Exubera vs Lispro in a Lantus-based Regimen for Improved Glycemic Control in Type 2 Diabetes: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Exubera vs Lispro in a Lantus-based Regimen for Improved Glycemic Control in Type 2 Diabetes (NCT00348374), a randomized phase 3 study. The primary outcome is Change From Baseline in Glycosylated Hemoglobin A1c (HbA1c) at End of Treatment.

2. Clinical Question

3. Trial Design

This was a Phase 3b, Phase 3, randomized, open-label, parallel-group, multicenter trial. The registry record lists enrollment of 191 and marks the study completed. The listed start was 2006-06, and primary completion was 2008-08. The sponsor was Pfizer. The study compared Exubera with insulin lispro when each was added to an existing Lantus-based regimen in patients with type 2 diabetes whose control was described as suboptimal.

Both treatment descriptions specify a weight-based initiation dose and individual dose adjustment according to blood glucose, alongside insulin glargine and oral agents. The trial summary describes dose titrations intended to allow a large proportion of subjects to reach target HbA1c levels. That statement describes the study’s stated intent; the registry data do not establish whether the targets were reached by any particular proportion.

Masking was listed as none, so the trial was open-label. The registry record does not report further details about how the randomization sequence was generated, how allocation was concealed, or whether outcome assessors were masked. It also does not report the intended allocation ratio. The parallel-group structure means the arms were followed as separate groups rather than in a crossover sequence.

The primary outcome was change in HbA1c at Week 24. The registry also lists secondary outcomes involving HbA1c targets, glucose measurements, lipids, and cardiovascular risk markers. While the record names these outcome topics, detailed results for all of them are not included in the registry record for this account. The original trial’s sample-size assumptions and statistical analysis plan are not reported in the registry record.

4. Participant Flow and Accounting

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

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: Full analysis set (FAS): subjects who received at least one dose of study medication, had a baseline glycosylated hemoglobin A1c (HbA1c%) measurement, and had a post-baseline HbA1C measurement; last (post-baseline) observation carried forward (LOCF). Number of subjects with HbA1c values at Baseline and Week 24: Exubera® n=81, Lispro n=90.

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 Glycosylated Hemoglobin A1c (HbA1c) at End of Treatment, measured over Baseline, Week 24 (End of Treatment).

The primary endpoint was mean change from baseline in glycosylated hemoglobin A1c (HbA1c), measured at Week 24, the stated end of treatment. The registry defines change as the mean value at Week 24 minus the mean value at baseline. The unit is percent. A negative change therefore represents a lower mean HbA1c at Week 24 than at baseline; comparing the arms asks whether those mean changes differed.

HbA1c is used as a measure of glycemic control over time. In general, a continuous endpoint retains information about the magnitude of measured change, rather than reducing every participant’s result to whether a chosen threshold was reached. The registry lists secondary analyses of HbA1c target attainment as well, but those threshold outcomes address a different question from the continuous primary endpoint.

The posted means were -1.4 for Exubera and -1.6 for insulin lispro. The unadjusted recalculated difference was defined as lispro minus Exubera and was -0.2. This is a comparison of group means, not a statement about each participant’s response. The registry record’s endpoint definition identifies the Week 24 time point and the change calculation, but it does not report additional details about the measurement schedule or the handling of particular clinical events beyond the stated last-observation-carried-forward approach for the full analysis set.

7. Results Posted in the Registry

• Insulin Lispro: mean -1.60 versus -1.40 on control; difference -0.20 (95% CI -0.55 to 0.15), standardized effect d = -0.17; z-test p = 0.263.

8. Statistical Interpretation

9. Statistical Methods Explained

The main statistical comparison concerns a continuous outcome: change in HbA1c from baseline to Week 24. For a continuous outcome, each participant’s change can take many values, and a group mean summarizes the center of those values. The posted primary summaries give an arm mean and standard deviation for each group. In general, the standard deviation describes how dispersed individual observations are around the group mean. It is not the same as the standard error, which describes uncertainty in an estimated mean.

What does the mean difference represent?

The comparison is expressed as insulin lispro minus Exubera. Subtracting one group mean from the other gives the unadjusted difference of -0.2. Because the endpoint is a change score, a more negative value denotes a larger decrease from baseline in the group average. The sign of a difference depends on the order of subtraction; it does not assign a value judgment to either treatment. A mean difference also does not describe the distribution of individual responses, and group averages can conceal substantial variation.

How should the uncertainty estimate be understood?

A confidence interval gives a range of values consistent with the data and the calculation’s assumptions at the stated confidence level. In general, an interval that includes zero leaves open the possibility of no difference in the underlying group means, while also indicating other values compatible with the data. The recalculated interval for this comparison includes zero and has limits on both the negative and positive sides. Its width reflects uncertainty in the estimated difference; the interval is not a range containing a fixed proportion of individual participant outcomes.

