Safety and Efficacy of Saxagliptin Plus Insulin With or Without Metformin: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Safety and Efficacy of Saxagliptin Plus Insulin With or Without Metformin (NCT00757588), a randomized phase 3 study. The primary outcome is Adjusted Mean Change From Baseline in A1C Levels (Last Observation Carried Forward [LOCF]).

2. Clinical Question

3. Trial Design

This was a randomized, double-blind, parallel-group Phase 3 trial. The official title describes it as multicenter. The registry lists an enrollment of 455 participants and a completed status. The study began in 2008-11, with primary completion in 2010-04. These registry entries describe the trial structure and record status; they do not by themselves establish how participants were recruited, how the study was monitored, or how its original sample size was chosen. Those details are not reported in the registry record.

The intervention was saxagliptin, 5 mg, added to insulin. The comparison group received placebo added to insulin. Participants could be using insulin alone or insulin together with metformin, and metformin use was included as an adjustment factor in the primary ANCOVA model. The record does not describe separate treatment estimates for the insulin-only and insulin-plus-metformin subgroups, so the posted primary estimate should be read as the reported overall comparison, adjusted for metformin use.

The trial’s stated purpose was to compare the effects of saxagliptin and placebo as add-on therapy to insulin-based treatment on glycemic control at 24 and 52 weeks. The primary outcome assessed adjusted mean change in A1C from baseline to Week 24, using last observation carried forward. The registry also lists outcomes related to postprandial glucose during a meal tolerance test, fasting plasma glucose, therapeutic glycemic response, and insulin dose. The record describes 14 outcome measures and 5 analyses, but does not give a complete statistical analysis plan in the information reported.

The primary analysis population was restricted to randomized participants who received at least 1 dose of double-blind medication and had both baseline and postrandomization measurement data in the 24- and 52-week periods. This eligibility definition matters when interpreting the result: it is not simply a summary of every participant enrolled. The number of participants contributing to an outcome can vary, and the posted outcome counts should be interpreted in the context of each endpoint’s analysis population.

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: Of 455 participants randomized, 402 received treatment and completed the 24-week phase. Two participants were mistakenly identified as not completing the short-term (ST) treatment period, although they did. The discrepancy was identified after the ST phase database lock. In reality, 404 completed the ST phase.

Period 1: Short-term (24-week) Phase

Reasons for not completing this period:

Period 2: Long-term (52-week) Phase

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: Randomized participants who received at least 1 dose of double-blind study medication. Participants must also have had both a baseline and at least 1 postrandomization measurement in the 24- and 52-week periods.

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 Adjusted Mean Change From Baseline in A1C Levels (Last Observation Carried Forward [LOCF]), measured over Baseline to Week 24.

A1C is a measure of glycemic control, and the primary outcome was the adjusted mean change in A1C levels from baseline to Week 24. The registry gives the unit as percentage of change and describes the outcome as a change from baseline, calculated as post-treatment minus baseline. A negative value therefore represents a decrease from baseline in the measure as recorded. The endpoint compares the average change in the saxagliptin-plus-insulin group with the average change in the placebo-plus-insulin group.

The outcome is adjusted rather than a simple comparison of raw group means. The registry describes an ANCOVA model that includes baseline values, treatment, and metformin use. Adjustment for baseline is a way to account for the starting measurement when estimating the treatment-group contrast; inclusion of metformin use accounts for that recorded treatment characteristic in the model. The registry does not report further detail about the model’s assumptions, diagnostics, or any additional covariates.

Using a continuous outcome preserves information about the magnitude of change across participants, rather than reducing the result to whether each person did or did not meet a response threshold. The primary result is an estimate of an average difference between groups, not a description of the change experienced by every participant. It also concerns the Week 24 measurement and should not be treated as a result for every listed follow-up time. The registry’s stated purpose includes assessment at 24 and 52 weeks, but the primary outcome specified is baseline to Week 24.

