A Study in Participants With Type 2 Diabetes Mellitus (AWARD-2): Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for A Study in Participants With Type 2 Diabetes Mellitus (AWARD-2) (NCT01075282), a randomized phase 3 study. The primary outcome is Change From Baseline to 52 Weeks Endpoint in Glycosylated Hemoglobin (HbA1c).

2. Clinical Question

3. Trial Design

AWARD-2 is a phase 3, randomized, open-label, parallel-arm trial listed as completed. It compared LY2189265 at 1.5 mg and 0.75 mg with insulin glargine in participants with type 2 diabetes mellitus. Participants were also taking metformin and glimepiride. The registry record lists an enrollment of 810, a start date of 2010-02, and a primary completion date of 2012-05.

The LY2189265 regimens were subcutaneous injections once weekly, while insulin glargine was administered subcutaneously once daily and titrated using blood glucose measures. All three treatment descriptions specify continued metformin at at least 1500 mg/day and glimepiride at at least 4 mg/day for 78 weeks. The record describes no masking, so the study was open-label; allocation was randomized and the model was parallel-arm.

The primary endpoint was change in HbA1c from baseline to the 52-week endpoint. Secondary outcomes included HbA1c changes at other listed time points, the number of participants reaching specified HbA1c thresholds, self-monitored blood glucose profiles, measures of beta-cell function and insulin sensitivity, glucagon concentration, body weight, and body mass index. The registry lists 26 outcome measures, while 25 secondary outcomes are posted.

The reported primary analysis used an analysis of covariance (ANCOVA) and included country, treatment, and baseline HbA1c as specified in the statistical description. The record does not report the original sample-size justification, the noninferiority margin, or a full protocol-level statistical analysis plan. Those omissions limit how completely the design rationale and analysis decisions can be assessed from 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.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: Participants who were randomized and received at least one dose of LY2189265 or Insulin Glargine with evaluable HbA1c data. Only pre-rescue measurements were used. Last observation carried forward (LOCF) was used to impute missing postbaseline values. If there were no data after the date of randomization, the endpoint was considered missing.

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.

Population: Participants who were randomized and received at least 1 dose of LY2189265 or Insulin Glargine.

6. Primary Endpoint

The registry lists the primary outcome as Change From Baseline to 52 Weeks Endpoint in Glycosylated Hemoglobin (HbA1c), measured over Baseline, 52 weeks.

The primary endpoint is change from baseline to the 52-week endpoint in glycosylated hemoglobin (HbA1c), expressed as a percentage of glycosylated hemoglobin. It describes how HbA1c changed over the stated interval in each treatment group. HbA1c is used in the trial as a measure of glycemic control; the registry also lists other measures of glucose control and metabolic status as secondary outcomes.

The posted arm estimates are least-squares means from an ANCOVA, rather than simple raw averages. The model description specifies country and treatment as fixed effects and baseline HbA1c as a covariate. Adjusting for baseline and country allows the reported estimates to reflect the model’s treatment comparisons while accounting for those specified factors. The posted standard error for each arm’s estimate is 0.06.

Change in HbA1c is useful for comparing the groups’ glycemic measurements over the trial interval, but it is not by itself a complete description of participants’ health or experience. It does not summarize safety, every secondary outcome, or the longer-term course beyond the reported endpoint. The primary endpoint should therefore be read alongside the trial’s other reported outcomes and the limits of the registry record.

7. Results Posted in the Registry

• LY2189265 0.75 mg: mean -0.76 versus -1.08 on control; difference 0.32 (95% CI 0.15 to 0.49), standardized effect d = 0.33; z-test p = <0.001.
• Insulin Glargine: mean -0.63 versus -1.08 on control; difference 0.45 (95% CI 0.28 to 0.62), standardized effect d = 0.46; z-test p = <0.001.

8. Statistical Interpretation

9. Statistical Methods Explained

The primary outcome is continuous: it is a numerical measurement of change in HbA1c. The posted arm results are least-squares means, which are model-adjusted estimates rather than unadjusted averages. Their reported standard errors describe the uncertainty of each arm estimate under the fitted model. The registry states that ANCOVA was used, with country and treatment as fixed effects and baseline HbA1c as a covariate.

What does ANCOVA contribute?

