A Phase III Study of BMS-512148 (Dapagliflozin) in Asian Patients With Type 2 Diabetes Who Are Not Well Controlled on Metformin Alone: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for A Phase III Study of BMS-512148 (Dapagliflozin) in Asian Patients With Type 2 Diabetes Who Are Not Well Controlled on Metformin Alone (NCT01095666), a randomized phase 3 study. The primary outcome is Adjusted Mean Change From Baseline in Hemoglobin A1C (HbA1c) at Week 24 (Last Observation Carried Forward [LOCF]).

2. Clinical Question

3. Trial Design

The registry identifies NCT01095666 as a completed Phase III trial sponsored by AstraZeneca. It lists a randomized, quadruple-masked, parallel-group design and an enrollment of 1484. The study started in 2010-06, and primary completion was recorded as 2013-03. The official title describes a multicenter, randomized, double-blind, placebo-controlled trial of dapagliflozin in combination with metformin in Asian subjects with type 2 diabetes and inadequate glycemic control on metformin alone.

The posted primary outcome compares placebo plus metformin with dapagliflozin 5 mg plus metformin and dapagliflozin 10 mg plus metformin. The outcome-specific analysis groups contain 139, 146, and 149 participants, respectively. These analysis counts are not the trial’s overall enrollment; they describe participants included in the primary endpoint analysis. The primary population is defined as randomized participants who received study medication and had nonmissing baseline and Week 24 HbA1c values using LOCF.

The registry lists dapagliflozin, metformin, dapagliflozin placebo, and pioglitazone among interventions. However, the arm descriptions are blank, and the posted primary comparisons concern the three groups named above. The role of pioglitazone in the study is not reported in the registry record. Additional details about dose schedules, treatment duration beyond the endpoint timeframe, eligibility criteria, allocation procedures, and the exact masking arrangements are not reported in the registry record. The title and endpoint specify Week 24 assessment; this description does not establish details beyond those recorded.

The record lists five outcome measures and ten analyses. It includes the primary HbA1c outcome and secondary measures of fasting plasma glucose, post-meal glucose, body weight, and achievement of HbA1c below 7.0%. Results are posted for the first three secondary measures, while the result for achievement of HbA1c below 7.0% is not reported in the registry record. The registry’s overall enrollment should not be confused with the smaller endpoint-specific analysis populations.

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.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: All randomized participants who received study medication and had nonmissing HbA1c values at baseline and Week 24 (LOCF)

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.

Population: All randomized participants who received at least 1 dose of study medication

6. Primary Endpoint

The registry lists the primary outcome as Adjusted Mean Change From Baseline in Hemoglobin A1C (HbA1c) at Week 24 (Last Observation Carried Forward [LOCF]), measured over From Baseline to Week 24.

The primary endpoint is the adjusted mean change from baseline in hemoglobin A1c (HbA1c) at Week 24, expressed as a percentage of hemoglobin. HbA1c is a continuous measure of glycemic status. Using a change from baseline focuses the comparison on how the measure shifted over the study period, while an adjusted mean comparison can account for covariates included in the sponsor’s statistical model. The registry describes the sponsor’s analysis as ANCOVA, but does not report the full model specification or the covariates included.

Baseline was defined using the last assessment before the first dose of double-blind study medication, with a date-based fallback when timing information was unavailable. HbA1c measurements were taken during qualification and lead-in periods and at several points during the double-blind period, including Week 24. The primary endpoint description states that data after rescue medication were excluded. Week 24 values were handled using last observation carried forward (LOCF), meaning the last available measurement was carried forward for participants without a Week 24 value under the stated approach.

This endpoint is used to compare mean change across randomized groups, rather than to count how many participants met a threshold. The registry separately lists achieving HbA1c below 7.0% as a secondary outcome, but that result is not reported in the registry record. The endpoint is a measure of an intermediate clinical outcome; by itself, it does not describe long-term health outcomes or establish what happened after Week 24.

7. Results Posted in the Registry

• Dapagliflozin 5 mg + Metformin: mean -0.82 versus -0.23 on control; difference -0.59 (95% CI -0.76 to -0.42), standardized effect d = -0.80; z-test p = <0.001.
• Dapagliflozin 10 mg + Metformin: mean -0.85 versus -0.23 on control; difference -0.62 (95% CI -0.79 to -0.45), standardized effect d = -0.85; 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.