7. Results Posted in the Registry

• Saxagliptin, 5 mg + Insulin: mean -0.73 versus -0.32 on control; difference -0.41 (95% CI -0.59 to -0.23), standardized effect d = -0.44; 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 From Baseline in Postprandial Glucose (PPG) Area Under the Curve (AUC) Response to an Meal Tolerance Test (MTT)

• 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 errors.
• Result:
  • Saxagliptin, 5 mg + Insulin: mean -4,548 versus -718.80 on control; difference -3,830 (95% CI -6,179 to -1,481), standardized effect d = -0.35; z-test p = 0.001.
• Interpretation:
  • The 95% confidence interval excludes no difference, so the mean is statistically lower for Saxagliptin, 5 mg + Insulin than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.

Change From Baseline in 120-minute PPG Values During an MTT

• 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 errors.
• Result:
  • Saxagliptin, 5 mg + Insulin: mean -27.20 versus -4.20 on control; difference -23.00 (95% CI -37.65 to -8.35), standardized effect d = -0.33; z-test p = 0.002.
• Interpretation:
  • The 95% confidence interval excludes no difference, so the mean is statistically lower for Saxagliptin, 5 mg + Insulin than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.

Change From Baseline in Fasting Plasma Glucose Values

• 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 errors.
• Result:
  • Saxagliptin, 5 mg + Insulin: mean -10.10 versus -6.10 on control; difference -4.00 (95% CI -13.62 to 5.62), standardized effect d = -0.08; z-test p = 0.415.
• Interpretation:
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between Saxagliptin, 5 mg + Insulin and the control. The interval gives the range of effects that remain compatible with the data.

The secondary outcomes cover related aspects of glucose response and treatment, including postprandial glucose during a meal tolerance test, fasting plasma glucose, therapeutic glycemic response, and insulin dose. Among the outcomes with numerical continuous estimates, the two postprandial measures show larger decreases in the saxagliptin group than in the placebo group. The fasting plasma glucose contrast is less conclusive: its reported confidence interval includes zero, and its posted p-value does not indicate a clear group difference under the unadjusted comparison. The registry also lists other secondary outcomes, but the available numerical results do not provide a basis for describing every listed measure.

These results are broadly related to the primary question because they assess other dimensions of glycemic response, while the primary endpoint is change in A1C. Agreement in direction across related measures can add context, but it does not make them interchangeable with A1C or establish that each represents the same underlying effect. Differences in timing, measurement scale, analysis population, and precision matter when comparing endpoints.

The secondary results carry less inferential weight than the primary result because multiple outcomes are assessed, creating more opportunities for apparently notable findings. The registry record does not report a multiplicity adjustment or a complete hierarchy governing secondary tests. Without such information, individual secondary p-values should be treated cautiously and not read as if each were a single isolated primary test. The posted secondary findings are best viewed as contextual evidence, not as independent confirmation of the primary result or as a substitute for a prespecified analysis plan.

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

• Change From Baseline in Postprandial Glucose (PPG) Area Under the Curve (AUC) Response to an Meal Tolerance Test (MTT)
• Change From Baseline in 120-minute PPG Values During an MTT
• Change From Baseline in Fasting Plasma Glucose Values
• Percentage of Participants Achieving a Therapeutic Glycemic Response
• Change From Baseline in Mean Total Daily Dose of Insulin (MTDDI) (LOCF)

9. Statistical Interpretation

10. Statistical Methods Explained

The primary outcome is continuous: it is a mean change in A1C, rather than a count or time-to-event measure. For a continuous endpoint, the treatment contrast is expressed in the same units as the outcome. The posted contrast of -0.41 is the difference between the group mean changes as recorded, with the intervention group compared with the placebo group. Its direction is interpreted in light of the outcome’s definition as post-treatment minus baseline.

What does adjustment do?

The registry describes the primary analysis as ANCOVA and specifies adjustment for baseline value and metformin use. In general, analysis of covariance estimates a group contrast while accounting for one or more measured variables included in the model. Baseline adjustment can improve comparability in the estimate by incorporating where participants started on the outcome; it does not change the trial’s randomized allocation. Including metformin use means that this factor is also part of the reported model. The record does not report the coding of metformin use, the treatment of any other covariates, or model diagnostics.