The interval and p-value are recalculated from the posted summary statistics using an unadjusted comparison. The registry record does not report the sponsor’s own primary-comparison model or its inferential result in the information. Consequently, the recalculation should be kept distinct from any trial analysis that may have been performed but is not reported in the registry record. No discrepancy is identified between a sponsor result and this recalculation in the recorded details.

What does the p-value tell the reader?

The unadjusted p-value is 0.263. In general, a p-value evaluates how unusual results at least as discordant with a null hypothesis would be if that null model and its assumptions held. It is not a direct measure of effect magnitude, clinical importance, or the probability that a result arose by chance. It does not prove that the arms are equivalent. Equivalence or noninferiority requires an appropriate prespecified margin and analysis; neither is reported in the registry record.

Why does variation matter?

The posted standard deviations, 1.3 and 1.0, show that participant-level changes varied around the corresponding arm means. In general, more variability makes a group mean less precise when the sample size is fixed. If data are more dispersed than the statistical model assumes, standard errors and confidence intervals can be misleading unless the analysis accounts for that feature. The registry does not report over-dispersion as an issue for this continuous endpoint; this is a general principle, not a claim about the trial.

The standardized mean difference is another way to express a mean contrast relative to the variability in the groups. It can help compare effects measured on different scales, but it is less directly interpretable than the HbA1c percentage-point change for this endpoint. The posted recalculation includes a pooled standard deviation of 1.152; the registry record does not give a clinical interpretation threshold for the standardized effect. Standardization does not remove uncertainty or correct for bias.

What assumptions are not established?

An unadjusted comparison of group means has assumptions about the independence of observations and the behavior of the outcome distribution, especially for a small-sample approximation. The registry record does not report diagnostic checks, a detailed model specification, a covariate-adjusted analysis, or sensitivity analyses. It also does not report the original protocol’s sample-size justification, interim analyses, or the trial’s full statistical analysis plan. These omissions prevent reconstruction of the complete sponsor analysis from the registry summary alone.

Randomization is intended to support a fair comparison by assigning participants to groups, but the registry entries do not describe the randomization procedure or concealment. Open-label status means treatment assignment was not masked. In general, lack of masking can affect care, behavior, or outcome assessment, although an objective laboratory measurement may be less subjective than a self-reported outcome. The record does not provide information to quantify any such influence in this trial.

The full analysis set definition and last-observation-carried-forward method are also part of the statistical interpretation. Last observation carried forward places the most recent post-baseline measurement in place of a later missing measurement. In general, this assumes that the carried-forward value is a reasonable substitute for the unobserved value, an assumption that may not hold if the outcome would have changed. The registry does not report sensitivity analyses evaluating alternative missing-data assumptions.

10. Confidence Intervals and Precision

Precision concerns how much uncertainty surrounds an estimate. The posted arm means are -1.4 for Exubera and -1.6 for insulin lispro, and the recalculated unadjusted mean difference is -0.2. The confidence interval for that difference includes zero, with one limit below zero and the other above zero. This means the data do not isolate a single direction for the underlying mean difference with high precision under this calculation.

The interval should be interpreted as uncertainty about the group-level contrast, not as a prediction interval for individual outcomes and not as a range containing most participants’ changes. The arm standard deviations, 1.3 and 1.0, describe participant-level dispersion; the confidence interval instead reflects uncertainty in the estimated difference. These quantities answer different questions.

The unadjusted p-value of 0.263 is consistent with the interval’s inclusion of zero. Neither statistic establishes that the groups are equal, and neither establishes that a difference of any particular size is clinically important. A more precise conclusion would require a narrower interval and a prespecified interpretation threshold, neither of which can be posted by the posted summary alone.

The interval is a recalculation from the posted summary statistics, rather than a reported sponsor model. It does not account for possible adjustment for baseline values, other covariates, repeated measurements, or multiplicity. The registry record does not report whether the sponsor used any such methods or whether conclusions differed. The primary endpoint analysis used last observation carried forward in the full analysis set, and the record does not report a sensitivity analysis that would show how missing-data assumptions affect precision.

11. 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 planning figures are illustrative calculations based on the observed effect, not the trial’s original design assumptions. For this continuous-outcome comparison, the figures are participants per arm, not total events. The calculation 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 at the stated fraction of the observed effect.

The 80% power column gives the number of participants per arm that the illustration estimates would be needed for an 80% chance of detecting the specified effect under those assumptions. The 90% power column gives the corresponding number per arm for a 90% chance. These are planning concepts: power is the probability of detecting an effect of a specified size if that effect and the calculation assumptions are correct. The values are not observed power for this completed trial.