Adjusted Mean Change From Baseline in Fasting Plasma Glucose (FPG) at Week 24 (Last Observation Carried Forward [LOCF])

• 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:
  • Dapagliflozin 5 mg + Metformin: mean -21.60 versus 0.50 on control; difference -22.10 (95% CI -28.57 to -15.63), standardized effect d = -0.79; z-test p = <0.001.
  • Dapagliflozin 10 mg + Metformin: mean -26.60 versus 0.50 on control; difference -27.10 (95% CI -33.51 to -20.69), standardized effect d = -0.97; z-test p = <0.001.
• Interpretation:
  • The 95% confidence interval excludes no difference, so the mean is statistically lower for Dapagliflozin 5 mg + Metformin than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.
  • The 95% confidence interval excludes no difference, so the mean is statistically lower for Dapagliflozin 10 mg + Metformin than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.

Adjusted Mean Change From Baseline in 2-hour Post Meal Glucose (PMG) (mg/dL) at Week 24 (Last Observation Carried Forward [LOCF])

• 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:
  • Dapagliflozin 5 mg + Metformin: mean -57.80 versus -15.50 on control; difference -42.30 (95% CI -53.82 to -30.78), standardized effect d = -0.94; z-test p = <0.001.
  • Dapagliflozin 10 mg + Metformin: mean -64.60 versus -15.50 on control; difference -49.10 (95% CI -60.45 to -37.75), standardized effect d = -1.09; z-test p = <0.001.
• Interpretation:
  • The 95% confidence interval excludes no difference, so the mean is statistically lower for Dapagliflozin 5 mg + Metformin than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.
  • The 95% confidence interval excludes no difference, so the mean is statistically lower for Dapagliflozin 10 mg + Metformin than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.

Adjusted Mean Change From Baseline in Total Body Weight (kg) at Week 24 (Last Observation Carried Forward [LOCF])

• 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:
  • Dapagliflozin 5 mg + Metformin: mean -1.84 versus -0.74 on control; difference -1.10 (95% CI -1.64 to -0.56), standardized effect d = -0.47; z-test p = <0.001.
  • Dapagliflozin 10 mg + Metformin: mean -2.56 versus -0.74 on control; difference -1.82 (95% CI -2.36 to -1.28), standardized effect d = -0.78; z-test p = <0.001.
• Interpretation:
  • The 95% confidence interval excludes no difference, so the mean is statistically lower for Dapagliflozin 5 mg + Metformin than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.
  • The 95% confidence interval excludes no difference, so the mean is statistically lower for Dapagliflozin 10 mg + Metformin than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.

The posted continuous secondary results follow a broadly consistent pattern across the glucose and body-weight measures: the dapagliflozin-plus-metformin groups have more negative mean changes than the placebo-plus-metformin group. The registry reports comparisons for fasting plasma glucose, two-hour post-meal glucose, and total body weight. These outcomes provide related but distinct information alongside the primary HbA1c measure; they do not replace the primary endpoint or constitute independent confirmation of every aspect of it.

The secondary findings are directionally compatible with the primary result in the sense that the posted group mean changes favor a larger reduction for the dapagliflozin groups on these measures. This description is limited to the posted estimates. Each endpoint has its own measurement scale, analysis population, missing-data handling, and potential sources of uncertainty, so effect sizes should not be compared directly across outcomes without accounting for those differences.

The registry lists achievement of HbA1c below 7.0% as a secondary outcome, but its result is not reported in the registry record. It also lists four secondary outcomes in total, while the posted numerical secondary results cover the three continuous measures. The absence of a posted result for the threshold outcome means no conclusion about that outcome can be drawn from this record.

Secondary outcomes generally carry less evidential weight than a designated primary outcome, especially when many outcomes or comparisons are examined. Multiple testing increases the chance that at least one result appears notable by chance if no adjustment is made. The registry record does not report whether the sponsor adjusted secondary analyses for multiplicity, how outcomes were prioritized, or whether a hierarchical testing strategy was used. The secondary results are therefore best treated as supportive descriptive evidence rather than as a separate definitive test. The group-level consistency does not resolve missing multiplicity information or establish long-term effects.

