Efficacy of Once-Weekly Exenatide Versus Once or Twice Daily Insulin Detemir in Patients With Type 2 Diabetes: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Efficacy of Once-Weekly Exenatide Versus Once or Twice Daily Insulin Detemir in Patients With Type 2 Diabetes (NCT01003184), a randomized phase 3 study. The primary outcome is Percentage of Patients Achieving Glycosylated Hemoglobin (HbA1c) Concentration ≤7.0% With Weight Loss (≥1.0 kg) at Endpoint (Week 26).

2. Clinical Question

The clinical question concerns a combined target rather than HbA1c or body weight in isolation. A participant counted as meeting the primary endpoint only if the HbA1c and weight conditions were both met at the endpoint measurement. This makes the endpoint a measure of joint target attainment: a change in either component alone would not qualify a participant as a responder.

The registry describes the study as a Phase 3 randomized parallel-group trial. It was not masked. The posted primary result therefore compares the proportion meeting the combined target in the two assigned groups, rather than estimating an average change in a continuous measure. The comparison is specific to the recorded treatment context and Week 26 endpoint; the registry record does not establish how results would differ in other populations, treatment settings, or follow-up periods.

3. Trial Design

The registry identifies NCT01003184 as a completed Phase 3 randomized, parallel-group study sponsored by AstraZeneca. The record lists 222 participants as enrolled. The study start is recorded as 2009-10 and primary completion as 2011-07. The arms were Exenatide Once Weekly and Insulin Detemir, and masking was listed as none.

The stated purpose was to compare exenatide once weekly with insulin detemir with respect to glycemic control, body weight, lipids, safety, tolerability, and patient-reported outcomes. The registry title describes the population as people with type 2 diabetes treated with metformin alone or in combination with sulphonylurea. The arm descriptions contain no additional detail. The record does not report the randomization procedure, allocation concealment, site structure, eligibility criteria beyond the information in the title, or the procedures used to manage concomitant treatment.

The primary endpoint was assessed at Week 26 and combined an HbA1c threshold with a minimum weight-loss threshold. The primary analysis population was defined as randomized participants who received at least one dose of study drug, analyzed according to their assigned treatment. The primary endpoint denominators were 111 for exenatide and 105 for insulin detemir. Those arm-specific endpoint denominators should be distinguished from the registry’s enrollment count; they describe the participants included in the posted endpoint result.

The record lists 14 outcome measures and 14 analyses, and secondary outcomes include further HbA1c, weight, fasting glucose, and blood-pressure measures. The stated study purpose also mentions lipids, safety, tolerability, and patient-reported outcomes. The details of any additional outcomes beyond those listed are not reported in the registry record. The sample-size justification, planned analysis populations beyond the primary definition, and any interim-analysis procedures are not reported in the registry record.

4. Participant Flow and Accounting

The registry reports how participants moved through each stage of the study. All counts below are exactly as posted; missing counts are not inferred from other stages.

Before assignment: 6 patients who were enrolled and randomized, subsequently discontinued the study before receiving study drug. These patients were not included in analysis.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: The full analysis set (FAS) includes all data from all randomised patients receiving at least one dose of the study drug according to the treatment the patients were assigned.

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 Percentage of Patients Achieving Glycosylated Hemoglobin (HbA1c) Concentration ≤7.0% With Weight Loss (≥1.0 kg) at Endpoint (Week 26), measured over Baseline, Week 26.

The primary endpoint is the percentage of participants who achieved both HbA1c at or below 7.0% and weight loss of at least 1.0 kg at the Week 26 endpoint. It is a binary response outcome: each participant is classified as meeting both conditions or not meeting them. The posted result is therefore a proportion in each treatment group, not an average HbA1c change or average weight change.

A combined endpoint is used when the question concerns attaining multiple targets at once. Here, it captures the joint occurrence of the specified glycemic and weight criteria. A participant who meets only one of the two criteria does not meet the endpoint definition. This joint requirement gives the result a different interpretation from the separate secondary outcomes that assess HbA1c and weight individually.

