Research Study Comparing a New Medicine Semaglutide to Liraglutide in People With Type 2 Diabetes: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Research Study Comparing a New Medicine Semaglutide to Liraglutide in People With Type 2 Diabetes (NCT03191396), a randomized phase 3 study. The primary outcome is Change in HbA1c.

2. Clinical Question

The registry record describes the study as a comparison of blood sugar levels after 30 weeks of treatment. HbA1c is the primary outcome, expressed as a percentage of glycosylated haemoglobin. The statistical question is about the average change in this continuous measure, rather than the proportion of participants who crossed a clinical threshold or experienced a particular event.

The primary outcome was evaluated during the on-treatment without rescue medication period. In that definition, participants contributed while they were considered treated with the trial product and had not initiated rescue medication. This makes the analysis period narrower than a simple comparison of every participant’s outcome regardless of treatment status or rescue medication use.

The registry record reports both a non-inferiority analysis and a superiority analysis for the primary outcome. It does not report the non-inferiority margin in the record, so the margin and the criteria used to judge non-inferiority cannot be assessed from the information recorded there. The sponsor’s reported ANCOVA estimate, confidence interval and p-value are the trial’s own inferential results; the simple comparison of posted group means is a separate, unadjusted calculation.

3. Trial Design

This was a phase 3, randomized, parallel-group study with no masking reported. Participants were assigned to semaglutide 1.0 mg once weekly or liraglutide 1.2 mg once daily, each as add-on treatment to 1-3 oral anti-diabetic drugs. The registry record describes the study as conducted in Europe. The recorded enrollment was 577.

The study was listed as completed. Its recorded start date was 2017-06-27 and its primary completion date was 2018-07-09. The brief study description says participation would last approximately 9 months (37 weeks), with 7 clinic visits and 3 phone calls with the study doctor. It describes general health checks, blood samples, heart and eye checks, and forms about health and satisfaction with diabetes treatment.

The primary outcome compared mean HbA1c change from week 0 to week 30. The primary population was the full analysis set, which the record defines as all randomized participants. The outcome description also specifies an on-treatment period without rescue medication, and a method for imputing missing primary-outcome data. Those analysis details matter because the recorded group summaries do not represent an unrestricted follow-up comparison across every possible treatment or rescue-medication experience.

Random allocation supports a group comparison by reducing systematic differences at the point of assignment, although the record does not describe the allocation procedure in further detail. Because masking is listed as none, the study was open-label. The registry record does not report the original sample-size justification, the allocation ratio in the formal design description, or further details of the randomization procedure. It also does not report interim analyses. These omissions limit how fully the design choices can be reconstructed from the registry record alone.

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: After screening, subjects were required to continue their oral anti-diabetic drug (OAD) pre-trial background medication (i.e. metformin, Sodium-glucose co-transporter-2 (SGLT-2) inhibitors, sulphonyl ureas, or combinations of these) throughout the entire trial.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: Full analysis set (FAS), which included all randomised participants.

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, measured over Week 0, week 30.

The primary endpoint is mean change in HbA1c from baseline, defined as week 0, to week 30. HbA1c is glycosylated haemoglobin and is recorded as a percentage. It is used as a continuous measure of blood sugar control over time, allowing the analysis to compare average changes between randomized groups rather than reducing each participant’s result to a category.

A negative mean change indicates that the average week-30 measurement was lower than the baseline measurement. The posted group means were -1.7 for semaglutide 1.0 mg and -1.1 for liraglutide 1.2 mg. The corresponding standard deviations were 0.9 and 1.0. These group summaries describe the observed change and its variability; they do not, by themselves, account for the sponsor’s adjusted analysis or establish why an individual participant’s measurement changed.

The endpoint was defined for the on-treatment without rescue medication period: participants were considered treated with trial product and had not yet initiated rescue medication. Missing data were imputed using observed data from participants in the same group defined by randomized treatment. The model included the stratification factor as a categorical effect and used baseline and all previous visits as covariates. The registry record does not report further detail about the imputation model or the sensitivity analyses used to examine its assumptions.

