Sitagliptin Added-on to Insulin Study (0431-051): Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Sitagliptin Added-on to Insulin Study (0431-051) (NCT00395343), a randomized phase 3 study. The primary outcome is Change From Baseline in A1C at Week 24.

2. Clinical Question

The primary endpoint addresses whether the average observed change in A1C differed between the two randomized groups over the study’s stated follow-up. The registry reports group estimates and a sponsor analysis of the net mean difference. A negative change represents a lower Week 24 A1C than at baseline, by the endpoint’s definition; the reported group estimates were negative for both groups.

The trial record describes a Phase III, randomized, double-blind study in people with type 2 diabetes. Total enrollment was 641. The primary endpoint population was defined using outcome availability and a stated approach to observations after glycemic rescue and missing Week 24 measurements. The record does not report further detail about baseline eligibility criteria or the clinical characteristics of participants.

This question is about the comparative statistical result for this endpoint and time point. It is not, by itself, a complete assessment of safety, other outcomes, longer-term effects, or how results might apply to people outside the trial population.

3. Trial Design

The registry describes a Phase III, multicenter, randomized, double-blind clinical trial with a parallel-group model. Participants were assigned to sitagliptin 100 mg once daily or placebo. The study concerned adding sitagliptin to insulin therapy, either alone or with metformin, in patients with type 2 diabetes and inadequate glycemic control. The record gives total enrollment as 641.

The primary outcome was change from baseline in A1C at Week 24. The outcome was defined as the Week 24 A1C percent minus the Week 0 A1C percent. For the primary endpoint, the stated Full Analysis Set included patients who had a baseline value and at least one post-baseline value. Data following glycemic rescue were treated as missing, and when a Full Analysis Set patient had no Week 24 data, the last non-baseline observed measurement was carried forward to Week 24.

The registry lists fasting plasma glucose, two-hour post-meal glucose, beta-cell sensitivity to glucose, and percentages of patients crossing two A1C thresholds among the secondary outcomes. Five secondary outcomes are posted in the record. The results described for the primary endpoint do not establish what happened on those secondary outcomes.

The listed study dates are 2006-12-11 for study start and 2008-10-13 for primary completion, and the status is completed. More detailed information about the randomization procedure, allocation concealment, masking procedures, study sites, treatment duration beyond the stated endpoint, adherence, and protocol deviations is not reported in the registry record. The record also does not report the planned statistical analysis document or a sample-size justification.

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: Patients at least 21 years of age with type 2 diabetes mellitus with inadequate glycemic control (A1C ≥7.5 and ≤11.0%) on stable-dose insulin (alone or in combination with stable-dose metformin) were eligible to enter the 24 week study. 1-week screening, followed by a 2-week single-blind placebo run-in.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: The Full Analysis Set (FAS) included all patients with a baseline value and ≥1 post-baseline value for this outcome. Data following glycemic rescue were treated as missing. For FAS patients with no data at Week 24, the last non-baseline observed measurement was carried forward to Week 24.

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 From Baseline in A1C at Week 24, measured over Baseline and Week 24.

The primary endpoint is change from baseline in A1C at Week 24, measured in percent. A1C is a continuous measure of glycemic status; the trial reports the mean change for each group and a between-group mean difference. By the registry’s definition, change is the Week 24 A1C value minus the Week 0 value. Thus, a negative change indicates that the measured A1C value at Week 24 was lower than the baseline value, while a value nearer zero indicates less change from baseline.

Using change from baseline allows the analysis to compare each participant’s follow-up measurement with their own starting measurement, then compare average changes between groups. The registry reports least-squares means for the groups and a sponsor-reported net mean difference from an analysis of covariance. General statistical context: analysis of covariance can account for specified variables when estimating a group contrast, but the registry record does not identify the covariates used for this analysis.

A1C is used as a clinical-trial outcome because it summarizes glycemic status over time rather than capturing only a single measurement occasion. The registry’s stated primary endpoint asks about the between-group contrast at Week 24; it does not, on its own, address other outcomes, longer follow-up, or all aspects of diabetes care. The outcome population and handling of missing measurements are therefore important to interpreting the estimate.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The primary outcome is continuous, so the registry presents a mean change for each group and a mean difference between groups. General statistical context: a mean summarizes the central level of a continuous measurement across participants, while the difference between group means describes the average contrast on the outcome’s measurement scale. Here, the scale is percentage-point change in A1C from baseline to Week 24. A group mean does not imply that all participants experienced the same change.