The registry says the endpoint value uses the last post-baseline measurement at which both HbA1c and weight are non-missing and measured at the same visit. It also states that participants without a baseline weight measurement, with a protocol violation of baseline HbA1c at or below 7.0%, or with missing post-baseline HbA1c or weight measurements are included as non-responders for the primary objective. This rule makes the handling of specified missing or ineligible measurements part of the endpoint definition, not simply a choice to omit those participants.

The endpoint can summarize how often participants met the combined target in each assigned group, but it does not show the distribution of HbA1c or weight changes among individuals. Nor does it reveal how many participants narrowly missed one criterion, or which component was responsible when the joint target was not achieved.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The primary outcome is a proportion because each participant is categorized as either achieving the joint HbA1c and weight target or not achieving it. A confidence interval around a proportion communicates the uncertainty in estimating the underlying response rate from the analyzed participants. It does not mean that a specified proportion of individuals had outcomes inside the interval; the interval concerns the estimated group-level response percentage.

The primary comparison is reported as an odds ratio from logistic regression. Logistic regression is a method for modeling a binary outcome, such as whether the combined endpoint was achieved. An odds ratio compares the modeled odds between groups. Odds are not the same as probabilities, so an odds ratio should not be read as a direct multiplier of the response percentage. The reported 95% confidence interval describes the range of values reasonably compatible with the observed data and model, subject to the assumptions of the analysis.

The p-value for the primary comparison is reported as <.0001. A p-value describes how incompatible the observed data, or more extreme data, would be with a specified null model if that model and the analysis assumptions held. It is not the probability that the null hypothesis is true, nor is it a measure of the size or practical importance of the observed result. The registry does not report the full logistic-regression specification, including any covariates or details of model diagnostics.

For continuous secondary outcomes, the registry reports least-squares means and standard errors, with sponsor-reported comparisons from mixed-model analyses or ANCOVA. A least-squares mean is a model-based group estimate; it need not be identical to a simple arithmetic average. A standard error describes the precision of an estimated mean, rather than the variability among individual participants. The sponsor-reported confidence intervals and p-values reflect the methods named in the registry, although full model specifications and adjustment details are not reported.

The recalculated continuous comparisons are simple, unadjusted calculations from the posted summary statistics. They are not a substitute for the sponsor’s fitted analysis. For HbA1c change, the recalculated mean difference is 0.41, with a 95% confidence interval from 0.198 to 0.622 and p <0.001. For body-weight change, the recalculated mean difference is 3.67, with a 95% confidence interval from 2.703 to 4.637 and p <0.001. For fasting serum glucose change, the recalculated mean difference is -0.1, with a 95% confidence interval from -0.636 to 0.436 and p = 0.715. These calculations do not adjust for baseline values or other covariates, and their precision depends on the summary-statistic assumptions used in the recalculation.

The planning illustrations use observed effects as hypothetical targets, not as estimates of the trial’s achieved power. Their stated assumptions are a two-sided significance level of 0.05, equal allocation, and a normal approximation; each planning target is a stated fraction of the observed effect. Such illustrations show how sample-size requirements can vary with the effect one aims to detect. They do not establish that the trial was designed using those assumptions, and the trial’s own sample-size justification is not reported in the registry record.

For example, planning for 75% of the observed HbA1c effect gives 93 participants per arm for 80% power and 125 participants per arm for 90% power. Planning for 75% of the observed body-weight effect gives 25 participants per arm for 80% power and 33 participants per arm for 90% power. For fasting serum glucose, the observed recalculated effect is small relative to its uncertainty, and the corresponding planning figures are much larger. These are illustrations conditional on the chosen effect and assumptions; they are not observed power, proof of adequacy, or evidence that the planned trial had those enrollment targets.

