Safety/Efficacy of Sitagliptin in Patient w/ Type 2 Diabetes (0431-801): Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Safety/Efficacy of Sitagliptin in Patient w/ Type 2 Diabetes (0431-801) (NCT00541775), a randomized phase 3 study. The primary outcome is Hemoglobin A1C (A1C) at Week 18.

2. Clinical Question

This is a short-term, between-group question about a continuous laboratory measure. It is not a question about whether an individual patient reaches a particular A1C target, nor does the primary endpoint report the proportion of patients who improve. The registered endpoint is the mean change from baseline, expressed in percentage points of glycosylated hemoglobin.

The registry's primary-hypothesis comparison is specifically sitagliptin versus placebo. The study is described as placebo and active controlled, but the posted primary analysis does not report a sitagliptin comparison with rosiglitazone or metformin. The registry record does not report an outcome estimate for a metformin arm in the primary results. Consequently, the central statistical interpretation concerns the placebo comparison, not every intervention named in the intervention list.

The question is framed for the all patients treated population, with missing A1C values handled by last observation carried forward. That definition matters: the result applies to the analysis population as defined in the record, rather than automatically to every randomized participant. The record does not report enough information to determine how the analysis population relates participant by participant to all randomized patients.

3. Trial Design

The study is registered as a Phase 3, randomized, double-blind, parallel-group trial. It is titled as a multicenter study evaluating the safety and efficacy of adding sitagliptin 100 mg once daily in patients with type 2 diabetes and inadequate glycemic control on metformin monotherapy. The registry lists 273 participants as enrolled and marks the study completed.

The listed arm descriptions include sitagliptin 100 mg, rosiglitazone 8 mg, and placebo. The intervention list also names metformin as a comparator, but the primary outcome results list sitagliptin, rosiglitazone, and placebo. How the metformin intervention entry relates to the study arms is not reported in the registry record. The posted primary-hypothesis analysis compares sitagliptin with placebo; it should not be read as a report of all possible pairwise comparisons.

The primary outcome is A1C at Week 18, with the timeframe specified as baseline and 18 weeks. Two secondary outcomes are listed: fasting plasma glucose and 2-hour post-meal glucose at Week 18. The registry reports three outcome measures and three analyses overall, but the record does not report the complete outcome-by-outcome analytic details in the registry record for this interpretation.

Enrollment began in 2006-06, and primary completion is recorded as 2007-03. These dates describe the registered study timeline; they do not establish when results were analyzed or posted. The record does not report the planned sample-size justification, randomization procedure, allocation ratio, site count, or a detailed protocol for treatment duration beyond the stated Week 18 endpoint. The term multicenter appears in the official title, but the number or locations of centers are not reported in the registry record.

Randomization and double masking are relevant design features because, in general, random allocation can support comparisons between groups and masking can reduce some forms of behavior or assessment bias. Those features do not by themselves resolve concerns about missing outcomes, analysis choices, or the scope of the reported comparisons. The specific implementation details needed to assess those issues are not reported in the registry record.

4. Participant Flow and Accounting

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

Before assignment: Patients 18-75 years of age with type 2 diabetes mellitus (T2DM) who were taking metformin monotherapy at a stable dose of ≥ 1500 mg/day for at least 10 weeks and had inadequate glycemic control (hemoglobin A1C ≥ 7.0% and ≤11%) were eligible to participate.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: The all patients treated population included all patients who took at least one dose of study medication and had both a baseline measurement and at least one post-randomization measurement for this outcome. Missing data were imputed using the last observation carried forward (LOCF) method.

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 Hemoglobin A1C (A1C) at Week 18, measured over Baseline and 18 Weeks.

The primary endpoint is change in hemoglobin A1C from baseline to Week 18, measured as percent of glycosylated hemoglobin. The registry defines the change as Week 18 A1C minus Week 0 A1C. A negative change therefore represents a lower measured A1C at Week 18 than at baseline for that arm; the group mean describes an average, not the outcome of every participant.