Planning for 100% of the observed effect gives 521 participants per arm for 80% power and 697 participants per arm for 90% power. Planning for 75% of the observed effect gives 926 participants per arm for 80% power and 1,240 participants per arm for 90% power. Planning for 50% of the observed effect gives 2,083 participants per arm for 80% power and 2,788 participants per arm for 90% power. Smaller effects need more participants to detect under otherwise unchanged assumptions, which explains the increasing planning figures as the effect fraction decreases.

These estimates use a recalculated observed effect and should not be read as evidence that the trial had a particular probability of success. The registry record does not report the original sample-size justification or its design assumptions.

12. 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 = -0.200; pooled standard deviation σ = 1.152.
2. n = 2 × (1.96 + 0.842)² × 1.152² / -0.200² = 520.6, rounded up to 521 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 <- -0.2                      # observed difference in means
sigma <- 1.1517930451843                     # 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(348374)
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 = 521 (80% power), 697 (90% power)
 75% of observed effect: n per arm = 926 (80% power), 1,240 (90% power)
 50% of observed effect: n per arm = 2,083 (80% power), 2,788 (90% power)
power.t.test: 521.6 
Simulated power at n = 521 per arm: 0.798 (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.

13. Multiplicity, Missing Data and Analysis Populations

The registry lists a primary HbA1c change endpoint and multiple secondary outcome topics, including HbA1c target attainment, glucose measures, lipids, and cardiovascular risk markers. It also describes HbA1c change at each visit among the secondary outcomes. When several outcomes, thresholds, or time points are tested, the chance of at least one apparently notable result can increase. In general, multiplicity procedures can limit that risk, but they may also affect the interpretation and precision of estimates.

Whether any multiplicity adjustment was planned or applied is not reported in the registry record. The unadjusted comparison described for the primary endpoint should therefore not be interpreted as adjusted for the family of listed analyses. The record does not provide enough detail to reconstruct a hierarchy of testing or to identify which outcomes, if any, were intended to support formal confirmatory conclusions.

The primary analysis population was the full analysis set. It included subjects who received at least one dose of study medication and had baseline and post-baseline HbA1c measurements. The primary endpoint description specifies last post-baseline observation carried forward. This approach uses an observed post-baseline value when a later measurement is unavailable; in general, it can misrepresent the unobserved outcome if the measurement would have changed over time. The registry record does not report the number of missing Week 24 values, reasons for missingness, or sensitivity analyses using alternative assumptions.

The record does not report detailed information about treatment adherence, protocol deviations, withdrawals, or the handling of participants who did not meet the analysis-set definition. It also does not report a separate intention-to-treat analysis or a per-protocol analysis. These are reporting gaps, not evidence that such procedures were or were not conducted outside the registry record.

14. 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: meniscus lesion (1.1%), ligament rupture (1.1%), medical device complication (1.1%), muscle spasms (1.1%), ileus (1.0%).

The registry lists serious adverse events among participants at risk in each arm. For Exubera, 3 serious adverse events affected participants among 88 at risk. For insulin lispro, 3 affected participants were reported among 96 at risk. These are counts of affected participants as presented in the registry, not a comparison adjusted for follow-up time or baseline risk. No rate comparison or inferential test is reported.

The registry also lists other adverse events affecting 68 participants among 88 at risk in the Exubera group and 67 among 96 at risk in the insulin lispro group. Those figures are separate from the serious-event counts. The listed serious-event terms include Meniscus lesion, Ligament rupture, Medical device complication, Muscle spasms, Ileus, and Labyrinthitis, with the registry giving each term a percentage of 1.1% or 1.0%. The record does not specify an arm attribution for those listed terms in the registry record.

Deaths are not reported in the registry record. The serious-event counts alone do not establish comparative safety: event type, timing, exposure duration, severity details, and a prespecified statistical comparison are not available in the registry record.

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

The posted endpoint results describe mean HbA1c changes over the study period in two randomized, open-label groups receiving different mealtime insulins on a Lantus-based regimen. The arm means were -1.4 for Exubera and -1.6 for insulin lispro. The unadjusted recalculated mean difference was -0.2 in the direction of a more negative mean change in the lispro group. The interval includes zero, and the unadjusted p-value was 0.263. These statistics do not establish a clear between-arm difference.

They also do not establish equivalence. A failure to demonstrate a clear difference is not proof that the regimens have the same effect. An equivalence or noninferiority conclusion requires an appropriate design, prespecified margin, and analysis; those details are not reported in the registry record. The result also cannot be translated into a claim about the experience of every participant, because the means summarize groups and the standard deviations indicate variation among individual changes.

