This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record.
1. Trial at a Glance
AWARD-11 was a randomized, double-blind, parallel phase 3 trial evaluating dulaglutide in participants with type 2 diabetes. The registry reports three dulaglutide treatment groups and an enrollment of 1,842 participants.
| Feature | AWARD-11 |
|---|---|
| Trial | AWARD-11 |
| ClinicalTrials.gov identifier | NCT03495102 |
| Phase | Phase 3 |
| Condition | Diabetes Mellitus, Type 2 |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Double |
| Primary purpose | Treatment |
| Enrollment | 1,842 |
| Intervention | Dulaglutide |
| Study status | Completed |
| Start date | April 5, 2018 |
| Primary completion date | May 28, 2019 |
| Lead sponsor | Eli Lilly and Company |
2. Clinical Question
The central statistical question was whether increasing dulaglutide dose was associated with greater improvement in glycemic control at Week 36. The primary endpoint was change in HbA1c from baseline to Week 36, with superiority comparisons of dulaglutide 3 mg and 4.5 mg against dulaglutide 1.5 mg.
Population
Participants with type 2 diabetes mellitus enrolled in the phase 3 AWARD-11 study.
Intervention
Dulaglutide, with three treatment groups represented in the registry results: 1.5 mg, 3 mg, and 4.5 mg.
Comparator
The primary statistical comparisons were dulaglutide 1.5 mg versus 3 mg and dulaglutide 1.5 mg versus 4.5 mg.
Primary question
Do the higher dulaglutide doses produce a greater change in HbA1c from baseline to Week 36 than dulaglutide 1.5 mg?
3. Trial Design
Dulaglutide 1.5 mg
- Dulaglutide intervention
- Serious adverse events: 53/612
- Primary comparisons against 3 mg and 4.5 mg
Dulaglutide 3 mg
- Dulaglutide intervention
- Serious adverse events: 45/616
- Compared with 1.5 mg in posted analyses
Dulaglutide 4.5 mg
- Dulaglutide intervention
- Serious adverse events: 38/614
- Compared with 1.5 mg in posted analyses
The enrollment counts associated with the three safety populations sum to 1,842, matching the reported trial enrollment. The registry describes the study as randomized, double-blind, and parallel.
4. Trial Timeline
Study start
The registry lists April 5, 2018 as the study start date.
Primary completion
The registry lists May 28, 2019 as the primary completion date.
Registry status
The ClinicalTrials.gov record lists the study status as completed.
5. Endpoints
The registry reports one registered primary endpoint and additional posted secondary outcome measures. The primary endpoint and the posted analyses focus on Week 36 comparisons.
| Endpoint | Time frame | Statistical method | Effect measure |
|---|---|---|---|
| Change in Hemoglobin A1c (HbA1c) From Baseline | Baseline, Week 36 | Mixed-effects model with repeated measures | Mean difference |
| Change in Body Weight From Baseline | Baseline, Week 36 | Mixed-effects model | Mean difference |
| Percentage of Participants Achieving HbA1c Target <7.0% | Week 36 | Logistic regression | Odds ratio |
| Change in Fasting Serum Glucose (FSG) From Baseline | Baseline, Week 36 | Mixed-effects model | Mean difference |
For the primary HbA1c endpoint, the registry defines HbA1c as a measure used to identify average plasma glucose concentration over prolonged periods. The registered definition specifies that the least-squares mean was estimated from a mixed-effects model with repeated measures containing baseline, pooled country, treatment, time, and treatment-by-time terms, using Type III sums of squares.
6. Statistical Methodology
Mixed-effects model with repeated measures
The primary HbA1c analysis used a mixed-effects model with repeated measures. This is appropriate for longitudinal outcomes because HbA1c is measured repeatedly over time and the model can represent the relationship between treatment, time, baseline value, and their interaction.
The registry states that the least-squares mean was estimated from this mixed-effects model with repeated measures using Type III sums of squares.
Why use a mixed-effects model?
A repeated-measures framework allows the analysis to use information from measurements collected over time rather than reducing each participant to a single observed value. The treatment-by-time term allows the model to represent whether the difference between treatment groups changes over the study period.