General explanation: analysis of covariance estimates group differences while accounting for selected covariates. A baseline measurement is often included because the outcome at follow-up can be related to the starting value. The model specified for AWARD-2 also includes country and treatment. The resulting adjusted means and contrasts are conditional on the model and analysis population; they are not the same quantity as a simple difference between unadjusted averages.

The reported population description says participants were randomized, received at least one dose of LY2189265 or insulin glargine, and had evaluable HbA1c data. It also says that only pre-rescue measurements were used and that last observation carried forward (LOCF) was used to impute missing postbaseline values. The registry text describing the population is truncated after an additional statement about absent postbaseline data. The remaining handling details are not reported in the registry record.

How should the confidence intervals be understood?

General explanation: a confidence interval gives a range of values compatible with the data and the assumptions of the analysis, at the stated confidence level. A narrower interval indicates greater precision under those assumptions; a wider interval indicates less precision. The sponsor’s primary comparisons are reported with 95% confidence intervals. The interval for the 1.5 mg comparison with insulin glargine is entirely negative, while the interval for the 0.75 mg comparison extends from a negative value to a positive value and includes zero.

A confidence interval is not a guarantee that the true effect lies in the interval, and it does not address limitations such as open-label treatment, the analysis population, or missing-data assumptions. It also does not identify whether a difference is clinically important. No threshold for clinical importance is reported in the registry record.

What do the p-values answer?

General explanation: a p-value describes how compatible the observed data are with a specified statistical null hypothesis, under the model and its assumptions. It is not the probability that the null hypothesis is true, and it does not measure the size or importance of an effect. The registry reports p-values for analyses classified as noninferiority or equivalence and for analyses classified as superiority or other. Because these are different analysis categories, their p-values address different questions. The noninferiority margin and the full decision rules are not reported in the registry record.

The sponsor’s ANCOVA estimates and the unadjusted re-calculations are not identical analytical procedures. The re-calculation uses posted summary estimates and approximate pooled variability to produce simple contrasts and intervals. It does not reproduce the full ANCOVA, its covariate adjustment, or every feature of the original analysis. The recalculated comparison against insulin glargine is expressed in the reverse direction from the sponsor’s reported least-squares mean difference. The direction of a contrast depends on which group is subtracted from which; the sign alone should not be interpreted without identifying that order.

Why does over-dispersion matter?

General explanation: over-dispersion arises when observed variability is greater than a statistical model expects. If a model assumes less variation than is present, standard errors can be too small and confidence intervals too narrow. This primary endpoint is continuous, and the posted model reports standard errors for adjusted means. No separate over-dispersion estimate or adjustment is reported in the registry record; this concept should not be mistaken for a documented feature of the AWARD-2 analysis.

The arm-level standard errors and adjusted contrasts are the most direct registry results for the primary endpoint. The planning illustrations use a normal approximation and an estimated effect derived from posted results; they are design scenarios, not a reconstruction of the trial’s original design calculation. The original sample-size assumptions and justification are not reported in the registry record.

10. Confidence Intervals and Precision

The sponsor-reported 95% confidence interval for the adjusted comparison of LY2189265 1.5 mg with insulin glargine runs from -0.60 to -0.29 around a least-squares mean difference of -0.45. This interval does not include zero, indicating that the reported adjusted estimate is separated from zero under the stated analysis. It does not by itself show whether the difference is clinically important, and the registry does not report a clinical-importance threshold.

For LY2189265 0.75 mg compared with insulin glargine, the reported estimate is -0.13 and the 95% confidence interval runs from -0.29 to 0.02. Because this interval includes zero, the posted interval is compatible with a range of contrasts on either side of zero. The interval is also consistent with greater uncertainty about the direction of the adjusted difference than the 1.5 mg comparison interval. The record’s p-values differ by analysis classification, so the confidence interval and each p-value should be read in the context of the specific question being tested.

The posted standard error for each arm’s least-squares mean change is 0.06. A standard error describes the precision of an estimate, not the spread of individual participants’ measurements. The arm estimates and comparison intervals are model-based, and their interpretation depends on the analysis population, covariates, and missing-data handling. The registry’s description of the population is incomplete, and the full missing-data rules are not reported in the registry record.

The recalculated intervals are based on simpler unadjusted summary-statistic methods and should not be substituted for the sponsor’s adjusted confidence intervals. The record flags a discrepancy, and the sponsor’s own analysis takes precedence. Statistical precision also does not resolve the unreported noninferiority margin or establish clinical importance.