What does the confidence interval communicate?

The reported 95% confidence intervals describe uncertainty around the estimated group means and the sponsor’s estimated net difference, according to the respective analyses. In general, a confidence interval is generated by a statistical procedure that, under its assumptions, would cover the underlying parameter at the stated rate over repeated studies. It is not a range containing a known proportion of individual participant outcomes. The sponsor’s interval for the net difference runs from -0.70 to -0.42. Its position below zero indicates that the estimated contrast is negative throughout that interval. The interval’s width also matters: it indicates the range of effect sizes reasonably compatible with the analysis, rather than proving one exact effect size.

What does the p-value add?

The sponsor reports p < 0.001 for the primary comparison. General statistical context: a p-value assesses how unusual data at least as incompatible with a specified null hypothesis would be if that hypothesis and the model assumptions held. It is not the probability that the null hypothesis is true, and it does not measure the size or practical importance of an effect. The reported p-value accompanies the estimated contrast and confidence interval; it does not replace either of them.

What is ANCOVA?

The sponsor’s reported comparison used ANCOVA, an abbreviation for analysis of covariance. In general, this method estimates a group contrast while accounting for one or more covariates included in the model. The registry reports the method and the net mean difference but does not report which covariates were included, the model specification, or diagnostic checks. It is therefore possible to describe the analysis label and reported estimate, but not to reconstruct the sponsor’s model from the registry record alone.

The group estimates are labelled least-squares means. In general, least-squares means are model-based estimates rather than necessarily the raw arithmetic means of observed values. They can depend on model choices and covariate adjustment. The registry identifies the estimate type but does not provide enough model detail to assess how those choices influenced the estimates. These values should therefore be read as the posted model-based group estimates, not as independently verified raw averages.

Why does the outcome population matter?

The Full Analysis Set for the primary endpoint included patients with a baseline value and at least one post-baseline value. This is the registry’s stated population definition, and it is not identical to a statement that every randomized participant contributed a Week 24 measurement. The endpoint-specific group estimates are reported for 305 in the sitagliptin group and 312 in the placebo group. These are the reported outcome group sizes; the registry record does not explain every difference between those counts and total enrollment.

The record states that observations after glycemic rescue were treated as missing. For patients in the Full Analysis Set without a Week 24 observation, the last non-baseline observed measurement was carried forward to Week 24. General statistical context: this approach fills a missing later value with an earlier observed value. It treats that carried-forward value as the endpoint measurement for the analysis, but the value was not actually measured at Week 24. If the outcome could change between the last observation and Week 24, the carried-forward value may not represent the unobserved endpoint. The registry does not report sensitivity analyses comparing this approach with alternatives.

What about over-dispersion?

Over-dispersion is a general concern for some count or binary-outcome models when observed variation exceeds what a model assumes. It is not the central issue for interpreting this continuous A1C mean comparison, and the registry does not report an over-dispersion analysis. The relevant questions for this endpoint are the continuous-outcome model, the covariates used, the analysis population, and the handling of missing values. Those details are not all reported.

How should the analysis be distinguished from planning?

Sample-size calculations are planning illustrations, not observed trial outcomes and not evidence that a study had a particular probability of detecting its result. General statistical context: a planning calculation depends on assumptions about the effect to detect, variability, allocation, significance level, and analysis. The registry record does not report the trial’s own sample-size justification or the assumptions used in its design. Therefore, any planning illustration based on the observed effect should not be interpreted as the trial’s original design rationale, or as observed power.

Multiplicity is another general consideration when a trial evaluates several outcomes or analyses. Testing multiple hypotheses can increase the chance of at least one nominally significant result unless an approach to multiplicity is specified. The record lists five secondary outcomes in addition to the primary endpoint, but the multiplicity strategy is not reported in the registry record. The reported primary comparison can be described as posted; the record does not establish whether or how the family of outcomes was controlled.

10. Analyses Reported by the Sponsor

The registry reports the sponsor’s own primary comparison as an ANCOVA net mean difference of -0.56, with a 95% confidence interval from -0.70 to -0.42 and p < 0.001. This is the sponsor-reported inferential analysis for the primary endpoint. The group-specific least-squares means are -0.59 for sitagliptin and -0.03 for placebo, each reported with its own 95% confidence interval.

