A Study of an Investigational Drug Sitagliptin for Type 2 Diabetes Mellitus (0431-044): Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for A Study of an Investigational Drug Sitagliptin for Type 2 Diabetes Mellitus (0431-044) (NCT00127192), a randomized phase 2 study. The primary outcome is Change From Baseline in HbA1c at Week 12.

2. Clinical Question

This is a short-term, dose-response comparison rather than a comparison of long-term clinical outcomes. The primary outcome focuses on a laboratory measure over the study period. The registry also lists change in glycosylated albumin and fasting plasma glucose at Week 12 as secondary outcomes, but the primary statistical interpretation concerns HbA1c. The information in the registry record does not establish what effects might occur beyond the measured period or on outcomes that were not reported.

3. Trial Design

The registry classifies this as a Phase II, randomized, double-blind, parallel-group study of sitagliptin in type 2 diabetes mellitus. Enrollment was 363. The five groups were placebo and sitagliptin at 25 mg, 50 mg, 100 mg, or 200 mg once daily; the arm descriptions specify a 12-week treatment period. The study status is completed. The registry lists a study start of 2005-07 and primary completion of 2006-03.

Randomization means allocation to the study groups was determined by a random process, while double masking indicates that treatment assignment was masked as recorded by the registry. The record does not report the randomization procedure, the people masked, or details of how masking was maintained. It identifies a parallel design, in which the groups are distinct rather than each participant receiving multiple study treatments in sequence.

The primary measure was assessed at baseline and Week 12. The primary analysis population was the Full Analysis Set, defined as participants with a baseline value and at least one post-baseline value for the outcome. For a participant in that set with no Week 12 data, the last non-baseline observed measurement was carried forward to Week 12. That rule defines how the registry handled missing Week 12 outcome data for this analysis; it does not reveal how many measurements were carried forward.

The record names Merck Sharp & Dohme LLC as sponsor. It does not report the planned sample-size rationale, the intended allocation ratio, details of eligibility criteria, or the full protocol. Those design details cannot be inferred from the enrollment total or the observed group sizes.

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 20-75 years of age with type 2 diabetes mellitus and inadequate glycemic control (HbA1c ≥6.5% and \<10% at Week -2) were eligible for randomization following at least 8 weeks of diet/exercise and antihyperglycemic agent (AHA) wash-off (for patients previously on an AHA), including a 2-week 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. For FAS patients with no data at Week 12, the last non-baseline observed measurement was carried forward to Week 12.

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 HbA1c at Week 12, measured over Baseline and Week 12.

The primary endpoint was change from baseline in HbA1c at Week 12, measured as a percent. The registry defines this change as the Week 12 HbA1c percent minus the Week 0 HbA1c percent. Thus, a negative change represents a lower measured value at Week 12 than at baseline, while a positive change represents a higher measured value. This describes the direction of the measurement, not a clinical recommendation.

HbA1c is used in this trial as a continuous laboratory outcome. A continuous measure can describe not only which group has a higher or lower average result, but also the size of the difference between group averages. The registry reports least-squares means for the treatment groups and ANCOVA-based least-squares mean differences for comparisons with placebo. Least-squares means are model-based group estimates; the registry record does not provide enough detail to reconstruct the adjustment variables or other model specifications.

The outcome's timeframe is baseline and Week 12. It therefore speaks to change over the treatment period recorded for this study, rather than persistence beyond that period. The registry also lists glycosylated albumin and fasting plasma glucose as secondary outcomes at Week 12. Their presence does not change which endpoint was designated primary. Results for those secondary measures are not described in the primary-outcome results.

For participants in the Full Analysis Set without a Week 12 measurement, the last non-baseline measurement was carried forward. This makes the analysis depend partly on an imputation rule rather than exclusively on measurements made at Week 12. The record does not report how many participants were affected by that rule.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The registry reports two types of inferential analysis for the primary outcome: pairwise comparisons based on ANCOVA and a Tukey linear trend test based on ANCOVA. The pairwise analyses compare each sitagliptin dose group with placebo. The trend analysis considers the set of dose groups and placebo together. No separate unadjusted recalculation is reported in the registry record for this continuous two-arm comparison.