Least-squares means
The primary HbA1c results are expressed as mean differences derived from the model. A least-squares mean is a model-adjusted mean corresponding to the covariate and design structure specified in the analysis model. It is therefore not necessarily identical to a simple arithmetic mean calculated separately within each treatment group.
Logistic regression
The binary secondary endpoint of achieving an HbA1c target below 7.0% at Week 36 was analyzed with logistic regression. The effect measure was an odds ratio.
An odds ratio above 1 indicates higher estimated odds of achieving the target in the group represented in the numerator. It is not the same quantity as a risk ratio or a difference in percentages.
Superiority testing
The posted analyses use a superiority hypothesis. The question is therefore whether the higher-dose group and the dulaglutide 1.5 mg group differ in the specified direction sufficiently to support a statistical treatment difference under the model and test used.
7. Primary Results: HbA1c
The primary endpoint was change in HbA1c from baseline to Week 36. Two primary comparisons were posted, both using mixed-model analysis and a mean difference as the effect measure.
Dulaglutide 1.5 mg vs 3 mg
Mean difference in HbA1c change
95% CI: −0.29 to −0.06 · P = 0.003
Baseline to Week 36; mixed-effects model; superiority comparison.
The estimated mean difference in change in HbA1c was −0.17 percentage points for the dulaglutide 1.5 mg versus 3 mg comparison as reported by the registry. The negative estimate indicates a greater reduction in HbA1c for the group represented by the first treatment in the reported comparison.
The 95% confidence interval extends from −0.29 to −0.06. This interval describes uncertainty around the model-based estimated difference; it does not mean that individual patients' HbA1c changes are confined to that range.
The P-value of 0.003 quantifies the evidence against the relevant null hypothesis under the specified statistical model. It does not measure the magnitude or clinical importance of the HbA1c difference. The estimate and its confidence interval are needed to understand the size and precision of the treatment difference.
Because this is a repeated-measures model, interpretation also depends on the model specification, the observed longitudinal data, and the assumptions underlying the mixed-effects analysis. The registry reports the analysis model but does not provide additional diagnostic information about those assumptions.
Dulaglutide 1.5 mg vs 4.5 mg
Mean difference in HbA1c change
95% CI: −0.45 to −0.22 · P < 0.001
Baseline to Week 36; mixed-effects model; superiority comparison.
The estimated mean difference in HbA1c change was −0.34 percentage points for the dulaglutide 1.5 mg versus 4.5 mg comparison. The larger absolute difference compared with the 1.5 mg versus 3 mg comparison indicates a larger estimated separation in HbA1c change for this comparison.
The 95% confidence interval of −0.45 to −0.22 provides the model-based uncertainty interval for the estimated difference. It is narrower in absolute width than the corresponding interval for the 1.5 mg versus 3 mg comparison, indicating greater numerical precision for this posted estimate.
The P-value of <0.001 indicates strong statistical evidence against the null hypothesis specified for the superiority comparison. It does not by itself quantify treatment benefit or establish whether the numerical difference is clinically important.
As with the other HbA1c analysis, the result is conditional on the prespecified analysis population and the mixed-effects model used for the repeated measurements.
8. Comparing the Primary HbA1c Estimates
| Comparison | Mean difference | 95% CI | P-value |
|---|---|---|---|
| Dulaglutide 1.5 mg vs 3 mg | −0.17 | −0.29 to −0.06 | 0.003 |
| Dulaglutide 1.5 mg vs 4.5 mg | −0.34 | −0.45 to −0.22 | <0.001 |
Both posted primary comparisons have confidence intervals entirely below zero. In the context of the reported superiority analyses, that indicates a statistically supported difference in the modeled change in HbA1c for each comparison.
The two estimates should not be interpreted as evidence that the treatment effect increases linearly with dose. The registry reports the pairwise comparisons, but the posted information does not provide a dose-response model, a test for linear trend, or an estimate of incremental effect per dose unit.