7. Results Posted in the Registry

• Liraglutide 1.2 mg: mean -1.10 versus -1.70 on control; difference 0.60 (95% CI 0.44 to 0.76), standardized effect d = 0.63; 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 Body Weight (kg)

• 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 deviations.
• Result:
  • Liraglutide 1.2 mg: mean -2.00 versus -5.80 on control; difference 3.80 (95% CI 3.08 to 4.52), standardized effect d = 0.86; z-test p = <0.001.
• Interpretation:
  • The 95% confidence interval excludes no difference, so the mean is statistically higher for Liraglutide 1.2 mg than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.

Change in Fasting Plasma Glucose (FPG)

• 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 deviations.
• Result:
  • Liraglutide 1.2 mg: mean -1.46 versus -2.65 on control; difference 1.19 (95% CI 0.81 to 1.57), standardized effect d = 0.52; z-test p = <0.001.
• Interpretation:
  • The 95% confidence interval excludes no difference, so the mean is statistically higher for Liraglutide 1.2 mg than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.

Change in Self-measured Plasma Glucose (SMPG), 7 Point Profile: Mean 7-point Profile

• 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 deviations.
• Result:
  • Liraglutide 1.2 mg: mean -2.10 versus -3.00 on control; difference 0.90 (95% CI 0.55 to 1.25), standardized effect d = 0.42; z-test p = <0.001.
• Interpretation:
  • The 95% confidence interval excludes no difference, so the mean is statistically higher for Liraglutide 1.2 mg than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.

The secondary outcome results give a broader picture of the measured changes during the study, while remaining subordinate to the primary HbA1c analysis. The registry record lists outcomes involving body weight, fasting plasma glucose, self-measured plasma glucose and fasting blood lipids. The posted numerical summaries for body weight, fasting plasma glucose and the mean 7-point self-measured plasma glucose profile show negative mean changes in both randomized groups. Their recalculated between-group contrasts are reported with unadjusted p-values of <0.001.

That broad pattern is directionally consistent with the primary outcome in the limited sense that both groups’ reported means decreased on the listed continuous measures, with larger decreases in the semaglutide group for the reported summaries. These secondary measures address different aspects of health and are recorded in different units, so their magnitudes should not be compared directly with the HbA1c estimate. Nor do they independently establish a broader treatment conclusion.

The record reports 40 secondary outcomes, but a complete inferential account requires more than observing which individual comparisons have small unadjusted p-values. When many outcomes are examined, the chance of at least one apparently notable result due to random variation can increase. The registry record does not report a multiplicity adjustment for the secondary comparisons. Consequently, unadjusted secondary results carry less inferential weight than the designated primary analysis, and should be interpreted as supportive or exploratory rather than as a set of independent confirmatory findings.

The registry record does not provide a sponsor-reported ANCOVA result for every secondary outcome; for fasting plasma glucose and the mean 7-point profile, the listed comparisons have no reported sponsor analysis in the record. This difference between summary-statistic recalculations and sponsor analyses is another reason to avoid treating all secondary comparisons as equally definitive. The outcome descriptions also do not establish that each listed measure was analyzed using the same model or assumptions.

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

• Change in Body Weight (kg)
• Change in Fasting Plasma Glucose (FPG)
• Change in Self-measured Plasma Glucose (SMPG), 7 Point Profile: Mean 7-point Profile
• Change in Self-measured Plasma Glucose (SMPG), 7 Point Profile: Mean Post Prandial Increment (Over All Meals)
• Change in Fasting Blood Lipids: Total Cholesterol
• Change in Fasting Blood Lipids: Low-density Lipoprotein (LDL)-Cholesterol
• Change in Fasting Blood Lipids: High-density Lipoprotein (HDL)-Cholesterol
• Change in Fasting Blood Lipids: Triglycerides

9. Statistical Interpretation

10. Statistical Methods Explained

The primary outcome is continuous: it is a measured quantity with a mean and standard deviation, rather than a count of events. The posted arm summaries give the mean change and its standard deviation for each randomized group. A mean difference expresses the separation between group averages, but its sign depends on which group is subtracted from which. The unadjusted recalculation is stated as liraglutide minus semaglutide, whereas the sponsor’s ANCOVA treatment difference is reported with the reverse sign relative to that contrast.