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

• Adjusted Mean Change From Baseline in Fasting Plasma Glucose (FPG) at Week 24 (Last Observation Carried Forward [LOCF])
• Adjusted Mean Change From Baseline in 2-hour Post Meal Glucose (PMG) (mg/dL) at Week 24 (Last Observation Carried Forward [LOCF])
• Adjusted Mean Change From Baseline in Total Body Weight (kg) at Week 24 (Last Observation Carried Forward [LOCF])
• Participants Achieving a Therapeutic Glycemic Response (Hemoglobin A1c [HbA1C]) <7.0% at Week 24 (Last Observation Carried Forward [LOCF])

9. Statistical Interpretation

10. Statistical Methods Explained

What does the mean change represent?

HbA1c is treated as a continuous outcome, so the primary result is expressed as a group mean change rather than a count of participants meeting a criterion. Each group’s mean change compares its baseline measurement with its Week 24 measurement as handled under the endpoint definition. A negative value indicates a decrease from baseline on this scale. The difference between group means expresses how far apart the average changes are, with the comparison defined as dapagliflozin plus metformin relative to placebo plus metformin.

The group summaries are adjusted mean changes. The registry’s reported comparison method is ANCOVA, a general statistical method that models an outcome while accounting for specified covariates. In general, adjustment can improve precision or account for baseline and other modeled differences, but what adjustment achieves depends on the model and data. The registry record does not report the model covariates, assumptions checks, or full analysis specification.

What does ANCOVA contribute?

The sponsor reports mean differences from an ANCOVA analysis. This is distinct from a simple unadjusted calculation using the posted group means and standard errors. The recalculated comparisons are useful as a transparent check of the direction and approximate size of the group contrast from the summary statistics, but they do not reproduce the sponsor’s model. The sponsor’s reported estimates and confidence intervals are therefore the primary record of the sponsor’s analysis. No discrepancy is flagged in the registry-derived comparison.

The ANCOVA results report 95% confidence intervals and p-values. A confidence interval describes a range of effect estimates compatible with the observed data and analysis assumptions at the stated confidence level. A p-value assesses how compatible the observed result is with a specified null model. Neither statistic measures clinical importance, and a p-value is not the probability that the null hypothesis is true.

How should LOCF and rescue medication be understood?

The endpoint uses last observation carried forward for the Week 24 assessment. In general, LOCF substitutes an earlier available measurement for a missing endpoint measurement. It can preserve participants in an analysis when the target visit measurement is unavailable, but it relies on an assumption that the carried-forward value is a reasonable stand-in for the missing value. If measurements would have changed after the last observed visit, that assumption may not reflect the unobserved outcome. The registry does not report sensitivity analyses using alternative missing-data methods or the pattern and reasons for missingness.

The endpoint description also excludes data after rescue medication. Rescue treatment can occur when additional intervention is needed, but the registry does not report the protocol rules for rescue, how many participants received it, or the timing of such treatment. Excluding post-rescue measurements defines the data used for this endpoint; it does not show what the outcome would have been had those measurements been retained or handled differently.

What do the recalculated comparisons represent?

The recalculated contrasts are unadjusted comparisons derived from posted summary statistics. They are not the sponsor’s ANCOVA estimates and should not be interpreted as a replacement for them. The recalculated HbA1c mean differences are -0.59 and -0.62, matching the sponsor’s reported central estimates to the displayed precision. The recalculated intervals are based on summary-statistic methods and are close to, but not identical in all endpoints to, the sponsor’s reported intervals. Small differences can arise because the analyses are not identical and because displayed summaries are rounded.

The recalculated p-values are unadjusted and should be read as descriptive checks, not as the sponsor’s inferential procedure. The registry reports the sponsor’s primary comparisons as ANCOVA with p < 0.001. A complete account of whether and how the sponsor adjusted for multiple comparisons is not reported in the registry record.

Why does dispersion matter?

For a continuous outcome, the amount of variation among individual measurements affects how precisely a group mean and group contrast can be estimated. Greater variability generally makes an estimate less precise when other features are held constant; more information can improve precision. The posted group summaries include standard errors, which describe uncertainty in the estimated means. They are not standard deviations of individual measurements. The registry’s sponsor analysis provides confidence intervals for the between-group contrasts, which are more directly useful for judging the range of compatible effects.

These methods rely on assumptions about the data and model. The registry record does not report diagnostic checks, distributional assessments, or departures from model assumptions. The posted results can therefore be interpreted in the context of the stated methods, but the record does not permit a full independent evaluation of every analytical assumption.