The primary outcome description gives the model as the difference between the Week t and baseline values modeled with baseline values, treatment, and metformin use. The stated primary timeframe is baseline to Week 24. The registry reports a separate analysis result described as a mean difference in final values. Readers should distinguish the reported adjusted analysis from a simple calculation using the group summary means. A simple calculation can be useful as a check on the direction and approximate size of a contrast, but it does not reproduce the covariate adjustment or necessarily the exact analysis population and model used by the sponsor.

What does last observation carried forward mean?

Last observation carried forward, abbreviated LOCF in the registry, is a method for handling a missing later measurement by carrying forward an earlier observed value. In general, this approach makes an assumption about how an unobserved outcome relates to a participant’s last measured outcome. It can produce a complete analysis dataset under that rule, but it does not establish that an individual’s unobserved value would actually have remained unchanged. The registry identifies LOCF in the primary outcome but does not report the timing or amount of missing data, the reasons measurements were missing, or sensitivity analyses using alternative missing-data assumptions. Those details are not reported in the registry record.

The stated primary analysis population required baseline and at least 1 postrandomization measurement in the 24- and 52-week periods, in addition to randomization and receipt of at least 1 dose of double-blind medication. This criterion defines who could contribute to the specified analysis; it is not the same thing as saying every randomized participant contributed an observed Week 24 value. The record does not provide enough detail to establish how the population rule interacted with LOCF in every case.

How should the p-value be read?

In general, a p-value measures how unusual data at least as discrepant as those observed would be if a stated null model were correct, subject to the model’s assumptions. The sponsor’s reported primary p-value is < 0.001. It does not quantify the probability that the null hypothesis is true, the probability of a future result, or the chance that the finding is due to a particular cause. It also does not indicate whether the size of the estimated difference is important to participants or clinicians.

The recalculated p-value is an unadjusted re-calculation from the posted summary statistics. It is not the same as the sponsor’s covariate-adjusted analysis. Its role is to provide a simple statistical cross-check; it should not supersede the sponsor’s reported analysis. Since the record indicates no discrepancy, the rounded estimates are broadly consistent, but their underlying methods remain distinct.

Why can a mean have a standard error?

The registry gives the primary arm summaries as means with standard errors. In general, a standard error describes the precision of an estimated mean, rather than the spread of individual participant measurements. A standard error is not interchangeable with a standard deviation: the latter describes variation among observations, while the former describes uncertainty in an estimated statistic under the analysis assumptions. The registry does not report standard deviations for the primary arm summaries, and they should not be inferred from the standard errors without additional information.

Confidence intervals add a range of values compatible with the estimate under the analysis method and its assumptions. A narrower interval indicates greater precision on that scale, while a wider interval indicates more uncertainty. The primary interval is reported from the sponsor’s ANCOVA analysis. The recalculated interval is based on posted summary statistics and is a separate unadjusted calculation. In general, estimates and intervals should be interpreted together: the point estimate gives a central contrast, while the interval conveys its precision and the range of effects not ruled out by the analysis.

The record does not report whether there were interim analyses, how the primary endpoint was selected relative to other endpoints, or whether any model assumptions were assessed. Such details are relevant to a full appraisal of statistical methods but cannot be inferred from the reported outcome result. The trial’s original sample-size justification and design assumptions are also not reported in the registry record.

11. Confidence Intervals and Precision

The sponsor’s reported primary estimate is a mean difference of -0.41, with a 95% confidence interval from -0.59 to -0.24. The interval communicates uncertainty around the estimated average contrast: values nearer either endpoint remain compatible with the data under the analysis and its assumptions. It does not mean there is a 95% probability that the true effect lies in this particular interval. Under the usual repeated-sampling interpretation, the confidence procedure is designed to cover the true parameter at the stated rate across repeated applications when its assumptions hold.

The primary interval does not include zero. That is consistent with the sponsor’s reported p-value of < 0.001 for a null comparison of no difference, although a confidence interval and p-value express related evidence in different forms. Neither statistic indicates whether the estimated difference is clinically important. The registry does not provide a threshold for a clinically meaningful A1C difference, so clinical importance cannot be determined from the estimate alone.