What does ANCOVA contribute?

ANCOVA, or analysis of covariance, estimates a group difference while accounting for covariates included in the model. In general, adjusting for baseline measurements or design factors can improve precision and can yield an estimate that differs from the raw comparison of posted means. The registry record identifies ANCOVA as the method for the sponsor’s primary analyses, but it does not fully describe the model specification, the covariates used in that analysis, or the coding of the treatment contrast. The endpoint description does state that the missing-data regression model included the stratification factor as a categorical effect and baseline and previous visits as covariates; that description should not be assumed to provide every detail of the sponsor’s ANCOVA model.

How should the unadjusted recalculation be read?

The recalculated mean difference and confidence interval use the posted group means and standard deviations to form a simple comparison. It is not the trial’s adjusted ANCOVA result. Its purpose is to describe the scale and precision of a straightforward summary-statistic contrast, not to replace the sponsor’s analysis. The recalculated p-value is also unadjusted. It should therefore be read as a separate statistical calculation rather than as the trial’s definitive inferential test.

What does a confidence interval express?

In general, a confidence interval describes the range of estimates compatible with the observed data under the specified statistical procedure. A narrower interval indicates greater precision for the estimated contrast under that procedure; a wider interval indicates less precision. The interval does not include all sources of uncertainty, such as uncertainty from missing-data assumptions or limitations in the analysis model. Nor does it quantify the probability that the true effect lies inside the particular interval.

Why can missing-data methods affect the result?

The primary outcome description says missing measurements were imputed using observed data from participants in the same group defined by randomized treatment. A regression model included the stratification factor as a categorical effect, with baseline and all previous visits as covariates. Imputation uses observed information to fill in missing outcome values for analysis; it does not turn those values into direct measurements. The validity of the result therefore depends in part on the assumptions behind the imputation model. The registry record does not report enough detail to assess those assumptions or identify sensitivity analyses for departures from them.

What do the p-values say?

In general, a p-value measures how incompatible the observed data are with a specified null model, assuming the model and analysis procedure are appropriate. A small p-value is not the probability that the null hypothesis is true, and it does not measure clinical importance. The sponsor’s primary p-value is reported as < 0.001. The unadjusted recalculation is reported as <0.001. These values arise from different analyses and should not be treated as interchangeable.

How do the non-inferiority and superiority labels fit?

Non-inferiority tests whether an intervention is not worse than a comparator by more than a prespecified margin. Superiority tests whether the data support a difference in favor of one group under the relevant model. The registry record labels the sponsor’s primary analyses as both non-inferiority and superiority and reports the same ANCOVA estimate, confidence interval and p-value for each. The non-inferiority margin, testing sequence and decision criteria are not reported in the registry record. Without those details, the labels can be described but their full inferential framework cannot be reconstructed.

More generally, trial interpretation depends on the endpoint definition, analysis population, model and handling of missing outcomes as well as the observed estimate. The randomized design gives a basis for comparing assigned groups, but the open-label status and the specific on-treatment analysis definition remain part of the context. The registry record does not report every detail of the statistical analysis plan, so the reported sponsor analysis is the most direct account of the trial’s inferential result, while the recalculation remains a transparent but simpler comparison.

11. Confidence Intervals and Precision

The unadjusted primary mean difference, expressed as liraglutide minus semaglutide, was 0.6 percentage points. Its 95% confidence interval was 0.445 to 0.755. This interval describes the precision of the simple summary-statistic comparison under its calculation method. Its endpoints are both on the positive side of zero, indicating that the recalculated contrast is estimated with a direction distinct from zero under that method.

The sponsor’s ANCOVA estimate was -0.69, with a 95% confidence interval from -0.82 to -0.56. The sign differs because the sponsor’s contrast is oriented oppositely to the stated unadjusted contrast; the ANCOVA estimate also comes from an adjusted analysis. The registry record does not give all model details needed to reproduce the sponsor’s interval. These intervals therefore should not be compared as if they were estimates from an identical model and contrast.