Several features of the design affect interpretation. Random allocation supports a comparison of assigned groups, but the registry does not describe the randomization sequence or concealment procedures. The open-label design means masking was absent. The primary endpoint’s specified non-responder handling addresses particular missing or invalid measurements, but the registry does not describe sensitivity analyses or alternative missing-data assumptions. The primary population excludes randomized participants who did not receive at least one dose, according to the stated full analysis set definition.

General statistical point: over-dispersion matters when observed variability exceeds what a model assumes. In a binary-outcome analysis, misspecification can make uncertainty estimates too narrow or too wide. The registry does not report an over-dispersion assessment, so no conclusion about that issue can be drawn for this analysis. More generally, uncertainty intervals and p-values depend on the model and its assumptions; a numerical result should be interpreted alongside the endpoint definition, analysis population, and available information about the model.

Why do the recalculated estimates differ in form from the sponsor’s estimates?

The recalculations use posted group summaries and are unadjusted. The sponsor’s secondary analyses are identified as mixed-model analyses or ANCOVA and report least-squares mean differences. Those methods can account for model structure or baseline information that is not represented by a simple calculation from group summaries. The estimates need not match exactly in magnitude, confidence limits, or p-value. The registry’s reported analysis is the trial’s own analysis; recalculated values are supplementary checks based on the available summaries.

What can a confidence interval establish?

A confidence interval communicates statistical precision under the analysis method. A narrower interval indicates less uncertainty in the estimated quantity than a wider interval would, while the interval’s location shows which values are compatible with the data and method. It does not account automatically for unmeasured bias, incomplete reporting, multiplicity, or departures from model assumptions. Those issues require separate information and assessment, some of which is not reported in the registry record.

10. Analyses Reported by the Sponsor

The sponsor’s primary analysis is the logistic-regression odds ratio of 6.60, with a 95% confidence interval from 3.17 to 13.73 and p <.0001. This analysis compares the odds of meeting the combined HbA1c and weight-loss endpoint between the assigned groups. It is the registry’s reported inferential analysis for the primary endpoint.

The posted response percentages give the underlying group-level context: 44.1 for exenatide once weekly and 11.4 for insulin detemir. The odds ratio should not be interpreted as a direct ratio of those percentages. It is a model-based comparison of odds, while the percentages are observed group response estimates.

For the secondary continuous outcomes, the sponsor reports least-squares mean differences from mixed-model analyses for HbA1c and body weight, and from ANCOVA for fasting serum glucose. The simple recalculations use unadjusted differences based on the posted group summaries. Although they help readers check the broad direction and scale of the comparisons, they do not reproduce the sponsor’s models and may differ because they do not account for the same model structure or adjustment.

The registry does not flag a discrepancy between the sponsor’s reported results and the recalculated summaries. The sponsor’s fitted analyses remain the appropriate reference for the trial’s reported inferential results; the recalculations should be understood as supplementary unadjusted checks, not replacements for those analyses.

11. Multiplicity, Missing Data and Analysis Populations

Multiplicity arises when a study tests several outcomes or comparisons. Even when each individual test uses a conventional significance threshold, examining many tests can increase the chance that one or more results appear statistically notable by chance. Common approaches include identifying a primary hypothesis in advance, adjusting inference across a family of tests, or treating secondary analyses as exploratory. The registry record identifies a primary endpoint and lists secondary outcomes, but multiplicity adjustment procedures are not reported in the registry record.

Consequently, the primary comparison has a clearer role in the confirmatory interpretation than the collection of secondary p-values. This does not make secondary outcomes uninformative; it means that their results should be read with greater caution when the handling of multiple testing is unknown. The registry does not report a hierarchy of secondary hypotheses or specify which secondary comparisons, if any, were intended to support confirmatory claims.

The primary endpoint has an explicit rule for specified missing or invalid measurements. Participants lacking a baseline weight, with a stated baseline HbA1c protocol violation, or with missing post-baseline HbA1c or weight measurements are counted as non-responders for the primary objective. The endpoint uses the last post-baseline visit with both required measurements available at the same time point. This rule makes the primary classification more explicit than simply analyzing only participants with complete endpoint data.

