The DURABLE Trial: Evaluating the Durability of Starter Insulin Regimens in Patients With Type 2 Diabetes (IOOV): Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for The DURABLE Trial: Evaluating the Durability of Starter Insulin Regimens in Patients With Type 2 Diabetes (IOOV) (NCT00279201), a randomized phase 4 study. The primary outcome is INITIATION: 24-Week Endpoint Glycosylated Hemoglobin (HbA1c).

2. Clinical Question

3. Trial Design

The DURABLE trial (NCT00279201) was a Phase 4, randomized, open-label, parallel-group study sponsored by Eli Lilly and Company. The registry lists enrollment of 2091 and a study period beginning in 2005-12, with primary completion in 2009-11; status is completed. Participants had type 2 diabetes with inadequate glycemic control while using oral diabetes medicines.

The primary comparison focused on the initiation phase: insulin glargine versus lispro low mix, each added to existing oral therapy, with HbA1c assessed at Week 24. The posted endpoint means are based on the intent-to-treat population with at least one post-baseline assessment in that phase, using last observation carried forward for the endpoint analysis. This outcome population is distinct from the overall enrollment figure.

The treatment pathways described in the registry extend beyond initiation. Participants with HbA1c at or below 7.0 at 24 weeks could enter maintenance on their assigned initiation regimen for up to an additional 2 years. For participants with HbA1c greater than 7.0 after 24 weeks, the record describes randomization to addendum regimens, including lispro mid mix, lispro low mix, or basal-bolus therapy, depending on the preceding regimen and addendum comparison. These later pathways address treatment intensification and maintenance; they do not change the definition of the primary initiation endpoint.

Masking is listed as none. The record identifies random allocation and a parallel model, but the method used to generate the allocation sequence, allocation concealment, and the trial’s original sample-size justification are not reported in the registry record.

4. Participant Flow and Accounting

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

Before assignment: Initiation: Glargine or Lispro LM - 24 weeks. Maintenance: Up to 2 more years treatment if HbA1c \<=7.0 at 24 weeks. After Initiation, if HbA1c \>7.0 after 24 weeks, instead of entering Maintenance, those from Lispro LM could receive Basal bolus or Mid Mix; glargine could receive Basal bolus or Lispro LM for 24 weeks in Intensification Addendum.

Initiation

Reasons for not completing this period:

Maintenance

Reasons for not completing this period:

Intensification Addendum

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: Last observation carried forward (LOCF) method was performed on the intent-to-treat (ITT) population who had at least 1 post-baseline assessment in the initiation phase. Initiation baseline: Week 0.

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 INITIATION: 24-Week Endpoint Glycosylated Hemoglobin (HbA1c), measured over Endpoint (Initiation: Week 24).

The primary endpoint is mean glycosylated hemoglobin (HbA1c), expressed as percent glycosylated hemoglobin, at the initiation-phase endpoint at Week 24. The posted endpoint means are 7.33 for insulin glargine and 7.23 for lispro low mix. The analysis population is the intent-to-treat population with at least one post-baseline assessment during initiation. The record says that last observation carried forward was used to obtain endpoint values; initiation baseline was Week 0.

HbA1c is used in diabetes research as a measure of longer-term blood glucose exposure, rather than a reading at just one moment. Comparing group means at a fixed endpoint provides a summary of the average measured level in each group. It does not describe the full distribution of participant outcomes, the proportion reaching a particular target, or how long an individual maintained a level of control. Those are distinct questions and require distinct outcome definitions.

The endpoint’s mean and standard deviation describe the group result and its variability. The primary endpoint is not the same as the secondary outcome of change from baseline to Week 24: the endpoint analysis compares the Week 24 values, while a change analysis uses both baseline and follow-up measurements. The registry also reports an ANCOVA comparison, but does not provide its adjusted estimate or confidence interval.

7. Results Posted in the Registry

• Lispro Low Mix: mean 7.23 versus 7.33 on control; difference -0.10 (95% CI -0.19 to -0.01), standardized effect d = -0.09; z-test p = 0.039.

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.