What does ANCOVA contribute?

ANCOVA is a statistical model used to compare outcome means while accounting for specified covariates. In general, adjustment can improve precision when the covariates are relevant and the model is appropriate. The reported treatment estimates are called least-squares means, and the pairwise results are least-squares mean differences. These are model-based quantities rather than simple differences between the displayed group means. The registry record does not identify the covariates, model formula, assumptions checks, or variance estimator, so those aspects of the analysis cannot be evaluated from the reported description alone.

The group estimates summarize average change in HbA1c over the stated timeframe. The pairwise difference is the estimated contrast between a sitagliptin group and placebo. A negative difference corresponds to a lower estimated change for the sitagliptin group relative to placebo, given the endpoint's definition as Week 12 minus Week 0. Its confidence interval expresses uncertainty around that model-based contrast. A confidence interval is not a guarantee that the true value falls within the interval; it is an interval constructed by a procedure that has a stated coverage interpretation under its assumptions.

What does the trend test assess?

A linear trend test evaluates whether the results show a systematic pattern across the ordered groups, according to the test's model. The registry identifies the method as a Tukey linear trend test based on ANCOVA and reports p < 0.001. A p-value describes how compatible the observed data are with the test's null hypothesis, under the analysis assumptions. It is not the probability that the null hypothesis is true, nor does it quantify the size or clinical importance of a trend.

The registry's description does not state the exact null hypothesis, the dose coding used, or additional details of the trend model. The group estimates are therefore important context: they show the observed model-based means at the individual doses, while the trend test supplies a separate overall test of pattern. A significant overall trend does not imply that each successive dose estimate must be more negative than the preceding one.

How should the confidence intervals and p-values be read?

The pairwise results include 95% confidence intervals and p-values reported as p < 0.001 for all four dose-versus-placebo comparisons. The intervals describe the precision of the estimated contrasts under the ANCOVA model. Because each reported interval lies below zero, the estimated contrast is consistently negative across its stated interval. The interval width also communicates uncertainty, but interpretation depends on the model and the handling of missing data.

A small p-value is evidence against the null hypothesis used for that particular test, conditional on the analysis assumptions. It is not a measure of the probability that a result occurred by chance in everyday language, and it does not measure effect size. Effect size is conveyed by the contrast estimate, while the confidence interval conveys its uncertainty. The registry does not report a clinical threshold against which to judge the importance of these differences.

What role does missing-data handling play?

The Full Analysis Set included participants with a baseline value and at least one post-baseline value for the outcome. For those participants without a Week 12 value, the last non-baseline observed measurement was carried forward. This is commonly described as last observation carried forward. In general, an imputation rule fills in an unobserved outcome using an assumption about what that outcome would be; it does not turn an unobserved measurement into an observed one.

Last observation carried forward can affect group estimates and uncertainty if the carried-forward values differ systematically from the values that would have been measured at Week 12. The record does not report how often this occurred, the timing of the carried-forward measurements, reasons for missingness, or results under alternative missing-data assumptions. The consequences of this rule therefore cannot be quantified from the reported information.

What is not established by the listed methods?

The registry record does not report the full statistical analysis plan, ANCOVA covariates, diagnostics, interim analyses, or a detailed multiplicity strategy. It lists a trend test and four pairwise comparisons, but does not state how the family of tests was handled or whether the reported p-values were adjusted. The term Tukey appears in the trend-test description; that alone is not enough to infer the handling of every reported comparison. No assumptions or unreported analysis details should be inferred from the names of the methods.

Sample-size calculations based on an observed effect, when used as planning illustrations, are not observed power and do not describe the trial's original design. The registry does not report the original sample-size justification. Such illustrations depend on assumptions about the effect to detect, significance level, allocation, and statistical approximation; they should not be treated as evidence that the study was designed around its observed results.