However, the registry does not report how many participants were assigned to each of these missing-data categories, whether additional sensitivity analyses were conducted, or how missing values were handled for each secondary outcome. The secondary outcome denominators vary across measures, and the registry does not provide a complete account of missingness or reasons for unavailable measurements. Results may depend on how missing outcomes are handled, particularly if missingness is related to participant characteristics or outcomes.

The primary population is described as the full analysis set, including all randomized participants who received at least one dose and analyzed according to assigned treatment. This is not the same as an analysis of every randomized participant if some did not receive treatment. The registry does not report a separate intention-to-treat analysis including all randomized participants regardless of treatment received, nor does it describe other analysis populations in sufficient detail. The implications of excluding randomized participants who received no dose cannot be assessed from the registry record alone.

12. 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: chest pain (1.9%), chest discomfort (1.0%), pneumonia (1.0%), small cell lung cancer stage unspecified (1.0%), renal injury (1.0%).

The safety record lists 111 participants at risk in the exenatide group and 105 in the insulin detemir group. Serious adverse events were recorded for 6 participants in each group. These are participant counts, not a comparison adjusted for exposure time, event recurrence, or baseline risk. The registry does not provide those details in the posted safety summary.

The record also lists other adverse events for 87 participants receiving exenatide and 58 receiving insulin detemir. The registry’s summary does not provide enough detail to determine the precise classification rules for this category or to make a full comparison of event severity, duration, or causal attribution. The serious-event terms listed include chest pain, chest discomfort, pneumonia, small cell lung cancer stage unspecified, renal injury, and hyperglycaemia. The listed percentages are 1.9%, 1.0%, 1.0%, 1.0%, 1.0%, and 1.0%, respectively.

The serious-event terms should not be treated as a complete account of all safety outcomes. The registry record does not report deaths in the posted adverse-event summary. Counts alone do not establish whether an event was caused by a study drug, and the available information is not sufficient for a comprehensive comparative safety assessment. Safety figures describe the events recorded in the registry; they do not establish an overall balance of benefits and harms.

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

The result describes how often participants in the analyzed groups met a specific combined target at Week 26. The posted percentages and odds ratio indicate a difference between the groups on that endpoint under the registry’s stated analysis. They do not show that every participant in one group had a particular outcome, nor do they reveal the range of individual responses within each group.

The endpoint joins an HbA1c threshold and a weight-loss threshold. It therefore cannot identify whether the primary contrast arose mainly from glycemic outcomes, body-weight outcomes, or the way both criteria occurred together. Separate secondary measures offer some context, but they address different statistical questions and are subject to the limitations of secondary analyses and unreported multiplicity procedures.

The reported p-value does not give the probability that the result is true, and it is not a measure of clinical importance. The confidence interval communicates uncertainty under the fitted analysis, but it cannot eliminate possible bias or uncertainty arising from missing information about trial conduct, modeling, or outcome reporting. Randomization is an important design feature, but the registry does not report enough operational detail to assess every aspect of its implementation.

The comparison is specific to the recorded population, treatment regimens, analysis set, and follow-up endpoint. The registry title describes people treated with metformin alone or with a sulphonylurea; it does not establish that the result applies to other populations or treatment contexts. The primary completion date and Week 26 assessment do not provide evidence about outcomes beyond the recorded follow-up.

These results are a statistical description of the posted trial record, not medical advice. They do not, by themselves, determine an individual treatment choice or establish a comprehensive assessment of efficacy, safety, or long-term outcomes. Such conclusions would require broader evidence and clinical context beyond this registry comparison.

14. Why This Trial Matters Statistically

This trial illustrates how the choice of endpoint determines what a treatment comparison means. The primary outcome is not simply a change in HbA1c or a change in weight: it is the proportion achieving both a specified glycemic target and a specified weight-loss target. That combined definition makes the primary result directly about joint target attainment, while the secondary outcomes address component measures separately.