INITIATION: Change in HbA1c From Baseline to 24 Weeks

• 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 deviations.
• Result:
  • Lispro Low Mix: mean -1.83 versus -1.68 on control; difference -0.15 (95% CI -0.26 to -0.04), standardized effect d = -0.12; z-test p = 0.009.
• Interpretation:
  • The 95% confidence interval excludes no difference, so the mean is statistically lower for Lispro Low Mix than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.

INITIATION: Change From Baseline to Endpoint in 1,5 Anhydroglucitol (1,5 AG)

• 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 deviations.
• Result:
  • Lispro Low Mix: mean 5.25 versus 4.61 on control; difference 0.64 (95% CI 0.20 to 1.08), standardized effect d = 0.13; z-test p = 0.004.
• Interpretation:
  • The 95% confidence interval excludes no difference, so the mean is statistically higher for Lispro Low Mix than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.

INITIATION: Participant Demographics of Participants Who Did Versus Did Not Achieve HbA1c Goal at Week 24 - Age

• 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 deviations.
• Result:
  • Did Not Meet Goal: mean 55.85 versus 58.21 on control; difference -2.36 (95% CI -3.24 to -1.48), standardized effect d = -0.24; z-test p = <0.001.
• Interpretation:
  • The 95% confidence interval excludes no difference, so the mean is statistically lower for Did Not Meet Goal than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.

The secondary outcomes cover additional aspects of glycemic response and related measures during initiation, including change in HbA1c, attainment of HbA1c thresholds, self-monitored glucose profiles, 1,5 anhydroglucitol, body weight, and self-reported hypoglycemic episodes. The registry also posts analyses about participant characteristics by whether the HbA1c goal was met. Together, these outcomes broaden the description beyond the primary endpoint mean, but they do not all measure the same construct or answer the same question.

The posted continuous secondary comparisons include changes in HbA1c and 1,5 anhydroglucitol; their reported group summaries point in the same broad direction of greater average change in the lispro low-mix group. That directional consistency can be considered alongside the primary endpoint result, but it does not make the measures interchangeable or establish that every secondary outcome supports the same conclusion. Some secondary outcomes concern thresholds, variability, safety-related experiences, or participant characteristics rather than mean glycemic levels.

The registry lists 56 outcome measures and 327 analyses, and 53 secondary outcomes are posted. When many outcomes or analyses are examined, the chance of obtaining apparently notable findings by chance can increase. Unless an analysis plan specifies how multiple comparisons are handled, isolated secondary p-values should receive less weight than the primary comparison. The multiplicity approach is not reported in the registry record.

Secondary findings are therefore best treated as additional descriptive or supportive evidence. Their interpretation depends on outcome definitions, analysis populations, missing-data handling, and whether comparisons were planned. The available registry record does not provide enough information to establish a hierarchy among all secondary analyses or to determine how their results should be adjusted for multiplicity.

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

• INITIATION: Change in HbA1c From Baseline to 24 Weeks
• INITIATION: Percentage of Participants With HbA1c < or = 7.0%, HbA1c <7.0%, and HbA1c < or = 6.5% at Endpoint
• INITIATION: HbA1c
• INITIATION: 7-point Self-monitored Plasma Glucose (SMPG) Profiles and Postprandial Excursions
• INITIATION: Change From Baseline to Endpoint in 1,5 Anhydroglucitol (1,5 AG)
• INITIATION: Incremental Change From Baseline in Body Weight
• INITIATION: Body Weight
• INITIATION: Percentage of Participants With Self-reported Hypoglycemic Episodes

9. Statistical Interpretation

10. Statistical Methods Explained

The primary outcome is continuous: HbA1c is measured on a numeric scale, and the registry reports a group mean and standard deviation for each regimen. A mean gives the arithmetic average for the analyzed participants. A standard deviation describes the spread of individual observations around that mean; it is not the uncertainty in the mean itself. The uncertainty in a difference between means is summarized by a confidence interval.

What does the unadjusted comparison represent?

The recalculated comparison uses the posted group means and standard deviations to estimate a simple difference between group averages. The direction is lispro low mix minus insulin glargine, so a negative estimate corresponds to a lower observed mean in the lispro low-mix group. This is an unadjusted summary comparison: it does not account for baseline values or other covariates. It is useful for describing the raw endpoint contrast, but it is not interchangeable with the sponsor’s ANCOVA analysis.