A1C is used as a measure of glycemic control over time, rather than a single point-in-time glucose reading. In general, a continuous endpoint retains information about the size of changes in the measurement, whereas a binary endpoint would classify participants into categories. The registry's endpoint is a mean change, so it answers a question about average change between groups and not the frequency of reaching a threshold.

The arm estimates are least-squares means with 95% confidence intervals. In general, a least-squares mean is a model-based estimate, which may account for terms specified in the analysis model. The registry record does not identify the covariates or full model specification used for the arm estimates. Thus, the posted estimates should be read as reported model-based means, not assumed to be simple raw averages.

The analysis population included patients who took at least one dose of study medication and had both a baseline measure and at least one post-randomization measure for the outcome. Missing data were imputed using last observation carried forward. These choices define how the endpoint was calculated and which participants contributed to the reported analysis. The record does not report sensitivity analyses that test whether the result changes under different assumptions about missing A1C measurements.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The primary comparison is reported as an ANCOVA analysis. In general, analysis of covariance models a continuous outcome or change score while accounting for one or more covariates. This can improve precision or adjust for specified baseline differences, depending on the model and its assumptions. The registry record does not report the full ANCOVA specification, including the covariates, whether baseline A1C was included, or the diagnostic checks used. It is therefore not possible to reconstruct the adjusted result from the record alone.

What does the reported contrast represent?

The posted contrast is a net mean difference of -0.51 for sitagliptin versus placebo. Its confidence interval is -0.70 to -0.32 at the 95% level, and the reported p-value is p <= 0.001. The contrast summarizes the modeled separation between the two groups on the change-from-baseline scale. The negative direction is consistent with the arm-level mean changes, where the sitagliptin estimate is -0.73 and the placebo estimate is -0.22. A net contrast is not a separate treatment-group mean; it is an estimate of the difference between groups.

In general, the p-value measures how incompatible the observed data, or more extreme data, are with a specified null model, assuming the model and analysis are appropriate. It is not the probability that the null hypothesis is true, and it does not measure the size or importance of the estimated difference. Here the p-value is reported as p <= 0.001. Its interpretation depends on the analysis having been specified and conducted as assumed; the registry record does not report the complete statistical analysis plan or whether this was the only inferential test considered.

Why use a confidence interval?

The 95% confidence interval communicates uncertainty around the estimated contrast, alongside its direction and size. In general, a confidence interval is generated by a procedure that would cover the true parameter at the stated rate over repeated samples if the assumptions hold. For an individual study, it is most useful as a measure of estimation uncertainty, not as a probability statement about the fixed true value. The reported interval does not include zero. That is consistent with the small reported p-value for the corresponding null comparison, subject to the model and test definitions.

The endpoint arm results also report 95% confidence intervals: -0.87 to -0.6 for sitagliptin, -0.92 to -0.65 for rosiglitazone, and -0.36 to -0.08 for placebo. Those intervals quantify uncertainty around each arm's least-squares mean. Comparing arm-level intervals visually is not equivalent to testing the difference between arms, because the uncertainty of a contrast depends on the joint model and covariance structure. The direct ANCOVA result is the reported inferential comparison for sitagliptin versus placebo.

What does least-squares mean imply?

In general, a least-squares mean is a fitted-model estimate of a group mean, often adjusted to a common value or distribution of model covariates. It may differ from an unadjusted arithmetic mean. Because the registry does not list the model terms or covariate values used for standardization, the precise adjustment cannot be described. These estimates should therefore be identified by their reported label and not treated as raw sample averages.

General statistical models also rely on assumptions about the outcome distribution, the relationship between covariates and outcome, and the adequacy of the fitted form. For continuous outcomes, residual behavior and variance structure can matter. The registry record does not report model diagnostics, transformations, or checks of these assumptions. That absence is a limitation in assessing the analysis, not evidence that any particular assumption failed.