11. Planning a Similar Trial: Sample Size

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

The planning figures are participants per arm, not total events. They illustrate sample sizes under a normal approximation, equal allocation between two arms, and a two-sided significance level of 0.05. Each scenario uses a stated fraction of the observed effect as the effect to detect. These are planning illustrations based on the observed effect, not observed power, and they are not the trial’s original sample-size justification.

The 80% power column gives the participants per arm implied by the stated assumptions for an 80% chance of detecting the chosen effect if that effect is present. The 90% power column gives the corresponding planning number for a 90% chance under those same assumptions. Power is a property of a design scenario and its assumptions; it is not a measure of how likely the trial result is to be true.

For the 0.75 mg versus 1.5 mg re-calculated comparison, planning for 100% of the observed effect gives 147 participants per arm for 80% power and 196 for 90% power. Planning for 75% of the observed effect gives 261 and 349 participants per arm, respectively. Planning for 50% of the observed effect gives 586 and 784 participants per arm, respectively.

For the insulin glargine versus 1.5 mg re-calculated comparison, planning for 100% of the observed effect gives 73 participants per arm for 80% power and 98 for 90% power. Planning for 75% gives 130 and 174 participants per arm, respectively. Planning for 50% gives 292 and 391 participants per arm, respectively. The pattern reflects a general statistical principle: smaller effects require more participants to detect under otherwise unchanged assumptions. The registry’s original design assumptions are not reported in the registry record.

12. How the Sample Size Is Derived

Power is the probability that a study of a given size rejects the null hypothesis when a specific true effect exists. It depends on four things: the effect to detect, the variability of the outcome, the significance level (here a two-sided α of 0.05, so z_1−α/2 = 1.960) and the desired power (z_1−β = 0.842 for 80% and 1.282 for 90%).

Formula

n is the number of participants in each arm, Δ is the difference in means to detect and σ is the common standard deviation.

Worked calculation for this trial

1. Observed difference Δ = 0.320; pooled standard deviation σ = 0.977.
2. n = 2 × (1.96 + 0.842)² × 0.977² / 0.320² = 146.3, rounded up to 147 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 <- 0.32                      # observed difference in means
sigma <- 0.976757157872183                     # 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(1075282)
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 = 147 (80% power), 196 (90% power)
 75% of observed effect: n per arm = 261 (80% power), 349 (90% power)
 50% of observed effect: n per arm = 586 (80% power), 784 (90% power)
power.t.test: 147.2 
Simulated power at n = 147 per arm: 0.796 (20000 replicates)

The simulation draws many sets of group means and variances at the planned sample size and counts how often the two-sample t-test rejects the null. It should land close to the target power; at very small required sample sizes it can fall below it, because the normal approximation is rough there.

13. Analyses Reported by the Sponsor

The registry reports adjusted ANCOVA comparisons from the sponsor’s analysis. For LY2189265 1.5 mg versus insulin glargine, the least-squares mean difference is -0.45, with a 95% confidence interval from -0.60 to -0.29 and p < 0.001. For LY2189265 0.75 mg versus insulin glargine, the difference is -0.13, with a 95% confidence interval from -0.29 to 0.02. The corresponding posted p-values are < 0.001 for the noninferiority-or-equivalence classification and 0.05 for the superiority-or-other classification.

The re-calculation is a simple unadjusted analysis based on posted summary statistics, rather than a reproduction of the sponsor’s covariate-adjusted ANCOVA. Its insulin glargine comparison with the 1.5 mg arm is +0.45 because its subtraction order is reversed from the sponsor’s -0.45 contrast. The re-calculated 0.75 mg comparison is against the 1.5 mg arm, so it is not the direct 0.75 mg versus insulin glargine contrast reported by the sponsor.

The record flags a discrepancy between the posted analysis and the re-calculation. The sponsor’s own analysis takes precedence for interpreting the trial’s reported result. The simplified estimates can serve as a separate check on the posted arm summaries, but they do not replace the adjusted results or clarify the unreported noninferiority margin.

14. Multiplicity, Missing Data and Analysis Populations

The registry lists multiple outcomes and analyses. Secondary measures include HbA1c changes at other time points, counts reaching HbA1c thresholds, self-monitored blood glucose, HOMA2 measures, glucagon, body weight, and body mass index. It lists 26 outcome measures and 36 analyses, with 25 secondary outcomes posted. Testing multiple outcomes or contrasts can increase the chance of at least one apparently notable finding by chance if no adjustment is made.