It also shows why effect measures need careful interpretation. The primary result is reported as an odds ratio, whereas the arm results are percentages. Those quantities are related but not interchangeable. An odds ratio describes relative odds under a logistic-regression analysis; the percentages describe group response proportions. Reading both together gives a clearer account of the result than treating the odds ratio as a change in percentage points.

The trial also provides a useful contrast between model-based sponsor analyses and simple unadjusted recalculations. The sponsor’s named methods use mixed models or ANCOVA for the continuous secondary outcomes, while recalculations from summary statistics do not reproduce those models. The distinction matters because an unadjusted estimate can be a useful numerical check without being the trial’s primary inferential result.

Finally, the record illustrates the importance of transparent reporting. The primary endpoint and its specified non-responder rule are described, but details about multiplicity, sample-size justification, randomization procedures, sensitivity analyses, and some aspects of missing data are not reported in the registry record. Those omissions limit how fully readers can evaluate the analysis plan and the robustness of the findings. Statistical interpretation depends not only on estimates and p-values, but also on knowing how outcomes and analyses were selected and conducted.

15. 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 identifies the study as open label, with no masking. This is an important design limitation to keep in mind when interpreting trial conduct and outcomes. The record does not state whether any outcome assessment was masked independently of participants and investigators.

The primary endpoint denominators are smaller than the recorded enrollment count, and the registry does not give a full participant-flow account explaining each difference. Although the primary endpoint specifies how several kinds of missing or invalid measurements are classified, the number of affected participants and the reasons for missing measurements are not reported in the registry record. Missing-data methods for the secondary outcomes are also not described in sufficient detail.

The registry provides only limited descriptions of the treatment arms. The title describes insulin detemir once or twice daily, but dosing procedures, treatment adjustment, adherence, and other regimen details are not reported in the registry record. The arm descriptions themselves contain no additional information. The record also does not provide the randomization method or allocation-concealment procedures.

The analysis information is incomplete for independent evaluation. The primary comparison is identified as logistic regression, and the secondary analyses are identified as mixed models or ANCOVA, but full model specifications, covariates, diagnostics, and analysis-plan details are not reported. Multiplicity handling and the trial’s sample-size justification are also not reported in the registry record. The planning illustrations based on observed effects cannot fill that gap or establish the design assumptions used by the investigators.

The follow-up described for the primary endpoint is Week 26. The posted result therefore addresses the specified endpoint at that time point and does not establish the durability of outcomes or later safety. The registry’s safety summary provides participant counts and selected serious-event terms, but not enough information to compare event timing, exposure, recurrence, attribution, or overall safety comprehensively. These limitations do not erase the reported estimates, but they constrain the scope and certainty of conclusions drawn from them.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Efficacy of Once-Weekly Exenatide Versus Once or Twice Daily Insulin Detemir in Patients With Type 2 Diabetes — NCT01003184.
• Registry record last updated 2015-04-07; status last verified 2015-03.
• Results first posted 2012-12-18.
• 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.

19. Record Summary

NCT01003184 is a completed Phase 3 randomized, open-label, parallel-group study sponsored by AstraZeneca. The registry records 222 participants enrolled and lists exenatide once weekly and insulin detemir as the study interventions. The population is described as people with type 2 diabetes receiving metformin alone or with a sulphonylurea.

The primary endpoint was the percentage achieving HbA1c at or below 7.0% together with weight loss of at least 1.0 kg at Week 26. The posted percentages were 44.1 for exenatide and 11.4 for insulin detemir; the reported odds ratio was 6.60, with a 95% confidence interval from 3.17 to 13.73 and p <.0001. The primary analysis used the stated full analysis set and counted specified missing or invalid endpoint measurements as non-responders.

Secondary outcomes cover further glycemic and weight measures and other listed outcomes. Their interpretation is limited by unreported multiplicity procedures and incomplete detail on analysis methods and missing data. The registry record does not report the trial’s sample-size justification or several design and analysis details needed for a fuller assessment.