How should the planning illustration be read?

Sample-size calculations associated with observed effects are planning illustrations, not observed power and not evidence that the original trial was designed around the effect ultimately reported. The stated illustration assumptions are a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation; the effect to detect is a stated fraction of the observed effect. Such calculations answer a hypothetical design question under those assumptions. They do not show the probability that this completed trial detected an effect, nor do they recover the trial's original sample-size rationale.

The registry record does not report the trial's own sample-size justification or the target effect used in planning. Any illustrative sample size depends on the assumed effect, variability, significance level, allocation, and model. If an observed effect is used as the planning input, the resulting number is conditional on that estimate and may not reflect a future study's true effect. The illustration should not be read as a recommendation or as an account of the original protocol.

What do the arm counts tell us?

The endpoint analysis lists 91 participants for sitagliptin, 87 for rosiglitazone, and 88 for placebo. In general, estimates from continuous outcomes become less precise when information is limited, though precision also depends on outcome variability and model structure. The registry lists confidence intervals, but does not provide the standard deviations or full model details needed to assess precision independently. The difference between the outcome counts and safety denominators is not explained in the registry record.

Finally, the reported statistical comparison is one part of interpretation. Model-based evidence, uncertainty, outcome definition, missing-data handling, and the extent of reported analyses all matter. A small p-value does not resolve questions about multiplicity, missingness assumptions, practical importance, or generalizability. Those matters require details not fully reported in this registry record.

10. Analyses Reported by the Sponsor

The registry reports the sponsor's primary comparison as an ANCOVA net mean difference of -0.51 for sitagliptin versus placebo, with a 95% confidence interval from -0.70 to -0.32 and p <= 0.001. The arm-level least-squares mean changes are -0.73 for sitagliptin and -0.22 for placebo. These are distinct reported quantities: the first is the modeled between-group contrast, while the latter are the separate group estimates.

A simple unadjusted re-calculation from posted summary statistics, when available, is not the same as the sponsor's fitted analysis. It may use less information, omit covariate adjustment, and depend on rounding of the posted arm estimates and intervals. No independently recalculated contrast is reported in the registry entries for this interpretation, so no numerical re-calculation should be substituted for the sponsor's ANCOVA estimate.

The registry marks no discrepancy between the sponsor's analysis and a recalculation. That does not mean the analysis can be fully reproduced: the ANCOVA terms and detailed model specification are not reported in the registry record. For the primary inference, the sponsor's own reported analysis is the relevant estimate. The arm means provide context for its direction, but they do not reproduce the adjusted contrast or replace its confidence interval and p-value.

11. Multiplicity, Missing Data and Analysis Populations

The registry record lists one primary outcome, A1C at Week 18, and two secondary outcomes: fasting plasma glucose and 2-hour post-meal glucose at Week 18. It also records three outcome measures and three analyses. The primary hypothesis comparison is sitagliptin versus placebo. The record does not report a complete testing hierarchy, whether the secondary outcomes were confirmatory or exploratory, or whether any multiplicity adjustment was applied.

Multiplicity arises when several outcomes, time points, groups, or analyses are examined. In general, conducting multiple tests can increase the chance of at least one apparently unusual result under a collection of null hypotheses. A prespecified hierarchy or adjustment can address this issue, but the necessary plan is not reported in the registry record. The reported p-value should therefore be interpreted as the result for the stated comparison, without assuming it was adjusted for other analyses or that it was the sole test in the study.

The primary analysis population included all patients who took at least one dose of study medication and had a baseline A1C measurement and at least one post-randomization measurement for that outcome. Missing data were imputed using last observation carried forward (LOCF). In general, LOCF carries a participant's last recorded value forward to a later assessment when the later value is missing. This makes an assumption that the last observed measurement is an appropriate substitute for the unobserved outcome; that assumption may not fit every participant's trajectory.