11. Confidence Intervals and Precision

The sponsor’s 95% confidence interval for the difference in HbA1c change between dapagliflozin 5 mg plus metformin and placebo plus metformin runs from -0.76 to -0.42. For dapagliflozin 10 mg plus metformin versus placebo plus metformin, it runs from -0.79 to -0.45. Both intervals are entirely below zero, consistent with a greater mean decrease in the dapagliflozin groups under the reported ANCOVA analyses.

A confidence interval expresses uncertainty about the estimated group difference. In general, an interval gives a range of values compatible with the observed data and the analysis assumptions at its stated confidence level. It does not mean that there is a 95% probability that the particular interval contains the true effect, nor does it capture uncertainty from every possible limitation, such as missing-data assumptions or model misspecification.

The interval widths indicate that the estimates are not exact. The 5 mg and 10 mg intervals overlap substantially, but overlap is not a formal test of whether the doses differ. The registry does not report a direct comparison between doses. Similarly, the fact that each interval excludes zero does not establish that the estimated differences are clinically important; the registry record does not report a clinical-importance threshold.

The sponsor’s reported intervals are based on ANCOVA. Recalculated intervals from posted summary statistics are unadjusted approximations and are not identical in every displayed endpoint to the sponsor’s results. For the primary endpoint, the central estimates match to the displayed precision, while the sponsor’s stated intervals and p-values should be used to describe the sponsor analysis. The registry record does not report additional interval estimates under alternative missing-data methods or model specifications, so precision under those alternatives cannot be assessed from the posted information.

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 illustrations express the number of participants per arm for 80% power and the number of participants per arm for 90% power. They are planning illustrations based on the observed effect, not the original trial’s design calculations. The planning 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 set to the stated fraction of the observed effect. These illustrations do not describe the sponsor’s original sample-size justification, which is not reported in the registry record.

For a given assumed effect, the 80% power column describes the approximate per-arm sample size associated with an 80% chance of detecting an effect of that size under the stated assumptions, if that effect is real. The 90% power column gives the corresponding planning target for 90% power. Higher target power generally requires more participants under otherwise similar assumptions. These are prospective planning quantities, not a statement about the probability that the completed trial’s findings are correct.

The rows represent planning for 100%, 75%, and 50% of the observed effect. Smaller assumed effects need more participants to detect than larger assumed effects, all else equal. The observed effect can be unstable and may not be a reliable planning value for a future study, particularly when taken from a single trial. The illustrations do not include unreported features such as the original protocol’s expected dropout, multiplicity strategy, analysis model, or any other design allowance. These figures are not observed power.

The figures are participants per arm for these continuous-outcome comparisons, not total event counts. The registry’s planning note distinguishes this from time-to-event illustrations, which would express total events; that distinction does not change the interpretation of these continuous-outcome figures. The original trial’s own sample-size assumptions and justification are not reported in the registry record.

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 = -0.590; pooled standard deviation σ = 0.733.
2. n = 2 × (1.96 + 0.842)² × 0.733² / -0.590² = 24.3, rounded up to 25 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 <- -0.59                      # observed difference in means
sigma <- 0.733385446167414                     # 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(1095666)
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 = 25 (80% power), 33 (90% power)
 75% of observed effect: n per arm = 44 (80% power), 58 (90% power)
 50% of observed effect: n per arm = 98 (80% power), 130 (90% power)
power.t.test: 25.3 
Simulated power at n = 25 per arm: 0.793 (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 comparisons as ANCOVA estimates. For dapagliflozin 5 mg plus metformin versus placebo plus metformin, the sponsor reports a mean difference of -0.59, a 95% confidence interval from -0.76 to -0.42, and p < 0.001. For dapagliflozin 10 mg plus metformin versus placebo plus metformin, the reported difference is -0.62, with a 95% confidence interval from -0.79 to -0.45 and p < 0.001.

The simple unadjusted recalculations from the posted summary statistics give central estimates of -0.59 and -0.62, respectively. These agree with the sponsor’s estimates to the precision displayed. The recalculated intervals and p-values arise from a different, unadjusted approach and should not be substituted for the sponsor’s ANCOVA results. The sponsor’s analysis is the relevant description of the analysis reported by the trial record; no discrepancy is flagged.

The registry reports the method as ANCOVA but does not provide the complete model specification, including the covariates, or enough information to reproduce the sponsor analysis fully. The posted summary statistics support a basic check of direction and approximate magnitude, not an independent replication of all modeling choices. Any difference between recalculated intervals and sponsor intervals should be understood in that context rather than treated as evidence of a reporting error.