Confidence intervals communicate uncertainty around an estimate, not certainty about an individual outcome. In general, the width of an interval reflects the precision delivered by the data and the statistical procedure. A confidence interval does not account for every possible source of uncertainty: model misspecification, missing-data assumptions, measurement limitations and the choice of analysis can also matter. The registry record’s primary endpoint description specifies an imputation approach, but does not report sensitivity analyses that would show how robust the estimate is to alternative missing-data assumptions.

The standard deviations for the posted primary group changes were 0.9 and 1.0. They describe variability among participant outcomes and are not confidence intervals for the group means. The sponsor’s interval and the recalculated interval address uncertainty in a treatment contrast. Keeping these quantities distinct helps avoid treating individual-level variability as uncertainty in the estimated group difference.

12. Planning a Similar Trial: Sample Size

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

The planning estimates are participants per arm, not total events. They are illustrations based on the observed effect, not the trial’s original sample-size design. The stated assumptions are a two-sided significance level of 0.05, equal allocation between the two arms, a normal approximation, and an effect to detect set at the stated fraction of the observed effect. The trial’s own original design assumptions are not reported in the registry record.

The 80% power column describes the number of participants per arm that the planning calculation estimates would be needed for an 80% chance of detecting the specified effect under those assumptions, if that effect and the calculation model were correct. The 90% power column uses the corresponding 90% chance. These are design-planning concepts, not guarantees about results in a future study. A larger power target generally requires more participants for the same effect and assumptions.

The rows that plan for 100%, 75% and 50% of the observed effect illustrate how the assumed effect size changes the planning estimate. When the target effect is smaller, it is harder to distinguish from random variation, so more participants per arm are needed under the same planning assumptions. The listed figures are illustrations computed from the observed effect and do not show the power actually achieved by this completed trial. They are not observed power.

The sample-size planning illustration is not a reconstruction of the original protocol. The registry record does not report the original sample-size justification or the assumptions used to choose enrollment. The calculation also uses simplified assumptions, including a normal approximation and equal allocation, and does not incorporate every possible design feature or analysis consideration. Its role is to make the relationship between effect size, power target and planned enrollment easier to understand, not to reassess the trial’s actual design adequacy.

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 = 0.600; pooled standard deviation σ = 0.951.
2. n = 2 × (1.96 + 0.842)² × 0.951² / 0.600² = 39.4, rounded up to 40 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 <- 0.6                      # observed difference in means
sigma <- 0.951054334844275                     # 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(3191396)
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 = 40 (80% power), 53 (90% power)
 75% of observed effect: n per arm = 71 (80% power), 94 (90% power)
 50% of observed effect: n per arm = 158 (80% power), 212 (90% power)
power.t.test: 40.4 
Simulated power at n = 40 per arm: 0.795 (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 sponsor’s primary result is the ANCOVA treatment difference of -0.69, with a 95% confidence interval from -0.82 to -0.56 and p < 0.001. The registry record reports this same estimate and interval under both the non-inferiority and superiority labels. The trial’s own adjusted analysis is distinct from the simple recalculation based on the posted group means and standard deviations.

The recalculated mean difference is 0.6 when expressed as liraglutide minus semaglutide, with a 95% confidence interval from 0.445 to 0.755 and p <0.001. Its sign is opposite to the sponsor’s reported treatment difference, consistent with an opposite contrast orientation. The estimates also come from different methods: one is an unadjusted comparison of summary statistics and the other is ANCOVA. They should not be read as identical estimates or as conflicting results merely because their signs and values differ.

The registry record does not flag a discrepancy between the sponsor analysis and the recalculation. The sponsor’s result remains the trial’s primary reported inferential analysis; the recalculation supplies a transparent, unadjusted reference based on the posted summaries. The non-inferiority margin and the full ANCOVA specification are not reported in the registry record, which limits independent reconstruction of the sponsor’s analysis.