The registry record does not report the number of missing Week 18 outcomes, reasons for missingness, or sensitivity analyses using alternative assumptions. It also does not report enough information to determine whether missingness differed between treatment groups or how robust the result is to those patterns. The analysis population definition is not equivalent to an all-randomized analysis unless the record establishes that equivalence, which it does not.

These omissions matter because the inferential result depends on both who contributed data and how missing values were handled. The available record supports describing the stated LOCF analysis and its estimated contrast. It does not support claims about results under other missing-data methods, the effect of multiplicity adjustment, or the full statistical analysis plan.

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: overdose (3.4%), intentional overdose (1.1%), syncope (1.1%), prostate cancer (1.1%), endometrial hyperplasia (1.1%).

The registry lists serious adverse events affecting 5 of 94 participants in the sitagliptin safety group, 5 of 87 in the rosiglitazone group, and 5 of 91 in the placebo group. These figures count participants affected as recorded in the safety results; they are not a measure of the total number of serious events. The record does not report deaths for these groups.

For other adverse events, the listed counts are 5 of 94 for sitagliptin, 1 of 87 for rosiglitazone, and 0 of 91 for placebo. The registry separately lists serious-event terms and percentages, including overdose at 3.4%, intentional overdose at 1.1%, syncope at 1.1%, prostate cancer at 1.1%, endometrial hyperplasia at 1.1%, and asthma at 1.1%. The record does not assign these terms to particular treatment arms in the information, so they should not be attributed to an arm.

Safety counts are descriptive. In general, comparisons of adverse events require attention to definitions, exposure duration, event severity, and uncertainty, as well as the number of participants at risk. The registry record does not provide enough detail to assess event timing, exposure-adjusted rates, causal attribution, or statistical comparisons. The counts alone should not be interpreted as proof that an event was caused by a study drug or as a complete account of safety.

The denominators in the safety results differ from the outcome-analysis counts. The reason for those differences is not reported in the registry record. Thus, safety and efficacy denominators should be read as belonging to their respective reported analyses rather than assumed to describe identical participant sets.

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

The posted primary result indicates a modeled difference in mean A1C change at Week 18 between sitagliptin and placebo within the defined treated analysis population. The estimated net mean difference is -0.51, with a 95% confidence interval from -0.70 to -0.32 and p <= 0.001. The estimate and interval describe the size and statistical uncertainty of the contrast under the reported ANCOVA approach.

This finding does not mean that every participant's A1C changed by the group average, or that every participant experienced a decrease. Nor does the p-value state the probability that the result is due to chance or the probability that the treatment has a particular effect. In general, the p-value is conditional on a null model and assumptions about how the data were analyzed. The confidence interval likewise reflects uncertainty in the estimated group contrast, not the range of individual responses.

The registry record does not report a threshold for a clinically meaningful A1C difference. Statistical evidence that a modeled contrast differs from zero does not, by itself, establish how important that difference is in clinical practice. The endpoint is a laboratory measure at Week 18, and the primary result does not answer questions about longer-term outcomes, complications, quality of life, or persistence of the measured change. Those outcomes are not reported as part of this primary comparison.

The result also does not establish a direct advantage or disadvantage relative to rosiglitazone. Although the rosiglitazone arm has a reported least-squares mean change, the primary inferential result is for sitagliptin versus placebo. No direct sitagliptin-versus-rosiglitazone estimate is reported in the primary analysis. The intervention list names metformin, but the primary outcome results do not provide a metformin-arm estimate.

Interpretation is bounded by the analysis population and LOCF method, as well as by the absence of a complete statistical plan and sensitivity analyses in the registry record. The result describes the registered analysis, not every reasonable alternative analysis. It should be read as an estimate from a randomized, double-blind trial with a specified endpoint and comparator, while recognizing the uncertainty and reporting gaps that remain.