In general, an analysis of covariance (ANCOVA) compares groups while accounting for specified covariates, often including a baseline measure. Adjustment can improve precision when covariates predict the outcome and can account for chance baseline imbalance. The registry identifies ANCOVA and reports its p-value, but the covariates, adjusted estimate, and confidence interval are not reported in the registry record. Those missing details limit independent interpretation of the adjusted result.

Why distinguish the endpoint from change?

The primary outcome is endpoint HbA1c at Week 24. A separate secondary outcome measures change from baseline to Week 24. These answer related but not identical questions. Endpoint analysis compares follow-up values; change analysis compares within-participant differences from baseline. In general, baseline adjustment or change-score analysis can affect precision and estimates, especially when baseline measurements vary. The primary endpoint definition and the secondary change outcome should therefore remain distinct in interpretation.

What does the p-value tell the reader?

A p-value describes how incompatible the observed data are with a specified null hypothesis under the assumptions of the analysis. It does not measure the size or clinical importance of a difference, and it is not the probability that the null hypothesis is true. The unadjusted p-value of 0.039 belongs to the recalculated summary comparison; the registry-reported ANCOVA p-value is 0.005. These values should be interpreted alongside their respective methods and the estimate’s uncertainty, rather than as stand-alone evidence.

General statistical interpretation also depends on whether an analysis was prespecified, which covariates and populations were used, and whether multiplicity was addressed. The registry record does not report the full statistical analysis plan or enough detail to establish those points. Consequently, the reported p-values should not be treated as if all analysis choices and testing decisions were known.

How does last observation carried forward affect interpretation?

The registry says that last observation carried forward (LOCF) was used for the primary endpoint in the intent-to-treat population with at least one post-baseline assessment in initiation. Under LOCF, a participant’s last observed value is used in place of an unobserved endpoint value. In general, this creates a complete endpoint dataset from available follow-up observations, but it relies on the carried-forward measurement being a reasonable substitute for the missing endpoint. If participants’ measurements would have changed after their last observation, LOCF may misrepresent their endpoint values. The direction and magnitude of any effect cannot be determined from the posted information alone.

The intent-to-treat label indicates the analysis is framed around randomized treatment assignment, but the stated primary analysis population excludes people without any post-baseline assessment. The registry does not report the number excluded for this reason or provide sensitivity analyses under other missing-data methods. This makes the population definition and handling of missing observations important context for the endpoint estimate.

What do the planning illustrations mean?

The planning figures use the observed effect as a starting point and illustrate participant requirements under stated assumptions. They are not evidence about the original trial design or its actual power. In general, a sample-size calculation depends on the assumed effect, variability, significance level, allocation, and chosen power. If the true effect is smaller than the effect used for a calculation, more participants are usually needed to detect it under otherwise unchanged assumptions. The trial’s original design assumptions and sample-size justification are not reported in the registry record.

11. Confidence Intervals and Precision

The recalculated unadjusted mean difference is -0.1, with a 95% confidence interval from -0.195 to -0.005. The interval gives a range of effect estimates compatible with the posted summary statistics under the assumptions used for the unadjusted calculation. In repeated use of the same method, confidence intervals are designed to cover the true parameter at the stated rate; a particular interval is not a probability statement that the true value lies within its endpoints.

Both limits are below zero, while the upper limit is close to zero. Thus the interval indicates that the unadjusted data are consistent with a modest negative difference, with less support for a difference far from zero. The interval is not a measure of clinical importance: judging importance requires context about meaningful HbA1c differences, which is not reported in the registry record.

The interval applies to the unadjusted comparison reconstructed from the posted means and standard deviations. The registry’s ANCOVA result has a reported p-value of 0.005, but no adjusted effect estimate or confidence interval is reported in the registry record. The recalculated interval should not be presented as if it were the interval from ANCOVA.

Precision also depends on the analysis population and missing-data method. The endpoint analysis uses LOCF in the intent-to-treat population with at least one post-baseline assessment. The confidence interval does not account for uncertainty about whether LOCF adequately represents missing endpoint values, nor does it express variation in results that might arise under another analysis plan. It quantifies sampling uncertainty for the specified unadjusted comparison, not every source of uncertainty in the study.