The registry record does not report whether or how multiplicity was controlled across outcomes, time points, doses, and analysis categories. It also does not report a hierarchy of hypotheses or a complete statistical analysis plan. Therefore, the posted p-values should be interpreted as results for their specified analyses, not as proof that the family of outcome tests was controlled at a particular error rate.

Missing-data handling is partly described. The primary analysis used last observation carried forward for missing postbaseline values, and only measurements before rescue were used. The population text ends mid-sentence while describing circumstances with no postbaseline data. The additional rule is not reported in the registry record. The record also does not provide enough detail to assess the consequences of LOCF, the amount or pattern of missing data, or the sensitivity of conclusions to alternative assumptions.

The primary analysis population is described as randomized participants who received at least one dose and had evaluable HbA1c data. The full population definition is incomplete in the registry text. The registry record does not report a full accounting of participants included in each analysis or a complete account of exclusions from the primary analysis. These details matter because an analysis restricted to participants with evaluable measurements can differ from one retaining all randomized participants under other missing-data assumptions.

The reported noninferiority-or-equivalence analyses are especially difficult to interpret without the margin and the prespecified decision rules. A p-value alone does not identify the margin or establish how the analysis was planned. Those design and analysis details are not reported in the registry record.

15. Safety Results

The registry reports serious adverse events as affected participants divided by participants at risk.

Other (non-serious) adverse events

Participants with at least one non-serious event at or above the registry's 3% reporting threshold.

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: overdose (1.1%), hyperglycaemia (1.1%), maternal exposure during pregnancy (0.8%), uterine leiomyoma (0.8%), coronary artery disease (0.8%).

The serious adverse event figures give the number of participants at risk and the number affected in each group. For LY2189265 1.5 mg, 32 of 273 participants were affected; for LY2189265 0.75 mg, 28 of 272 were affected; and for insulin glargine, 33 of 262 were affected. These are descriptive counts from the posted safety results. They do not establish that a treatment caused an event or allow conclusions about comparative safety on their own.

The separate other-adverse-event figures report 199 of 273 participants for LY2189265 1.5 mg, 190 of 272 for LY2189265 0.75 mg, and 187 of 262 for insulin glargine. The registry lists serious-event terms including overdose and hyperglycaemia, each at 1.1%, and maternal exposure during pregnancy, uterine leiomyoma, coronary artery disease, and cholelithiasis, each at 0.8%. These term percentages are presented as posted; the record does not establish a causal relationship.

Deaths are not reported in the registry record. The safety denominators differ from the arm counts listed for the primary efficacy analysis, so the figures should be understood using their stated safety denominators. Counts alone do not describe event timing, duration, severity beyond the serious classification, or the full clinical context.

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

The posted primary results describe changes in HbA1c over the baseline-to-52-week interval in three randomized treatment groups. The adjusted mean change was -1.08 for LY2189265 1.5 mg, -0.76 for LY2189265 0.75 mg, and -0.63 for insulin glargine. The sponsor’s adjusted comparison for 1.5 mg versus insulin glargine was -0.45, with a 95% confidence interval from -0.60 to -0.29. The adjusted comparison for 0.75 mg versus insulin glargine was -0.13, with a 95% confidence interval from -0.29 to 0.02.

These estimates address the trial’s specified HbA1c endpoint and analysis. They do not show whether the observed differences matter to participants, because no clinical-importance threshold is reported in the registry record. The registry labels analyses as noninferiority or equivalence, but the noninferiority margin and complete decision rules are not reported. The result classifications therefore cannot be fully assessed from the registry record alone.

The findings also do not describe every outcome assessed. The registry lists secondary measures that include HbA1c thresholds, blood glucose profiles, body weight, and other metabolic outcomes. Nor do HbA1c estimates alone resolve questions about adverse events, missing data, longer-term outcomes, or how findings apply beyond the trial population and treatment regimens.

The study was open-label, and the primary analysis used a stated approach to rescue measurements and missing postbaseline values. These are relevant features when interpreting the estimates. The adjusted sponsor analysis is the primary basis for reading the reported comparison; the unadjusted re-calculation is a simpler analysis and is not a substitute. No medical advice or treatment recommendation follows from these statistical results alone.