14. Why This Trial Matters Statistically

This trial illustrates the distinction between a within-arm change and a between-arm treatment contrast. The arm means describe average change from baseline in each group. The primary question, however, is whether the mean change differs between sitagliptin and placebo. The ANCOVA net mean difference addresses that comparison directly; interpreting one arm's change alone would not answer the randomized comparison.

It also shows why model-based estimates and raw descriptions should not be treated as interchangeable. The least-squares means and ANCOVA contrast are reported as model-based quantities, but the full model specification is absent. A reader can interpret the stated estimate and interval while recognizing that reproducing the analysis requires details not in the registry record. This is a useful distinction between reading a registered result and independently verifying its calculation.

The record provides an opportunity to consider uncertainty alongside statistical evidence. The confidence interval gives a range of plausible values under the stated procedure, while the p-value addresses compatibility with a null comparison. Neither quantity answers whether the observed difference meets a clinical-importance threshold, which is not reported. Emphasizing both estimate and precision is more informative than focusing on the p-value alone.

Finally, the outcome illustrates how analysis-population definitions and missing-data methods affect the meaning of a result. The stated population excludes patients without the required measurements and uses LOCF for missing values. These details are integral to the estimand actually represented by the reported analysis. The record's incomplete reporting of multiplicity, model terms, and sensitivity analyses limits how far the findings can be generalized across analytic choices. These are not reasons to disregard the posted result, but they are reasons to state precisely what it estimates.

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 record does not report the ANCOVA covariates, full model specification, or model diagnostics. The sponsor's adjusted contrast can be quoted, but it cannot be fully reconstructed from the listed arm estimates alone. The absence of these details also limits assessment of how the adjustment affected the result.

Missing A1C data were handled by LOCF, but the record does not report how many Week 18 values were missing, why they were missing, or whether missingness differed by group. No sensitivity analyses under alternative missing-data assumptions are reported in the registry record for this interpretation. Consequently, robustness to missing-data choices cannot be assessed.

The record lists secondary outcomes and multiple analyses but does not report a complete multiplicity plan. It is not possible to determine whether the primary comparison was protected by a prespecified testing hierarchy or how secondary-outcome testing was addressed. The record also does not report the original sample-size justification, planned effect, or interim-analysis plan.

The primary hypothesis concerns sitagliptin versus placebo. The posted primary analysis does not provide a direct inferential comparison with rosiglitazone, and no primary outcome estimate for a metformin arm is listed. The registry's metformin intervention entry is not reconciled with the three arm descriptions and results summarized in this record.

Safety denominators differ from the outcome-analysis counts, and the reason is not reported. Death counts are not reported in the registry record. Serious-event terms and percentages are listed, but the information does not identify their treatment-arm attribution or provide causal assessments. Finally, the endpoint is measured at Week 18; longer-term outcome results are not reported in the primary endpoint information.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Safety/Efficacy of Sitagliptin in Patient w/ Type 2 Diabetes (0431-801) — NCT00541775.
• Registry record last updated 2015-08-24; status last verified 2015-08.
• Results first posted 2010-06-17.
• 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

NCT00541775 is a completed Phase 3 randomized, double-blind, parallel-group study of sitagliptin in patients with type 2 diabetes and inadequate glycemic control on metformin monotherapy. The registry lists 273 enrolled participants, with study start in 2006-06 and primary completion in 2007-03.

The primary endpoint was change in A1C from baseline to Week 18. In the stated treated analysis population, the reported sitagliptin-versus-placebo ANCOVA net mean difference was -0.51, with a 95% confidence interval from -0.70 to -0.32 and p <= 0.001. Missing outcome values were imputed using LOCF.

The registry also reports arm-level A1C estimates for rosiglitazone, but the primary inferential comparison is sitagliptin versus placebo. Detailed model terms, the trial's sample-size justification, a complete multiplicity plan, and sensitivity analyses for missing data are not reported in the registry record.