15. Multiplicity, Missing Data and Analysis Populations

Multiplicity arises when a study evaluates multiple outcomes, comparisons or hypotheses. In general, testing many hypotheses without adjustment can increase the chance that at least one result appears statistically notable even when there is no corresponding underlying effect. The registry record lists 40 secondary outcomes and reports three analyses. It does not report a multiplicity adjustment for the secondary results, a testing hierarchy, or a complete plan describing how the family of hypotheses was handled. The unadjusted secondary p-values therefore do not have the same confirmatory status as a result interpreted under a prespecified multiplicity strategy.

The primary outcome was analyzed in the full analysis set, defined as all randomized participants. The primary endpoint description specifies the on-treatment without rescue medication period, so the analysis also depends on the treatment and rescue-medication conditions used to define that period. The record does not report further details about how participants with treatment discontinuation or other intercurrent events were handled beyond that endpoint description.

Missing primary-outcome data were imputed using observed data from participants in the same randomized-treatment group. The regression model included the stratification factor as a categorical effect, with baseline and all previous visits as covariates. This is the method described for the primary endpoint; the registry record does not report enough detail to evaluate the imputation assumptions fully, nor does it report sensitivity analyses that test alternative assumptions. Missing-data methods for secondary outcomes are not reported in the registry record.

The non-inferiority margin and the details of the non-inferiority testing procedure are not reported in the registry record. The original sample-size justification, interim analyses and a complete statistical analysis plan are also not reported. These omissions do not change the recorded estimates, but they limit a reader’s ability to reconstruct the full inferential framework or assess how all planned analyses were controlled.

16. Safety Results

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

Deaths

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: pyelonephritis (0.7%), abdominal pain upper (0.3%), acute kidney injury (0.3%), bradycardia (0.3%), calculus urinary (0.3%).

The registry record lists serious adverse events for 289 participants at risk in the semaglutide group and 287 at risk in the liraglutide group. Serious adverse events were recorded for 17 and 22 participants, respectively. Deaths were recorded as 0 in each group. The record also lists other adverse events for 139 participants in the semaglutide group and 113 in the liraglutide group, using the same respective at-risk counts.

These are counts of participants with events in the listed categories, not a complete description of event severity, duration, relatedness to treatment or clinical course. The most frequent serious-event terms listed include pyelonephritis at 0.7%, and abdominal pain upper, acute kidney injury, bradycardia, calculus urinary and cholecystitis at 0.3% each. The registry record does not provide enough detail in these summary figures to compare the full safety profiles or attribute events to treatment.

Safety counts should be considered alongside the way events were collected and the size and duration of follow-up. The record gives the study duration description and recorded at-risk counts, but these summaries alone do not establish whether a difference in counts is meaningful or explain its cause. Zero recorded deaths does not mean that every possible safety outcome was ruled out.

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

The primary result describes average HbA1c change from week 0 to week 30 in the randomized groups, under the specified on-treatment without rescue medication analysis definition. The posted group means were -1.7 for semaglutide 1.0 mg and -1.1 for liraglutide 1.2 mg. The sponsor’s ANCOVA analysis reported a treatment difference of -0.69, with a 95% confidence interval from -0.82 to -0.56 and p < 0.001. These are statistical descriptions of the recorded outcome, not a complete account of every participant’s experience.

The result does not show that every participant had the group-average change, nor does it describe the range of individual outcomes beyond the reported standard deviations. The comparison also does not establish what would happen under different doses, different background treatments, longer follow-up, or in populations not represented by the study. The registry record describes participants with type 2 diabetes using 1-3 oral anti-diabetic drugs, and the study was conducted in Europe.

A small p-value is not a measure of clinical importance, and a confidence interval does not convey every source of uncertainty. The record does not report the non-inferiority margin, full analysis plan, original sample-size justification or detailed sensitivity analyses for missing data. Those details are relevant to judging how the trial’s statistical conclusions were framed and how dependent they may be on analytic assumptions.

Secondary results provide context on other measured outcomes, but the record does not report a multiplicity adjustment for the secondary comparisons. Their unadjusted results should not be treated as a separate collection of definitive findings. Safety summaries likewise describe recorded counts and selected event terms, not complete profiles or causal attribution. Together, these limits mean the results should be interpreted narrowly as a comparison of the recorded outcomes under the stated study conditions.