9. Secondary Results: Body Weight
Change in body weight from baseline to Week 36 was analyzed using mixed-model analysis. The registry reports two pairwise comparisons against dulaglutide 1.5 mg.
| Comparison | Mean difference | 95% CI | P-value |
|---|---|---|---|
| Dulaglutide 1.5 mg vs 3 mg | −0.9 kg | −1.4 to −0.4 kg | 0.001 |
| Dulaglutide 1.5 mg vs 4.5 mg | −1.6 kg | −2.1 to −1.1 kg | <0.001 |
The negative mean differences indicate greater modeled reduction in body weight for the higher-dose comparisons as represented by the reported treatment contrasts. The estimated differences were −0.9 kg for 1.5 mg versus 3 mg and −1.6 kg for 1.5 mg versus 4.5 mg.
The confidence intervals exclude zero for both comparisons. That provides statistical evidence of a difference under the reported mixed-model framework. The intervals should not be interpreted as ranges containing the body-weight change of individual participants.
The body-weight results are secondary endpoints. They therefore provide additional information about the trial's treatment effects but answer a different question from the primary HbA1c endpoint.
10. Secondary Results: HbA1c Target Achievement
The registry also reports the percentage of participants achieving an HbA1c target of <7.0% at Week 36. These outcomes were analyzed with logistic regression, with odds ratios as the effect measure.
| Comparison | Odds ratio | 95% CI | P-value |
|---|---|---|---|
| Dulaglutide 1.5 mg vs 3 mg | 1.49 | 1.12 to 1.98 | 0.006 |
| Dulaglutide 1.5 mg vs 4.5 mg | 2.23 | 1.65 to 3.01 | <0.001 |
An odds ratio of 1.49 means that the estimated odds of achieving HbA1c <7.0% were 1.49 times the corresponding odds for the comparison represented in the denominator, under the logistic regression model. An odds ratio of 2.23 similarly represents 2.23 times the odds under that model.
Odds are not probabilities. An odds ratio of 2.23 does not mean that 2.23 times as many participants achieved the target, nor does it mean that the probability of success increased by 123 percentage points.
The 95% confidence intervals for both comparisons exclude 1. The P-values of 0.006 and <0.001 provide statistical evidence against the corresponding null hypothesis of an odds ratio equal to 1 under the reported logistic regression framework.
The registry does not provide the corresponding percentages of participants achieving the target in each treatment group in the posted statistical analyses, so the odds ratios should not be converted into absolute risk differences from the reported information alone.
11. Secondary Results: Fasting Serum Glucose
Change in fasting serum glucose from baseline to Week 36 was also analyzed using mixed-model analysis.
| Comparison | Mean difference | 95% CI | P-value |
|---|---|---|---|
| Dulaglutide 1.5 mg vs 3 mg | −3.7 mg/dL | −7.8 to 0.5 mg/dL | 0.084 |
| Dulaglutide 1.5 mg vs 4.5 mg | −8.1 mg/dL | −12.3 to −3.9 mg/dL | <0.001 |
The estimated mean difference in fasting serum glucose was −3.7 mg/dL for the 1.5 mg versus 3 mg comparison and −8.1 mg/dL for the 1.5 mg versus 4.5 mg comparison.
The confidence interval for the 1.5 mg versus 3 mg comparison spans zero, from −7.8 to 0.5 mg/dL, and the P-value is 0.084. Under the reported superiority framework, this does not provide the same statistical evidence of a treatment difference as the corresponding 4.5 mg comparison.
For the 1.5 mg versus 4.5 mg comparison, the 95% confidence interval is −12.3 to −3.9 mg/dL and the P-value is <0.001. The interval excludes zero, providing statistical evidence of a difference under the reported model.
The contrast between these two FSG results illustrates why the estimated effect, confidence interval, and P-value should be considered together rather than interpreting a P-value in isolation.