The recalculation from summary statistics also reports a rounded mean difference of -0.41 and a 95% confidence interval from -0.59 to -0.24. Matching rounded values do not make the procedures identical: the recalculation is unadjusted, while the sponsor’s result is ANCOVA-adjusted. The sponsor’s analysis remains the reference for the trial’s reported result.

For the secondary outcomes, interval widths differ and some intervals include zero while others do not. This variation reflects differences in estimated contrasts and precision, as well as the measures’ distinct units. These confidence intervals should not be compared by their raw numerical widths across outcomes measured on different scales. The reported summaries also do not establish how uncertainty from considering multiple outcomes should be handled; a multiplicity adjustment is not reported in the registry record.

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 sample-size figures are planning illustrations based on the observed effect, not the trial’s original design calculation. For these continuous outcomes, the figures are participants per arm, not total event counts. The planning assumptions are a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation. The calculation uses the stated fraction of the observed effect as the effect to detect. The registry record does not report the original sample-size assumptions or justification.

The 80% power column gives the estimated participants per arm for a hypothetical study with an 80% chance of detecting the specified effect under the planning assumptions, if that effect and the calculation assumptions were correct. The 90% power column gives the corresponding planning figure for a 90% chance. These are prospective planning concepts: they describe a hypothetical design, not the probability that the completed trial’s finding is correct or the chance that its observed result occurred by chance. This is not observed power.

Rows that plan for 100%, 75%, or 50% of the observed effect use progressively smaller target effects. A smaller effect is harder to distinguish from random variation with the same design assumptions, so the planning illustration calls for more participants per arm. This relationship is general for the stated calculation, not a guarantee that a future trial will achieve its target power. The observed estimate may itself be uncertain, and the true effect in another study may differ.

The planning assumptions specify equal allocation, while the posted analysis groups do not have equal outcome counts. That is not a contradiction: the illustration describes a hypothetical design, whereas the trial’s reported counts reflect the posted analysis data. The sample-size values should not be interpreted as recommendations for a new study without considering its purpose, target effect, outcome variability, missing-data expectations, multiplicity strategy, and other design choices. Those choices are not fully reported for the original trial.

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 = -0.410; pooled standard deviation σ = 0.925.
2. n = 2 × (1.96 + 0.842)² × 0.925² / -0.410² = 79.9, rounded up to 80 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 <- -0.41                      # observed difference in means
sigma <- 0.924827757105721                     # 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(757588)
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 = 80 (80% power), 107 (90% power)
 75% of observed effect: n per arm = 142 (80% power), 191 (90% power)
 50% of observed effect: n per arm = 320 (80% power), 428 (90% power)
power.t.test: 80.8 
Simulated power at n = 80 per arm: 0.804 (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 registry reports the sponsor’s primary analysis as ANCOVA, with a mean difference in final values of -0.41, a 95% confidence interval from -0.59 to -0.24, and p < 0.001. The simple recalculation from posted group summaries gives a mean difference of -0.41, a 95% confidence interval from -0.59 to -0.24, and an unadjusted p-value of <0.001. The rounded estimates and intervals are close, and the registry record indicates no discrepancy.

The match at the displayed precision should not obscure the distinction in methods. The sponsor’s estimate comes from its reported ANCOVA model, adjusted for baseline value and metformin use. The recalculation is an unadjusted comparison derived from the posted summary statistics. It does not reproduce the sponsor’s model and is not an alternative official trial result. Where the two approaches differ, the sponsor’s own analysis takes precedence.

The reported sponsor analysis is the appropriate reference for describing the trial’s primary inferential result. The recalculation can serve as a transparent check on the broad direction and scale of the contrast, but it cannot validate unreported model details or resolve questions about missing data, analysis populations, or multiplicity. The record does not report additional information needed to assess those issues.

15. Multiplicity, Missing Data and Analysis Populations

The registry describes a primary A1C outcome and multiple secondary outcomes, but a multiplicity strategy is not reported in the registry record. Multiplicity arises when several hypotheses or outcomes are tested: even if each test is considered separately, the chance of at least one seemingly noteworthy result can increase across a collection of tests. In general, a prespecified testing sequence or adjustment can help control that issue. No such procedure should be assumed in the absence of a reported plan.