10. Analyses Reported by the Sponsor

The primary-outcome results reported by the sponsor are ANCOVA-based analyses: least-squares mean differences for each sitagliptin dose versus placebo, with 95% confidence intervals and p-values. Those estimates are the sponsor's analysis of the trial outcome. The registry does not report a separate unadjusted recalculation for a two-arm comparison of this continuous endpoint, so there is no such recalculated estimate to set against the sponsor's results in the reported record.

The four sponsor-reported contrasts were -0.69 for 25 mg QD, -0.99 for 50 mg QD, -0.96 for 100 mg QD, and -1.04 for 200 mg QD, each compared with placebo. Each had a 95% confidence interval entirely below zero and p < 0.001. These are model-based adjusted contrasts, not a fresh subtraction of the displayed group means. Differences between a simple unadjusted calculation and an ANCOVA estimate can arise because the analyses target different quantities or use different model information; the specific reason cannot be determined without the model details.

The registry record does not flag a discrepancy between sponsor and recalculated results. In any event, the sponsor's own analysis is the relevant reported analysis when interpreting the trial's prespecified statistical results. The absence of a separately reported recalculation should not be mistaken for evidence that the analyses agree exactly or that an unadjusted estimate is unavailable in the underlying data.

11. Multiplicity, Missing Data and Analysis Populations

The registry reports five primary-outcome inferential results: a Tukey linear trend test across the five groups and four pairwise comparisons of a sitagliptin dose with placebo. Each pairwise comparison is reported with p < 0.001. Because several tests are presented, the interpretation of their p-values can depend on how multiplicity was handled. When many hypotheses are tested, the chance of at least one statistically significant finding can increase if each test is interpreted without adjustment. This is a general statistical concern, not a claim that the reported tests were or were not adjusted.

The registry record does not report a complete multiplicity strategy. It does not state whether the pairwise p-values were adjusted, how the trend test was positioned relative to those comparisons, or which testing hierarchy was used. Although the trend-test description includes the word Tukey, the record does not specify enough to conclude how multiplicity was handled across all analyses. The reported p-values should therefore be described as posted, without assuming a correction that is not documented.

The primary analysis population is defined as the Full Analysis Set: participants with a baseline value and at least one post-baseline value for the endpoint. For members of that set without Week 12 data, the last non-baseline observed measurement was carried forward. This tells readers the stated inclusion rule and missing-data method, but not how many participants were excluded for lacking qualifying measurements or how many had a value carried forward. Those counts are not reported in the registry record.

The information does not describe other analysis populations, such as a per-protocol population, or sensitivity analyses using different missing-data assumptions. It also does not report the reasons for missing Week 12 values. As a result, the effect of missingness and the robustness of the result to alternative assumptions cannot be assessed from the registry record. The record does not report interim analyses or a broader statistical analysis plan, so the timing and prespecification of all inferential decisions are not available.

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

The posted primary-outcome results show that, in the Full Analysis Set analyzed with the registry's stated method, the estimated change in HbA1c was negative in each sitagliptin dose group and positive in the placebo group. The ANCOVA-based estimated contrasts for each dose versus placebo were negative, and their reported 95% confidence intervals were also below zero. The registry reports p < 0.001 for each of these comparisons and for its linear trend test.

These findings describe the measured endpoint over baseline to Week 12. They do not show what happened beyond that timeframe, establish results for unmeasured outcomes, or demonstrate how individual participants responded. Group means and their confidence intervals summarize estimates; they do not mean every participant experienced a change matching the group average. The record lists glycosylated albumin and fasting plasma glucose as secondary outcomes, but their results are not part of the primary-outcome estimates.

Statistical evidence is not the same as clinical importance. The record does not state a threshold for a clinically important difference, and the p-values do not answer whether the size of the observed difference is meaningful for a particular person or decision. No medical advice or treatment conclusion follows from these statistical results alone.

The registry also does not provide a complete account of the ANCOVA model, multiplicity handling, or the number and pattern of missing Week 12 measurements. The use of last observation carried forward means that some analyzed values may have been imputed rather than measured at Week 12. These limitations affect how confidently the estimates can be interpreted beyond the stated analysis.