12. Summary of Posted Statistical Results
| Endpoint | Comparison | Effect | 95% CI | P-value |
|---|---|---|---|---|
| HbA1c change | 1.5 mg vs 3 mg | Mean difference −0.17 | −0.29 to −0.06 | 0.003 |
| HbA1c change | 1.5 mg vs 4.5 mg | Mean difference −0.34 | −0.45 to −0.22 | <0.001 |
| Body weight change | 1.5 mg vs 3 mg | Mean difference −0.9 kg | −1.4 to −0.4 | 0.001 |
| Body weight change | 1.5 mg vs 4.5 mg | Mean difference −1.6 kg | −2.1 to −1.1 | <0.001 |
| HbA1c <7.0% | 1.5 mg vs 3 mg | Odds ratio 1.49 | 1.12 to 1.98 | 0.006 |
| HbA1c <7.0% | 1.5 mg vs 4.5 mg | Odds ratio 2.23 | 1.65 to 3.01 | <0.001 |
| FSG change | 1.5 mg vs 3 mg | Mean difference −3.7 mg/dL | −7.8 to 0.5 | 0.084 |
| FSG change | 1.5 mg vs 4.5 mg | Mean difference −8.1 mg/dL | −12.3 to −3.9 | <0.001 |
13. Statistical Methods Explained
Why was a mixed-effects model used for HbA1c?
HbA1c is a longitudinal measurement, and the registered analysis uses a mixed-effects model with repeated measures. The model incorporates baseline, treatment, time, pooled country, and treatment-by-time effects. This structure allows the analysis to compare treatment groups while accounting for the repeated measurement framework.
What does a mean difference of −0.34 mean?
A mean difference of −0.34 represents a 0.34 percentage-point difference in the modeled change in HbA1c between the two treatment groups in the direction represented by the reported contrast. It is a difference in mean change, not a percentage reduction in HbA1c and not a probability.
What does an odds ratio of 2.23 mean?
An odds ratio of 2.23 means that the estimated odds of achieving an HbA1c below 7.0% were 2.23 times the comparison odds under the logistic regression model. The odds ratio does not directly state the difference in the percentage of participants achieving the target.
Why is the confidence interval important?
A point estimate alone does not communicate statistical precision. The 95% confidence interval provides an interval describing uncertainty around the estimated treatment effect under the model and sampling framework. For the primary HbA1c comparison with 4.5 mg, the interval is −0.45 to −0.22, whereas for the 3 mg comparison it is −0.29 to −0.06.
Why should the P-value not be interpreted as effect size?
A P-value addresses evidence against a specified null hypothesis under the statistical model. It is affected by the magnitude of the effect, the variability of the data, and the amount of information available. The size and precision of the treatment effect are better described by the effect estimate and confidence interval.
What does the treatment-by-time term do?
The treatment-by-time term allows the model to represent different outcome trajectories across treatment groups. In a repeated-measures setting, this is important because the treatment contrast need not be assumed to be identical at every measured time point.
Why does the analysis population matter?
The posted primary HbA1c analyses include participants who received at least one dose and had baseline and post-baseline HbA1c measurements, with post-discontinuation and post-new-antihyperglycemic-medication data excluded. The treatment effect therefore describes the prespecified analyzed population and data handling specified in the registry, rather than an unrestricted comparison of every enrolled participant at every time point.
14. Primary Analysis Population and Data Handling
The primary HbA1c analysis population was defined as all participants who received at least one dose of study drug and had a baseline and at least one post-baseline HbA1c value. The registry specifies exclusion of data after treatment discontinuation and/or initiation of new antihyperglycemic medications.
At least one dose
Participants had to receive at least one dose of study drug to enter the posted analysis population.
Baseline measurement
A baseline HbA1c measurement was required for the primary analysis population.
Post-baseline measurement
At least one post-baseline HbA1c value was required.
Post-discontinuation data
The registry states that data after treatment discontinuation were excluded from the primary HbA1c analysis.
This population definition is important when interpreting the results. The analysis does not simply compare the raw Week 36 means among all 1,842 enrolled participants. It applies the specified eligibility and data-handling rules and then uses the longitudinal model to estimate treatment differences.
