooled standard deviation σ = 1.034.
2. n = 2 × (1.96 + 0.842)² × 1.034² / 0.050² = 6,707.3, rounded up to 6,708 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 <- 0.05                      # observed difference in means
sigma <- 1.03353761421634                     # 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(676338)
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 = 6,708 (80% power), 8,980 (90% power)
 75% of observed effect: n per arm = 11,925 (80% power), 15,964 (90% power)
 50% of observed effect: n per arm = 26,830 (80% power), 35,917 (90% power)
power.t.test: 6708.3 
Simulated power at n = 6,708 per arm: 0.803 (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 reports sponsor comparisons from mixed models analysis, expressed as least-squares mean differences. It also includes simple unadjusted re-calculations from the posted group summaries. Those calculations are not identical procedures: the re-calculations use the reported means, standard errors, and sample sizes, while the sponsor analysis uses a mixed model. The estimates also use opposite contrast directions, which accounts for reversed signs when the same groups are compared.

The registry record marks a discrepancy between the re-calculated results and the sponsor’s own analysis. The sponsor results therefore take precedence for interpreting the trial’s reported formal comparisons. For example, the sponsor estimate for exenatide versus metformin is -0.05, whereas the unadjusted re-calculation is 0.05 in the comparator-minus-exenatide direction. The estimates have similar magnitude but are not interchangeable; their intervals and p-values also differ. A similar distinction applies to the other comparisons.

The record classifies the sponsor analyses as noninferiority or equivalence analyses. The relevant margins and detailed decision criteria are not reported in the registry record, so a reader cannot independently verify the formal conclusion from the estimates and p-values alone. The re-calculations provide a useful check based on posted summaries, but do not supersede the sponsor’s model-based results or resolve missing details about the prespecified analysis.

15. Multiplicity, Missing Data and Analysis Populations

The record lists 11 outcome measures and 27 analyses. Multiple outcomes and pairwise comparisons create a multiplicity issue: general statistical explanation, the more hypotheses examined, the greater the chance that at least one result appears unusual by chance if each is judged separately. An adjustment, a prespecified testing hierarchy, or another multiplicity strategy can help control how results are interpreted. Such a strategy is not reported in the registry record. The posted p-values should therefore be read in light of the number of outcomes and analyses, especially for secondary findings.

The primary analysis population is described as all randomized patients who had taken at least one dose of study drug. The record says all scheduled post-baseline measurements were included and that unscheduled observations were carried forward to the next scheduled visits. This is the stated approach for the primary outcome. The registry record does not report a fuller missing-data method, the amount or pattern of missing primary measurements, or sensitivity analyses testing alternative assumptions about missing data.

The record names mixed models analysis for the sponsor’s primary comparisons but does not provide the full model specification. It also does not report the noninferiority or equivalence margins, the original sample-size rationale, or whether interim analyses were planned or conducted. These omissions limit an independent appraisal of the formal testing framework. They should not be filled in by assuming a conventional protocol or standard statistical practice. The same caution applies to the secondary outcomes: details about their testing order, adjustment, and missing-data handling are not reported in the registry record.

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: anaemia (0.6%), angina pectoris (0.6%), arteriovenous malformation (0.6%), arteritis (0.6%), arthritis bacterial (0.6%).

The adverse-event summary reports serious adverse events by treatment group, with both the number at risk and the number affected. The listed at-risk counts are 248 for exenatide, 246 for metformin, 163 for pioglitazone, and 163 for sitagliptin. The corresponding numbers affected are 4, 12, 9, and 3. These are counts, not estimates of treatment effect, and they should be read alongside their group-specific denominators. They do not identify whether an event was caused by treatment.

The registry also lists several serious-event terms at 0.6% each: anaemia, angina pectoris, arteriovenous malformation, arteritis, arthritis bacterial, and ascites. These entries alone do not describe all event types, duration, severity beyond the serious-event classification, or the clinical course of affected participants. Death counts are not reported in the registry record. The adverse-event summary should not be treated as a complete safety comparison or as evidence that one treatment is safer than another without additional definitions and analysis.

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

The results describe how group-average HbA1c changed from baseline to Week 26 among the trial’s stated analysis population, and how the sponsor’s model-based comparisons estimated differences between exenatide and three active treatments. The estimates are accompanied by confidence intervals and p-values, which communicate statistical uncertainty under the reported analyses. The comparison with sitagliptin had an interval that did not include zero, while the intervals for comparisons with metformin and pioglitazone did include zero. The registry classifies these as noninferiority or equivalence analyses, but the margins are not reported.