The registry’s broader outcome list includes target attainment, glucose profiles, meal tolerance tests, lipids, and cardiovascular risk markers. Detailed results for those outcomes are not reported in the registry record, so the primary endpoint result should not be taken as a summary of every outcome studied. The record does not report the original sample-size assumptions, sponsor analysis, multiplicity adjustment, or sensitivity analysis for missing data.

These findings are statistical descriptions of the posted data, not medical advice. They do not determine which treatment is appropriate for an individual, and they do not establish a treatment recommendation. Clinical importance cannot be judged from the mean difference alone because the registry entries do not supply a prespecified threshold for a meaningful difference or the full context needed to evaluate benefits and harms.

16. Why This Trial Matters Statistically

This trial illustrates how to distinguish a point estimate from the certainty around it. The mean difference is -0.2, but the recalculated confidence interval includes zero and the unadjusted p-value is 0.263. Reporting only the direction of the point estimate would omit the uncertainty; treating the p-value as proof of no difference would overstate what the analysis can establish.

It also shows why the analysis population and missing-data method matter. The primary endpoint summaries apply to a full analysis set defined by study exposure and available baseline and post-baseline HbA1c measurements, with last observation carried forward. That method is part of the result’s meaning, not merely a technical detail. Without the number and pattern of missing measurements or sensitivity analyses, readers cannot assess how strongly the estimate depends on that assumption.

The trial’s open-label design and the registry’s incomplete reporting of allocation procedures, multiplicity handling, and the sponsor’s comparative model limit the ability to reconstruct the complete inferential analysis. The standardized and mean-difference summaries can be informative, but they do not replace a reported analysis plan or a clinically defined interpretation threshold. The sample-size illustrations further demonstrate that planning requirements depend on the effect size chosen for detection: smaller effects require larger per-arm samples under the stated assumptions.

For clinical researchers, the central statistical lesson is to read the endpoint definition, population, estimate, uncertainty, and analysis assumptions together. A registry summary can support a careful description of posted results while leaving important questions unresolved.

17. Limitations

• The posted summary statistics are the only data these results rest on; patient-level data, full protocols and statistical analysis plans are not available.
• Recalculated results are simple, unadjusted calculations from posted summaries. They do not replace the sponsor's prespecified analysis.

The registry record does not report the original sample-size justification, the planned allocation ratio, the randomization sequence method, or allocation concealment. It identifies the study as randomized, open-label, parallel-group, and multicenter, but these labels do not provide the operational details needed to assess every potential source of bias.

The primary comparison is an unadjusted recalculation from arm-level summaries rather than a sponsor-reported model. The record does not report whether the sponsor’s analysis adjusted for baseline HbA1c or other factors, or whether repeated measurements were modeled. The registry also does not report the complete analysis plan, interim analyses, a multiplicity procedure, or a prespecified clinical threshold for interpreting the mean difference.

The full analysis set excludes subjects without the required baseline or post-baseline measurements by definition, and the primary endpoint used last observation carried forward. The numbers and reasons for missing measurements, withdrawals, and protocol deviations are not reported in the registry record. No sensitivity analysis for missing-data assumptions is included in the registry entries.

Open-label assignment may affect behavior or care, while the registry does not describe masking of outcome assessors. The trial’s secondary outcome topics are listed, but detailed findings are not reported in the registry record. Safety counts are provided, but deaths are not reported, and the registry summary does not include exposure-adjusted comparisons or a formal safety analysis. These limits constrain interpretation of both the efficacy comparison and the adverse-event summaries.

18. Related Tutorials

19. Related Statistical Calculators

20. Sources

• ClinicalTrials.gov: Exubera vs Lispro in a Lantus-based Regimen for Improved Glycemic Control in Type 2 Diabetes — NCT00348374.
• Registry record last updated 2010-03-16; status last verified 2009-07.
• Results first posted 2010-03-16.
• 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.

21. Record Summary

NCT00348374 was a completed Phase 3b, randomized, open-label, parallel-group, multicenter trial sponsored by Pfizer. It compared Exubera with insulin lispro as mealtime insulin added to insulin glargine and oral agents in suboptimally controlled patients with type 2 diabetes. The listed study start was 2006-06, and primary completion was 2008-08.

The primary endpoint was mean change in HbA1c from baseline to Week 24. Posted means were -1.4 for Exubera and -1.6 for insulin lispro. The unadjusted recalculated mean difference was -0.2; its confidence interval includes zero, and the unadjusted p-value was 0.263. The primary analysis used a full analysis set and last observation carried forward. Original sample-size assumptions, the sponsor’s comparative analysis, and multiplicity handling are not reported in the registry record.