Any sample-size calculations based on observed effects are planning illustrations, not evidence of the trial's original design or observed power. The original sample-size justification is not reported in the registry record. The most direct reading is therefore limited to the reported group estimates, model-based comparisons, and stated analysis population and missing-data rule.

13. Why This Trial Matters Statistically

This trial is statistically informative as a dose-ranging, parallel-group comparison with four active doses and a placebo group. Reporting group estimates alongside placebo contrasts allows readers to distinguish the pattern of average changes from the results of individual pairwise tests. The separate trend test addresses an overall pattern across the ordered groups, while the pairwise analyses quantify dose-specific contrasts against the comparator.

The results also illustrate why effect estimates, confidence intervals, and p-values serve different purposes. The estimated difference describes the direction and size of the modeled contrast. The confidence interval communicates uncertainty around that estimate under the stated analysis. The p-value describes evidence relative to a test's null hypothesis under its assumptions. None alone establishes clinical importance, and the registry does not report a clinical importance threshold.

The group estimates are useful context for interpreting the overall trend test. The test is statistically significant as reported, but the sequence of estimates is not strictly monotone across every active dose: the 50 mg QD estimate is -0.71, and the 100 mg QD estimate is -0.69. This makes it important not to translate an overall trend result into a claim that every dose increase produced a successively larger change.

The record also makes visible some limits of registry-level interpretation. ANCOVA is named, but its covariates and other model details are absent. The Full Analysis Set and last-observation-carried-forward rule are described, but the affected participant counts are not. The multiplicity strategy is not reported. These gaps show why numerical results should be read together with analysis details, missing-data information, and the limits of what the record documents.

14. 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 gives limited detail about the statistical model. It identifies ANCOVA and least-squares mean estimates but does not list covariates, model specification, diagnostic checks, or variance assumptions. Those omissions restrict independent assessment of the adjusted estimates. The record also does not report a full statistical analysis plan or explain the multiplicity strategy for the trend test and pairwise comparisons.

Missing-data handling is described as carrying the last non-baseline observation forward for Full Analysis Set participants without Week 12 data. However, the number affected, the timing of their last measurements, and reasons for missing Week 12 values are not reported in the registry record. No sensitivity analyses using alternative missing-data approaches are described. The potential impact of the stated method on the results therefore cannot be evaluated from these details alone.

The registry reports arm sizes and primary-outcome estimates, but the record does not provide the original sample-size justification, planned allocation ratio, or detailed eligibility criteria. It does not report how masking was implemented or the randomization procedure. These design details matter when assessing how well the analysis and trial procedures support the intended comparison, and should not be inferred from the study's randomized, double-blind labels.

The primary endpoint is measured at Week 12, and no longer-term primary outcome is reported. The record lists two secondary outcomes but does not include their results in the primary-outcome information. A clinical importance threshold is also not reported. Consequently, the results support interpretation of the stated HbA1c analysis over its specified timeframe, but do not by themselves resolve questions about longer-term outcomes, broader clinical importance, or unreported measures.

15. Related Tutorials

16. Related Statistical Calculators

17. Sources

• ClinicalTrials.gov: A Study of an Investigational Drug Sitagliptin for Type 2 Diabetes Mellitus (0431-044) — NCT00127192.
• Registry record last updated 2015-06-15; status last verified 2015-05.
• Results first posted 2009-09-25.
• 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.

18. Record Summary

This completed Phase II study randomized 363 participants with type 2 diabetes mellitus in a double-blind, parallel-group design. It compared once-daily sitagliptin at 25 mg, 50 mg, 100 mg, or 200 mg with placebo over 12 weeks.

The primary endpoint was change from baseline in HbA1c at Week 12. The Full Analysis Set required a baseline value and at least one post-baseline value; the last non-baseline observation was carried forward for participants without Week 12 data. The registry reports ANCOVA-based least-squares mean comparisons with placebo and a Tukey linear trend test. Each dose-versus-placebo comparison and the trend test had p < 0.001.

The registry record does not report the full model specification, a complete multiplicity plan, the number of imputed Week 12 values, or the original sample-size justification. The reported results therefore describe the stated short-term analysis, with those methodological details remaining uncertain.