17. Why This Trial Matters Statistically

AWARD-2 illustrates why a trial result should be read through its estimand, model, and contrast definition rather than by comparing arm estimates in isolation. Its primary outcome is a continuous change measure, and the registry reports model-adjusted least-squares means from an ANCOVA that includes baseline HbA1c and country. The sponsor’s adjusted contrasts answer a different statistical question from simple unadjusted contrasts reconstructed from arm summaries.

The comparison direction matters: the sponsor’s 1.5 mg versus insulin glargine estimate is -0.45, while the unadjusted recalculation is +0.45 because its subtraction order is reversed. The size is similar, but the signs describe opposite contrast orientations. For the lower dose, the re-calculated contrast is against the higher dose, whereas the sponsor reports a direct contrast with insulin glargine. Recognizing those distinctions prevents unlike comparisons from being treated as interchangeable.

The record also illustrates the importance of separating noninferiority from superiority questions. The posted analysis categories and p-values do not replace the noninferiority margin or its decision rules. Without those details, the noninferiority label cannot be fully interpreted from the registry alone.

Finally, planning calculations based on observed effects are not observed power and do not recover a trial’s original sample-size rationale. Likewise, confidence intervals convey precision under an analysis, not clinical importance or freedom from bias. These distinctions apply broadly when clinical researchers interpret registry results alongside the original analysis plan and study report.

18. Limitations

• The posted summary statistics are the only data these results rest on; patient-level data, full protocols and statistical analysis plans are not available.
• Recalculated results are simple, unadjusted calculations from posted summaries. They do not replace the sponsor's prespecified analysis.

The record is open-label, with no masking reported. Knowledge of assigned treatment can affect participant or investigator behavior and may influence outcomes that involve measurement or reporting. HbA1c is a laboratory measure, but the broader trial outcomes and safety reporting can still be affected by an open-label design.

The registry does not report the noninferiority margin, the original sample-size justification, or a complete statistical analysis plan. It also does not report whether multiplicity was controlled across the many listed outcomes and analyses. These gaps limit assessment of the prespecified hypotheses and the interpretation of the noninferiority-or-equivalence classifications.

The primary analysis used pre-rescue measurements and LOCF for missing postbaseline values. The registry’s description of the analysis population is truncated, and the amount and pattern of missing data are not reported in the record. It is therefore not possible to assess from this record how sensitive the estimates are to the missing-data method or to the population definition.

The unadjusted re-calculations rely on posted summary statistics and approximate variability. They do not reproduce the sponsor’s ANCOVA and are not the primary trial analysis. The registry flags a discrepancy, so the sponsor’s own analysis takes precedence. Safety results are reported as counts and selected event-term percentages; deaths are not reported in the registry record. These limits mean the posted results should be interpreted as a partial account of the trial rather than a complete evaluation of all efficacy and safety questions.

19. Related Tutorials

20. Related Statistical Calculators

21. Sources

• ClinicalTrials.gov: A Study in Participants With Type 2 Diabetes Mellitus (AWARD-2) — NCT01075282.
• Registry record last updated 2015-01-16; status last verified 2015-01.
• Results first posted 2015-01-16.
• Data retrieved from ClinicalTrials.gov on 2026-11-09.
• Group sizes, means, event counts and sponsor-reported p-values are reproduced as posted. Effect estimates, confidence intervals and p-values not attributed to the sponsor are independently calculated from the posted summary statistics.

22. Record Summary

AWARD-2 is a completed, phase 3, randomized, open-label, parallel-arm comparison of LY2189265 at 1.5 mg or 0.75 mg with insulin glargine in participants with type 2 diabetes taking metformin and glimepiride. The registry lists enrollment of 810 and treatment descriptions extending for 78 weeks.

The primary endpoint was change in HbA1c from baseline to the 52-week endpoint. The sponsor’s ANCOVA reported an adjusted difference of -0.45 for LY2189265 1.5 mg versus insulin glargine and -0.13 for LY2189265 0.75 mg versus insulin glargine. The registry also lists secondary outcomes and safety results.

The record does not report the noninferiority margin, original sample-size justification, or complete statistical plan. The sponsor’s adjusted analysis takes precedence over simplified unadjusted re-calculations.