15. Multiplicity, Missing Data and Analysis Populations

The registry lists five outcome measures and ten analyses. The record therefore describes more than one statistical comparison and more than one outcome. When many hypotheses are tested, the chance of obtaining at least one small p-value by chance can increase. In general, multiplicity procedures control or otherwise address this issue by specifying how evidence across tests will be evaluated. Whether such a procedure was used for the primary or secondary analyses is not reported in the registry record.

The primary outcome has two posted comparisons, each against the same placebo-plus-metformin group. The registry reports p < 0.001 for both sponsor ANCOVA comparisons, but it does not state whether these p-values were adjusted for the two dose comparisons or for the wider set of outcomes and analyses. The posted secondary comparisons likewise do not come with a reported multiplicity strategy. Thus the reported p-values should be read as the analyses are described, without assuming that family-wise error or false-discovery concerns were addressed in a particular way.

The primary endpoint uses an analysis population defined by receipt of study medication and nonmissing baseline and Week 24 HbA1c values, with LOCF. Data after rescue medication were excluded. This differs from an analysis that necessarily includes every randomized participant regardless of outcome availability. The registry does not report the number excluded from this analysis, the reasons for missing measurements, the number receiving rescue medication, or a sensitivity analysis using a different missing-data approach.

LOCF is an imputation approach, not a direct observation of a missing Week 24 value. In general, carrying a prior value forward assumes it remains an appropriate substitute at the endpoint. If the outcome changes after the last measurement, the imputed value can misrepresent the unobserved Week 24 result. Excluding post-rescue measurements also defines the estimand through the data included; the registry does not provide enough detail to judge how rescue use affected results.

The record does not report the full statistical analysis plan, interim analyses, original sample-size justification, covariates in the ANCOVA model, or analysis diagnostics. Those omissions limit assessment of prespecification, missing-data robustness, and the broader error-control strategy. They do not negate the posted estimates, but they narrow what can be concluded from the registry record alone.

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: joint injury (0.7%), squamous cell carcinoma (0.7%), angina unstable (0.7%), myocardial bridging (0.7%), abdominal hernia (0.7%).

The registry records serious adverse events in 6 of 145 participants in the placebo-plus-metformin group, 3 of 147 in the dapagliflozin 5 mg-plus-metformin group, and 3 of 152 in the dapagliflozin 10 mg-plus-metformin group. These are the posted serious adverse event counts and the corresponding at-risk group sizes. They are descriptive counts, not adjusted comparisons, and the registry record does not report a statistical test for differences between groups.

The listed serious adverse event terms each have a reported frequency of 0.7%: joint injury, squamous cell carcinoma, unstable angina, myocardial bridging, abdominal hernia, and pneumonia. The registry’s other adverse-event counts are 34 of 145, 43 of 147, and 42 of 152 for the respective groups. The record does not provide further detail in this summary about event severity, timing, duration, or attribution.

Deaths are not reported in the registry record. The serious-event counts alone cannot establish whether an event was caused by study treatment or whether the groups differ reliably in safety. Safety interpretation depends on the definitions, follow-up, event ascertainment, and analyses, some of which are not reported in the registry record.

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

The posted primary results describe a greater mean decrease in HbA1c by Week 24 in each dapagliflozin-plus-metformin group than in the placebo-plus-metformin group, according to the sponsor’s ANCOVA analyses. The reported mean differences are -0.59 for the 5 mg comparison and -0.62 for the 10 mg comparison, with both reported 95% confidence intervals below zero and p < 0.001. These estimates apply to the defined outcome, analysis population, and handling of rescue medication and missing Week 24 values.

They do not establish a direct difference between the two dapagliflozin doses. Each dose was compared with the same control, and no direct dose-versus-dose result is reported. They also do not establish that the doses are equivalent, nor do the similar point estimates establish a dose-response pattern. The registry record does not report a clinical-importance threshold, so statistical evidence of a group difference should not be equated automatically with a judgment about the size or importance of that difference.

The result does not establish outcomes beyond the endpoint timeframe or effects on long-term complications, quality of life, or other outcomes not reported. It does not demonstrate what would have been observed without LOCF, without excluding post-rescue observations, or under a different analysis population. Those questions require evidence not reported in the registry record.

Nor do the findings alone settle the overall benefit-risk balance. Serious adverse event counts are descriptive, deaths are not reported in the registry record, and the available summary does not supply the full safety analyses or event details. The record supports a focused statistical description of Week 24 HbA1c and related posted measures, not individualized clinical conclusions or advice.