This limitation is particularly relevant to interpretation of the secondary results. Their estimates can describe patterns in the posted data, but unadjusted p-values do not account for the fact that multiple outcomes are considered. The record does not state whether the secondary outcomes were intended as confirmatory, exploratory, or governed by a formal testing hierarchy. The number of outcome measures and analyses recorded does not, by itself, establish how multiplicity was handled.

The primary endpoint is identified, and its reported analysis is described as ANCOVA with baseline value and metformin use included. However, the full statistical analysis plan is not reported in the registry record. Details such as prespecification of all modeling choices, interim analyses, treatment of protocol deviations, and assessment of model assumptions are not reported in the registry record. They should not be inferred from the trial’s completed status or from the posted p-value.

Missing data also affect interpretation. The primary outcome uses last observation carried forward, and the analysis population requires baseline and postrandomization measurements in the 24- and 52-week periods. In general, missing-data methods rely on assumptions about unobserved outcomes. The registry does not report the amount or pattern of missingness, reasons for missing measurements, or sensitivity analyses examining alternative assumptions. Consequently, the robustness of the primary result to different missing-data approaches cannot be assessed from the record alone.

The described primary population is not identical to an unrestricted summary of all enrolled participants: participants had to be randomized, receive at least 1 dose, and meet the stated measurement criteria. The record does not provide enough detail to determine the impact of these criteria on the estimate or to compare the result with a different analysis population. These omissions limit a complete appraisal of both multiplicity and missing-data handling.

16. 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: abscess limb (1.3%), peripheral ischaemia (0.7%), peripheral vascular disorder (0.7%), jaundice (0.7%), cholangitis (0.7%).

The registry reports serious adverse events for the placebo-plus-insulin group as 13 affected participants among 151 at risk, and for the saxagliptin-plus-insulin group as 25 affected participants among 304 at risk. It also reports other adverse events for 64 of 151 and 96 of 304, respectively. These are counts within the stated groups and denominators; they should not be confused with the primary outcome analysis counts, which are defined for a different endpoint and analysis population.

Among the listed serious-event terms, the registry reports abscess limb at 1.3%, and peripheral ischaemia, peripheral vascular disorder, jaundice, cholangitis, and cholangitis acute at 0.7% each. These terms describe events recorded in the registry; the listed percentages do not establish that an event was caused by a study treatment. The record does not report deaths in the available safety summary, so death counts are not reported in the registry record.

Adverse-event counts alone do not establish a comparative safety profile. Interpretation depends on event definitions, observation time, severity, relatedness assessments, and the completeness of reporting. The registry record does not provide those details in the information. The counts should therefore be read as descriptive safety summaries, not as proof of a causal difference between groups.

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

The primary estimate describes a more negative adjusted mean change in A1C in the saxagliptin-plus-insulin group than in the placebo-plus-insulin group from baseline to Week 24. The sponsor’s confidence interval is entirely below zero, and its reported p-value is small under the stated model. Together, these results provide evidence of a difference in average A1C change between the randomized groups under the analysis used.

That statistical result does not establish the importance of the difference to patients. The registry record does not report a threshold for clinical importance, nor does the mean contrast show how effects were distributed across individuals. A group average cannot tell whether every participant changed in the same way, and it does not describe the likelihood of a particular outcome for an individual.

The findings are specific to the comparison, population definition, outcome, analysis method, and timeframe recorded. The intervention was saxagliptin, 5 mg, added to insulin; the comparator was placebo added to insulin. The primary result concerns baseline to Week 24, and the primary population required the stated treatment and measurement criteria. It should not be generalized automatically to different treatment settings, populations, endpoints, or follow-up periods.

The result also does not prove that the treatment caused every observed difference without qualification. Randomization supports a group comparison, but interpretation still depends on conduct of the trial, analysis choices, missing-data handling, and outcome reporting. The registry identifies the study as randomized and double-blind but does not report every detail needed to assess those elements fully.