A simple unadjusted recalculation from posted summary statistics, when available, is not necessarily identical to a sponsor’s adjusted model-based analysis. Differences can arise because model-based estimates may account for covariates, while an unadjusted contrast does not. The registry record does not report a recalculated contrast, so there is no separate recalculated result to compare with the sponsor’s estimate. The sponsor’s analysis therefore remains the posted comparative result; the record does not offer enough details to reproduce its ANCOVA independently.

The registry does not flag a discrepancy between the sponsor’s analysis and a recalculation. That absence does not establish that all model details can be reconstructed from the registry. The sponsor’s reported estimate, confidence interval, and p-value should be interpreted as the posted analysis, alongside the stated population and missing-data approach.

11. Multiplicity, Missing Data and Analysis Populations

The primary endpoint population was the Full Analysis Set, defined as all patients with a baseline value and at least one post-baseline value for the outcome. Patients without any post-baseline value for this outcome were not included under that definition. The registry does not provide a fuller accounting of the numbers excluded from this population or the reasons for missing outcome data.

For this endpoint, data following glycemic rescue were treated as missing. For Full Analysis Set patients with no Week 24 measurement, the last non-baseline observed measurement was carried forward to Week 24. This procedure means that some analyzed Week 24 values may be imputed from earlier measurements rather than recorded at Week 24. General statistical context: carrying an earlier value forward assumes, for purposes of the analysis, that it can stand in for the missing later value. If the outcome would have changed after that earlier measurement, the substitution may not reflect the unobserved Week 24 result. The registry record does not report the amount or timing of missingness, the reasons for missing measurements, or sensitivity analyses using alternative assumptions.

The listed secondary outcomes include changes in fasting plasma glucose and two-hour post-meal glucose, change in an index of static beta-cell sensitivity to glucose, and the percentages of patients below two A1C thresholds. Five secondary outcomes are posted. When multiple outcomes are examined, general statistical context is that the chance of at least one nominally significant finding can rise if no adjustment or prespecified testing strategy is used. Whether a multiplicity adjustment or hierarchy was planned or applied is not reported in the registry record. The primary result should not be treated as proof about the secondary outcomes.

The record does not report interim analyses, stopping rules, a detailed statistical analysis plan, or a sample-size justification. Those details should not be inferred from the posted primary result. The registry’s endpoint population and missing-data rules are reported, but the broader analysis plan remains incompletely described.

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: any cardiac disorders (1.6%), any infections and infestations (0.9%), cardiac failure congestive (0.6%), any nervous system disorders (0.6%), angina pectoris (0.6%).

For serious adverse events, the registry reports 20 affected participants among 322 at risk in the sitagliptin group and 11 among 319 at risk in the placebo group. These are participant counts, not event totals; the record does not provide a complete account of event recurrence or participant-level overlap across categories. The registry also lists serious-event terms and percentages, including Any Cardiac disorders at 1.6%, Any Infections and Infestations at 0.9%, and several other terms at 0.6%.

For other adverse events, the record reports 49 affected participants among 322 at risk in the sitagliptin group and 25 among 319 at risk in the placebo group. Death counts are not reported. The serious-event terms are categories recorded in the registry; their presence does not establish a causal relationship to the assigned intervention. The record does not provide enough detail to assess event timing, severity beyond the serious classification, exposure duration, or whether categories overlap.

These figures are descriptive safety data and should be read separately from the primary A1C analysis. Differences in counts alone do not establish why events occurred or whether they were caused by a study intervention. The registry record does not report a formal comparative statistical analysis of these adverse-event counts. The reported at-risk counts are the denominators for the posted safety figures and should not be confused with the primary endpoint analysis group sizes.

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

The posted primary result is an estimated difference between randomized groups in mean change from baseline in A1C at Week 24. The sitagliptin group had a reported least-squares mean change of -0.59, while the placebo group had a reported mean change of -0.03. The sponsor’s ANCOVA estimate for the net difference was -0.56, with a 95% confidence interval from -0.70 to -0.42 and p < 0.001. These statistics describe the trial’s primary endpoint under its stated analysis approach.

The result does not mean that every participant had the group-average change. It does not quantify individual response, and the confidence interval is not a range for individual outcomes. It also does not independently establish the clinical importance of the estimated difference: the registry record does not report a threshold for clinical importance. A small p-value is not a measure of that importance.

The result is specific to the study population, treatment comparison, endpoint definition, and Week 24 timeframe. It does not answer whether the same contrast would occur in populations not represented in the trial, with different treatment regimens, or at other follow-up times. The record does not provide results for the listed secondary outcomes in the registry record described for this primary analysis, so no conclusion about those endpoints follows from the primary result.