15. Multiplicity and Multiple Comparisons
The registry reports two primary statistical analyses for the same registered primary endpoint: dulaglutide 1.5 mg versus 3 mg and dulaglutide 1.5 mg versus 4.5 mg. It also reports multiple secondary endpoint comparisons involving body weight, HbA1c target achievement, and fasting serum glucose.
| Analysis family | Posted comparisons | Interpretive issue |
|---|---|---|
| Primary HbA1c endpoint | 1.5 mg vs 3 mg; 1.5 mg vs 4.5 mg | Two superiority comparisons address the same registered primary endpoint. |
| Body weight | 1.5 mg vs 3 mg; 1.5 mg vs 4.5 mg | Secondary endpoint with two posted comparisons. |
| HbA1c target <7.0% | 1.5 mg vs 3 mg; 1.5 mg vs 4.5 mg | Binary secondary endpoint analyzed with logistic regression. |
| FSG | 1.5 mg vs 3 mg; 1.5 mg vs 4.5 mg | Secondary longitudinal endpoint analyzed with mixed models. |
Multiplicity matters because the probability of observing at least one apparently positive result can increase as the number of statistical tests increases. The registry identifies the analyses and their P-values, but the ClinicalTrials.gov record does not report an alpha-allocation or multiplicity-adjustment procedure for the complete family of posted comparisons.
16. Safety Results
The registry reports serious adverse events by treatment arm. These counts are presented as affected participants divided by participants at risk.
| Treatment arm | Serious adverse events | At risk |
|---|---|---|
| Dulaglutide 1.5 mg | 53 | 612 |
| Dulaglutide 3 mg | 45 | 616 |
| Dulaglutide 4.5 mg | 38 | 614 |
The reported serious-adverse-event counts provide an arm-level safety description. They should not be interpreted as a formal between-group safety hypothesis test because the posted statistical analyses in the ClinicalTrials.gov record do not include a comparative inferential analysis of serious adverse events.
17. What the Mean Difference Does — and Does Not — Mean
A mean difference of −0.34 for HbA1c change represents the difference between the modeled mean changes for the two treatment groups in the reported contrast. The negative sign indicates the direction of the contrast; it is not itself a measure of statistical significance.
The estimate does not mean that every participant's HbA1c decreased by 0.34 percentage points. Individual changes can vary substantially around a treatment-group mean.
The 95% confidence interval of −0.45 to −0.22 communicates uncertainty around the estimated mean difference. It does not describe the distribution of individual participant responses.
The P-value of <0.001 indicates statistical evidence against the relevant null hypothesis under the specified analysis. It does not tell the reader whether the estimated difference is large or small, nor does it communicate the range of plausible treatment effects. Those questions are addressed by the estimate and confidence interval.
18. Interpreting the Odds Ratio Correctly
The HbA1c target endpoint is categorical: participants either achieved an HbA1c below 7.0% at Week 36 or they did not. Logistic regression models the log-odds of that binary outcome.
An odds ratio compares odds, not probabilities directly. The same odds ratio can correspond to different absolute probability differences depending on the baseline probability.
For the 1.5 mg versus 4.5 mg comparison, the posted odds ratio is 2.23 with a 95% confidence interval of 1.65 to 3.01. This indicates higher estimated odds of achieving the HbA1c target in the treatment group represented by the numerator of the reported comparison.
Because the registry does not provide the underlying group-specific percentages for this statistical analysis, the odds ratio cannot be translated into an absolute percentage-point difference from the posted result alone.
19. Statistical Interpretation of the FSG Results
1.5 mg vs 3 mg
Mean difference −3.7 mg/dL; 95% CI −7.8 to 0.5 mg/dL; P = 0.084.
1.5 mg vs 4.5 mg
Mean difference −8.1 mg/dL; 95% CI −12.3 to −3.9 mg/dL; P < 0.001.
These two analyses illustrate an important statistical principle: a larger estimated effect does not automatically determine statistical significance, and a smaller P-value does not by itself establish a larger effect. The first comparison has an interval that includes zero, whereas the second has an interval entirely below zero.