Secondary outcomes offer additional context but have less weight when multiplicity handling is not reported. Safety counts are descriptive and do not establish causation. Taken together, the record supports a carefully bounded reading of the reported A1C contrast, while leaving questions about clinical importance, missingness, multiplicity, and broader applicability unresolved.

18. Why This Trial Matters Statistically

This trial illustrates why a treatment comparison should be read through its estimand, model, and analysis population rather than through a p-value alone. The primary outcome is a continuous mean change from baseline, analyzed with adjustment for baseline values and metformin use. Those details define what the reported contrast represents and distinguish it from a raw comparison of group means.

It also demonstrates how a registry can present both an adjusted sponsor analysis and enough group-level summaries to support a simple recalculation. The rounded primary contrast and interval are close across these approaches, but the methods are not identical. Keeping that distinction visible helps readers use recalculations as checks without substituting them for the trial’s own analysis.

Precision and multiplicity matter alongside direction. The primary interval is on one side of zero, whereas not all secondary results have the same degree of certainty. The secondary outcomes concern related but distinct measures, and the registry does not report whether a multiplicity procedure was used. Reading each secondary p-value as an independent confirmation would overlook that broader testing context.

Finally, the primary result depends on the defined analysis population and LOCF handling of missing measurements. Neither the amount of missing data nor sensitivity analyses are reported. This makes the record a useful example of how a clear central estimate can coexist with important uncertainties about robustness and interpretation. Statistical evidence is strongest when the estimate, its precision, the analysis assumptions, and the scope of available reporting are considered together.

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 primary analysis uses last observation carried forward, but the registry does not report how many measurements were missing, when they were missing, why they were missing, or whether alternative missing-data analyses were performed. Because LOCF relies on an assumption about unobserved later values, the robustness of the estimate under other assumptions cannot be assessed from the record.

The analysis population requires randomization, receipt of at least 1 dose of study medication, and baseline and postrandomization measurements in the 24- and 52-week periods. This restriction is important when interpreting the result, and the record does not describe how estimates would compare under other population definitions. Posted counts differ across outcomes, so one endpoint’s analyzed group should not be assumed to be the same as another’s.

The registry reports an ANCOVA model adjusted for baseline value and metformin use, but does not report model diagnostics, additional modeling details, or the full statistical analysis plan. A multiplicity adjustment for secondary outcomes is not reported. Interim analyses, the original sample-size justification, and the protocol’s detailed assumptions are not reported in the registry record.

The primary endpoint is assessed from baseline to Week 24. Although the stated study purpose includes assessment at 24 and 52 weeks, the primary result discussed is not evidence about every time point. The registry’s outcome summaries also do not establish clinical importance, since a threshold for a meaningful A1C difference is not reported.

Safety summaries give event counts and selected serious-event terms, but the available record does not provide enough detail to assess duration of exposure, relatedness, severity criteria, or comparative causal risk. The secondary results are also difficult to interpret as confirmatory evidence without a reported multiplicity plan. These limitations do not negate the reported estimates; they define the questions that remain open when interpreting them.

20. Related Tutorials

21. Related Statistical Calculators

22. Sources

• ClinicalTrials.gov: Safety and Efficacy of Saxagliptin Plus Insulin With or Without Metformin — NCT00757588.
• Registry record last updated 2015-06-01; status last verified 2015-05.
• Results first posted 2011-12-13.
• 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 describes a completed, randomized, double-blind, parallel-group Phase 3 trial of saxagliptin, 5 mg, added to insulin versus placebo added to insulin in participants with type 2 diabetes and inadequate glycemic control on insulin alone or insulin with metformin. Enrollment was 455. The primary outcome was adjusted mean change in A1C from baseline to Week 24, using LOCF and an ANCOVA model adjusted for baseline value and metformin use.

The sponsor reports a mean difference of -0.41, with a 95% confidence interval from -0.59 to -0.24 and p < 0.001. The simple recalculation from posted summaries is close at the displayed precision but is unadjusted; the sponsor’s analysis is the reference result. Secondary measures provide additional context, while missing-data details and multiplicity handling are not reported in the registry record.