These findings do not, by themselves, establish that treatments are equivalent, clinically interchangeable, or different in a way that matters to patients. A nonsignificant result does not prove no difference, and statistical significance does not establish clinical importance. The registry record does not report a threshold for a clinically important HbA1c difference or the margins needed to interpret noninferiority or equivalence. The results also cannot be used to infer effects on outcomes that were not measured or whose numerical results are not reported.

The trial compared active monotherapies in drug-naive patients with type 2 diabetes treated with diet and exercise. Its conclusions should be understood within that population, design, endpoint, and follow-up. The posted summary does not answer how these treatments compare in other patient groups, under other treatment regimens, or over other time periods. Nor does it provide a complete accounting of all benefits and harms. Statistical estimates are evidence about the trial’s measured outcomes, not medical advice or a guarantee of an individual response.

18. Why This Trial Matters Statistically

This trial illustrates several important principles in reading clinical results. First, the estimated group changes and the estimated between-group contrasts answer different questions. Each group’s change from baseline describes its modeled average trajectory, while the contrast quantifies the relative difference between two groups. A treatment group’s change alone is not enough to establish how it compares with another active treatment.

Second, the statistical framework matters. The sponsor’s primary comparisons are identified as mixed models analyses and classified as noninferiority or equivalence analyses. Their interpretation depends on margins and decision rules, which are not reported in the registry record. The presence of p-values and confidence intervals does not substitute for those design details. The record also flags a discrepancy with simple unadjusted re-calculations, making it important to distinguish a summary-statistic check from the sponsor’s formal model-based analysis.

Third, uncertainty and multiplicity both affect interpretation. Confidence intervals show the precision of estimated contrasts, while numerous outcome measures and analyses make isolated secondary p-values less conclusive in the absence of a reported adjustment strategy. Finally, sample-size illustrations derived from observed effects are not the original trial’s sample-size rationale and are not observed power. Keeping these distinctions clear helps prevent an estimate from being interpreted as more definitive, more clinically meaningful, or more broadly applicable than the registry results support.

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.

The registry record leaves important details unavailable for independent assessment. It does not report the noninferiority or equivalence margins, the full statistical analysis plan, or the original sample-size justification. Although it names mixed models analysis, it does not give the full model specification. The record also does not report whether interim analyses were planned or conducted. These gaps limit evaluation of the formal comparison framework.

The primary analysis population required randomization and at least one dose of study drug. The record states how scheduled and unscheduled observations were handled, but does not describe the amount or pattern of missing data, the handling of missing scheduled visits, or sensitivity analyses. The posted group summaries also do not show participant-level distributions. The unadjusted re-calculations rely on estimated variation from standard errors and sample sizes, and the record flags a discrepancy with the sponsor analysis.

Multiplicity adjustment for the several outcomes and analyses is not reported. Numerical results are not reported in the registry record for some of the listed secondary outcomes, and death counts are not reported. Safety counts alone cannot establish causality or provide a complete safety comparison. Finally, the evidence pertains to the registered population, four active monotherapy groups, and Week 26 primary endpoint. Broader conclusions about other populations, treatment settings, time periods, or unmeasured outcomes are not supported by these posted results.

20. Related Tutorials

21. Related Statistical Calculators

22. Sources

• ClinicalTrials.gov: Safety and Efficacy of Exenatide Once Weekly Injection Versus Metformin, Dipeptidyl Peptidase-4 Inhibitor, or Thiazolidinedione as Monotherapy in Drug-Naive Patients With Type 2 Diabetes (DURATION-4) — NCT00676338.
• Registry record last updated 2015-04-09; status last verified 2015-03.
• Results first posted 2012-11-26.
• 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

NCT00676338, DURATION-4, is recorded as a completed randomized, double-masked, parallel phase 3 trial with 820 participants enrolled. It compared exenatide once weekly with metformin, pioglitazone, and sitagliptin as monotherapy in drug-naive patients with type 2 diabetes treated with diet and exercise.

The primary outcome was change in HbA1c from baseline to Week 26. Sponsor comparisons used mixed models analysis and are classified in the registry as noninferiority or equivalence analyses. The registry does not report the margins needed to independently evaluate that classification. It also flags a discrepancy between the sponsor results and unadjusted re-calculations; the sponsor analysis takes precedence. Multiple secondary outcomes and adverse-event summaries are listed, but not all corresponding numerical results or analysis details are reported in the registry record.