18. Why This Trial Matters Statistically

This trial illustrates why the analysis definition matters alongside the observed group means. The primary result is not simply a comparison of any week-30 measurements: it is defined for an on-treatment period without rescue medication, and missing values were imputed using a regression model based on observed data from the randomized-treatment groups. Those choices shape what the estimate represents and which assumptions support it.

It also illustrates why adjusted sponsor analyses and recalculations from summary statistics should be kept distinct. The posted means allow a simple unadjusted contrast, while the sponsor reports an ANCOVA treatment difference. Contrast direction affects the sign, and adjustment can affect the estimate and its uncertainty. Reading the methods as well as the numeric result avoids mistaking different statistical summaries for direct contradictions.

The registry record labels the primary analyses as both non-inferiority and superiority, but does not report the non-inferiority margin or complete testing framework. That is a useful reminder that an analysis label alone is not enough to reconstruct the decision rule. Similarly, the presence of numerous secondary outcomes makes multiplicity important: unadjusted secondary p-values require more caution than the designated primary analysis.

Finally, the planning illustrations show the relationship between the assumed effect size and the number of participants per arm needed under stated statistical assumptions. They are not the original design calculation and do not measure power observed in the completed study. Separating planning illustrations, posted summaries and sponsor analyses helps keep each statistic tied to the question it actually addresses.

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 does not report the non-inferiority margin, so the margin-based criterion cannot be assessed independently. It also does not provide a complete specification of the sponsor’s ANCOVA model, including all covariates and the treatment contrast coding. The sponsor’s estimate can be reported as recorded, but the registry entries do not permit full reproduction of that analysis.

The primary endpoint is restricted to an on-treatment period before rescue medication. That definition does not represent an unrestricted comparison regardless of treatment status or rescue use. The record describes imputation for missing primary data but does not report sensitivity analyses or enough detail to evaluate the robustness of the result to alternative missing-data assumptions. Methods for missing secondary outcomes are not reported in the registry record.

The trial was unmasked, with masking listed as none. The record does not provide enough detail to determine how the open-label design may have affected outcome measurement, participant behavior or reporting. The registry record also does not report the original sample-size justification, interim analyses, a complete statistical analysis plan, or a multiplicity adjustment for secondary results. The non-inferiority decision process therefore cannot be reconstructed fully, and unadjusted secondary findings warrant caution.

The safety summaries give counts and selected serious-event terms, but do not by themselves describe event timing, duration, relatedness, resolution or the full clinical context. The primary and secondary results concern the recorded study population and follow-up, not other doses, populations or treatment settings. These limitations define the scope of what can be concluded from the registry record.

20. Related Tutorials

21. Related Statistical Calculators

22. Sources

• ClinicalTrials.gov: Research Study Comparing a New Medicine Semaglutide to Liraglutide in People With Type 2 Diabetes — NCT03191396.
• Registry record last updated 2019-10-15; status last verified 2019-10.
• Results first posted 2019-08-28.
• 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

The registry record describes a completed phase 3 randomized, parallel-group study with no masking. It compared semaglutide 1.0 mg once weekly with liraglutide 1.2 mg once daily as add-on treatment to 1-3 oral anti-diabetic drugs in people with type 2 diabetes. Recorded enrollment was 577, and the study was conducted in Europe.

The primary outcome was mean change in HbA1c from week 0 to week 30, evaluated in the full analysis set during the on-treatment without rescue medication period. The sponsor reported an ANCOVA treatment difference of -0.69, with a 95% confidence interval from -0.82 to -0.56 and p < 0.001. The registry record labels this analysis as both non-inferiority and superiority, but does not report the non-inferiority margin.

Secondary outcomes and safety summaries add context, but multiplicity procedures, several analysis details, the original sample-size justification and sensitivity analyses for missing data are not reported in the registry record. The posted summary-statistic recalculations are unadjusted and should be distinguished from the sponsor’s primary analysis.