The width of the confidence interval also matters. The interval for the 1.5 mg versus 3 mg comparison extends from −7.8 to 0.5 mg/dL, while the interval for the 1.5 mg versus 4.5 mg comparison extends from −12.3 to −3.9 mg/dL. These intervals communicate both direction and precision.
20. Limitations
- Limited baseline detail: The ClinicalTrials.gov record summarized here does not report a detailed baseline demographic table, so treatment-group balance cannot be assessed from the posted information here.
- Limited dose-response characterization: The posted analyses provide pairwise comparisons but do not provide a formal dose-response trend model or a continuous dose-effect estimate.
- Multiplicity: Multiple primary and secondary comparisons are reported. The complete multiplicity-control strategy is not specified in the posted information summarized here.
- Analysis population: The primary HbA1c analyses exclude post-discontinuation and post-new-antihyperglycemic-medication data and require baseline and post-baseline measurements. Interpretation therefore depends on the prespecified analysis population.
- Model assumptions: Mixed-effects longitudinal models rely on their specified covariance and distributional framework. The posted registry information does not provide a detailed model-diagnostic assessment.
- Binary endpoint interpretation: Odds ratios for HbA1c target achievement should not be interpreted as risk ratios or absolute probability differences.
- Safety inference: Serious adverse-event counts are reported by arm, but the posted information summarized here does not provide a formal comparative safety hypothesis test for these events.
- Generalizability: The registry information available here identifies the condition and trial design but does not provide a detailed baseline description from which broader generalizability can be assessed.
21. Why This Trial Matters Statistically
AWARD-11 is a useful teaching case because it combines a three-arm randomized design with longitudinal continuous outcomes and binary responder outcomes. The same trial therefore demonstrates why the statistical method must follow the structure of the endpoint rather than applying a single analysis to every outcome.
| Concept | How it appears in AWARD-11 |
|---|---|
| Randomization | The study uses randomized allocation across three treatment arms. |
| Blinding | The registry describes the study as double-blind. |
| Repeated measures | HbA1c, body weight, and fasting serum glucose are evaluated from baseline to Week 36. |
| Mixed-effects model | The primary HbA1c endpoint and continuous secondary outcomes use mixed-model analysis. |
| Least-squares mean | The primary HbA1c endpoint uses a model-based least-squares mean framework. |
| Logistic regression | Achievement of HbA1c <7.0% is analyzed as a binary outcome. |
| Mean difference | HbA1c, body weight, and FSG treatment effects are reported as mean differences. |
| Odds ratio | HbA1c target achievement is reported using odds ratios. |
| Confidence interval | All posted comparative estimates include two-sided 95% confidence intervals. |
| P-values | Each posted statistical comparison includes a P-value. |
22. Related Tutorials
Learn more about the methods used in this trial:
23. Related Calculators
24. Sources
- ClinicalTrials.gov: NCT03495102, AWARD-11.
- PubMed: PMID 36894938.
- PubMed: PMID 35546790.
- PubMed: PMID 33397768.
Continue through Clinical Biostats
Explore the statistical methods represented in this trial through deeper tutorials and practical statistical tools.
25. Record Summary
AWARD-11 provides a clear example of how statistical methodology changes with endpoint structure. The primary HbA1c endpoint is a longitudinal continuous outcome analyzed with a mixed-effects model with repeated measures, while achievement of an HbA1c target below 7.0% is a binary outcome analyzed with logistic regression. Body weight and fasting serum glucose are also analyzed with mixed-model methods.
The primary HbA1c comparisons report mean differences of −0.17 for dulaglutide 1.5 mg versus 3 mg and −0.34 for dulaglutide 1.5 mg versus 4.5 mg, with corresponding two-sided 95% confidence intervals of −0.29 to −0.06 and −0.45 to −0.22. The posted secondary analyses extend the statistical picture to body weight, HbA1c target achievement, and fasting serum glucose.
The most useful interpretation combines the effect estimate, confidence interval, and P-value while keeping the endpoint definition and analysis population in view. The different results across HbA1c, body weight, HbA1c target achievement, and fasting serum glucose also demonstrate why a clinical trial cannot be reduced to a single P-value or a single summary measure.