More generally, a randomized design can support a causal comparison when the design, conduct, analysis, and assumptions are appropriate. The registry identifies randomization and masking, but does not report enough information to assess every element of implementation. Interpretation should therefore remain tied to the reported trial and its documented limitations rather than extended to other populations, settings, or treatment regimens.

18. Why This Trial Matters Statistically

This trial illustrates how a clearly defined continuous primary endpoint can be reported through group mean changes, between-group estimates, confidence intervals, and p-values. The separate group summaries show the scale and direction of change, while the sponsor’s ANCOVA contrasts address the randomized comparisons. Reading both kinds of information helps distinguish what happened within groups from how the groups differed.

It also illustrates why the analysis population and outcome handling matter. The primary result is based on participants who received study medication and had baseline and Week 24 HbA1c values under LOCF, with post-rescue data excluded. Those choices define which data contribute to the estimate. Without reporting missingness patterns, rescue use, and sensitivity analyses, readers cannot determine how robust the estimate is to alternative choices.

The trial has two dose-specific comparisons against one control, alongside several secondary measures. This makes multiplicity an important interpretive issue: a collection of analyses cannot automatically be read as a single test with one error rate. The registry does not report how multiplicity was handled, which limits the strength of conclusions drawn from multiple nominal p-values.

Finally, the record distinguishes sponsor analyses from unadjusted recalculations based on summary statistics. Agreement in the primary central estimates provides a useful consistency check, but does not make the simple recalculation equivalent to ANCOVA. Statistical transparency requires attention to the analysis method, precision, missing-data choices, multiplicity, and the limits of summary-level information—not just whether a p-value is small.

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.

Several limitations are specific to what the registry reports for this trial. Although enrollment is recorded as 1484, the primary HbA1c analysis includes outcome-specific groups of 139, 146, and 149 participants. The registry does not explain the difference between overall enrollment and the primary analysis counts, give the number excluded, or describe reasons for absent baseline or Week 24 values. The primary result therefore applies to its stated analysis population, not necessarily to all randomized participants.

The endpoint uses LOCF and excludes observations after rescue medication. The record does not report how many participants required rescue, when it occurred, or results using alternative approaches. These choices could affect the interpretation of the estimated Week 24 changes, but their impact cannot be quantified from the registry information.

The statistical model is described as ANCOVA, but the covariates and full model specification are not reported. The registry also does not report diagnostic checks or a complete statistical analysis plan. It does not state whether multiplicity adjustments were made across the dose comparisons and outcomes, and no original sample-size justification is reported. Planning illustrations based on observed effects do not fill that gap.

The registry lists pioglitazone as an intervention, but the posted primary comparisons concern placebo plus metformin and the two dapagliflozin doses; arm descriptions are blank. The role of pioglitazone is not reported in the registry record. In addition, a secondary outcome for achieving HbA1c below 7.0% is listed without a posted result. Death counts are also not reported. These gaps constrain interpretation of the broader efficacy and safety record.

20. Related Tutorials

21. Related Statistical Calculators

22. Sources

• ClinicalTrials.gov: A Phase III Study of BMS-512148 (Dapagliflozin) in Asian Patients With Type 2 Diabetes Who Are Not Well Controlled on Metformin Alone — NCT01095666.
• Registry record last updated 2017-09-11; status last verified 2017-09.
• Results first posted 2017-09-11.
• 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

NCT01095666 is a completed randomized, quadruple-masked, parallel-group Phase III trial sponsored by AstraZeneca, with enrollment of 1484. It evaluated dapagliflozin in Asian patients with type 2 diabetes whose control was inadequate on metformin alone. The primary endpoint was adjusted mean change in HbA1c from baseline to Week 24, using LOCF and excluding data after rescue medication.

The sponsor’s ANCOVA results report mean differences of -0.59 for dapagliflozin 5 mg plus metformin versus placebo plus metformin and -0.62 for dapagliflozin 10 mg plus metformin versus placebo plus metformin. Both reported 95% confidence intervals are below zero, with p < 0.001. Results are also posted for fasting plasma glucose, post-meal glucose, and body weight. The registry does not report the result for the listed HbA1c threshold outcome below 7.0%.

Interpretation is limited by unreported details on multiplicity, missing data, rescue use, model specification, and the role of pioglitazone. The results describe the posted analyses and do not establish long-term outcomes or conclusions beyond the measures and timeframe recorded.