Missing data and glycemic rescue handling are part of the result’s interpretation. The analysis treated post-rescue data as missing and carried the last non-baseline measurement forward for Full Analysis Set patients without a Week 24 observation. The resulting estimate reflects those rules and should not be read as if every participant had an observed Week 24 measurement.

Finally, the primary efficacy result is not a complete evaluation of safety. The registry lists serious and other adverse-event counts, but the posted counts alone do not establish causation. Considering the A1C result alongside safety, other outcomes, longer-term follow-up, and clinical context would require evidence beyond this endpoint estimate.

14. Why This Trial Matters Statistically

This trial illustrates why a randomized comparison is interpreted through the between-group estimate rather than through a single group’s change over time. Both reported mean changes were negative, but the comparative question is captured by the sponsor’s net mean difference. The placebo group supplies the concurrent reference for interpreting change during the study, rather than treating the intervention group’s before-and-after result as the whole answer.

It also shows how the reported estimand depends on the endpoint definition and analysis population. The primary measure is change from baseline to Week 24, and the Full Analysis Set required both a baseline value and at least one post-baseline value. The rules for glycemic rescue and missing Week 24 observations further specify what data contributed to the analysis. These choices affect what the estimate represents and should be read alongside its confidence interval and p-value.

The contrast between least-squares means and a sponsor-reported ANCOVA result is useful statistically: model-based estimates and a net comparison are related, but their interpretation depends on the analysis model. Since the registry does not state which covariates were included, readers cannot fully reproduce the sponsor’s adjustment from the record alone.

Finally, the record demonstrates the distinction between statistical evidence and broader judgment. The confidence interval and p-value summarize uncertainty under the reported analysis, while clinical importance, generalizability, secondary outcomes, and safety require separate information. The reported result is meaningful as a Week 24 comparative estimate, but it should not be expanded into claims that the registry does not support.

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 entry does not include enough detail to independently reconstruct the sponsor’s ANCOVA. Although the method and net mean difference are reported, the covariates, model specification, and diagnostic checks are not. The least-squares means are therefore best understood as the posted model-based estimates rather than independently verifiable raw group averages.

The missing-data approach may also affect interpretation. Data after glycemic rescue were treated as missing, and the last non-baseline observed measurement was carried forward for Full Analysis Set patients with no Week 24 data. The record does not report the amount, timing, or reasons for missing observations, nor sensitivity analyses to assess the effect of alternative assumptions. This limits assessment of how robust the estimated contrast is to missing-data handling.

The registry lists five secondary outcomes, but the multiplicity strategy is not reported. The primary result should not be taken as evidence for those outcomes. The statistical plan, interim analyses, sample-size justification, and detailed randomization procedures are also not reported in the registry record, so they cannot be inferred from the primary analysis.

The enrollment count is 641, while the reported endpoint group sizes are 305 and 312. The record does not explain all accounting between total enrollment and endpoint analysis counts, so these figures should not be treated as interchangeable. The safety denominators are also distinct from the endpoint group sizes.

Finally, the primary endpoint concerns A1C change at Week 24. It does not itself describe longer-term outcomes, participant-level variation, or the clinical importance threshold for the observed difference. Serious adverse-event counts are posted, but death counts are not reported, and the registry does not give a formal comparative safety analysis in the registry record.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Sitagliptin Added-on to Insulin Study (0431-051) — NCT00395343.
• Registry record last updated 2017-05-12; status last verified 2017-04.
• Results first posted 2009-10-23.
• 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

NCT00395343 is a completed Phase III, multicenter, randomized, double-blind, parallel-group study of sitagliptin added to insulin therapy, alone or with metformin, compared with placebo in patients with type 2 diabetes and inadequate glycemic control. Total enrollment was 641. The listed study start was 2006-12-11, and primary completion was 2008-10-13.

The primary endpoint was change from baseline in A1C at Week 24. The sponsor reported an ANCOVA net mean difference of -0.56, with a 95% confidence interval from -0.70 to -0.42 and p < 0.001. The stated endpoint population and missing-data rules are important to interpreting that estimate.

The registry lists five secondary outcomes and reports serious and other adverse-event counts. Further details about the statistical plan, multiplicity, missing-data sensitivity analyses, and some safety interpretation are not reported in the registry record.