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 planning illustrations express required sample size as participants per arm for this continuous-outcome comparison. They are not total event counts. The calculations use 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. Those are the assumptions for these planning illustrations; the trial’s own original design assumptions and sample-size justification are not reported in the registry record.

The 80% power column gives the participants per arm that the calculation estimates would be needed to have an 80% chance of detecting the specified effect under those assumptions, if that effect were the true one. The 90% power column expresses the corresponding planning target for a 90% chance. These are design probabilities for hypothetical studies, not probabilities that the observed result is correct. The figures are planning illustrations and are not observed power.

Rows planning for a smaller fraction of the observed effect require more participants per arm. This reflects a general property of statistical design: smaller differences are harder to distinguish from random variation when the other assumptions are held constant, so a larger sample is needed to reach the same planned power. The illustrations are sensitive to the estimated effect used as their starting point; an observed estimate can be imprecise and need not equal the effect in a future study.

These figures should not be read as proof that the completed study was adequately or inadequately sized. That judgment would require the original design assumptions and planned analysis, which are not reported in the registry record. They instead show how sample-size requirements change under the stated hypothetical assumptions.

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 = -0.100; pooled standard deviation σ = 1.075.
2. n = 2 × (1.96 + 0.842)² × 1.075² / -0.100² = 1,814.1, rounded up to 1,815 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 <- -0.0999999999999996                      # observed difference in means
sigma <- 1.07500149619368                     # 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(279201)
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 = 1,815 (80% power), 2,429 (90% power)
 75% of observed effect: n per arm = 3,226 (80% power), 4,318 (90% power)
 50% of observed effect: n per arm = 7,257 (80% power), 9,715 (90% power)
power.t.test: 1815 
Simulated power at n = 1,815 per arm: 0.801 (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 comparison of insulin glargine and lispro low mix using ANCOVA, with a p-value of 0.005. It does not report the adjusted mean difference or its confidence interval. The posted group means and standard deviations support a separate, simple unadjusted re-calculation: lispro low mix minus insulin glargine is -0.1, with a recalculated 95% confidence interval from -0.195 to -0.005 and an unadjusted p-value of 0.039.

The two analyses are not the same calculation. ANCOVA can account for specified covariates, whereas the re-calculation uses the displayed endpoint summaries without adjustment. The difference between their p-values is not itself evidence of an error; adjustment and analysis details can affect estimates of uncertainty and statistical evidence. Since the sponsor’s own analysis is the trial-reported analysis, it takes precedence when describing the trial’s inferential result. The unadjusted re-calculation is useful as a transparent check based on the posted summaries, but it should not replace or be attributed to ANCOVA.

Because the adjusted estimate, covariates, and confidence interval are not reported in the registry record, the extent to which adjustment changed the estimated magnitude or precision cannot be assessed. The p-value alone does not provide that information.

15. Multiplicity, Missing Data and Analysis Populations

The registry identifies a primary endpoint and lists numerous secondary outcomes and analyses. It reports 56 outcome measures, 327 analyses, and 53 posted secondary outcomes. With repeated testing across outcomes, time points, and analysis populations, the chance of at least one small p-value arising by chance can increase. Statistical plans commonly address this issue by specifying a primary testing hierarchy or an adjustment procedure, but the multiplicity strategy is not reported in the registry record. The number of analyses alone does not establish which tests were confirmatory or how results were selected for reporting.

The primary endpoint is defined for the initiation phase, with LOCF applied to the intent-to-treat population that had at least one post-baseline assessment. This is the reported missing-data approach. The number of participants lacking a Week 24 value, the amount and pattern of missingness by group, and sensitivity analyses using other assumptions are not reported in the registry record. In general, missingness can matter if the likelihood of missing follow-up or the unobserved outcome is related to treatment or participant status. LOCF may not capture changes that would have occurred after a participant’s last observed value.

The analysis population is not every enrolled participant: the primary outcome description specifies at least one post-baseline assessment in the initiation phase. The record does not report the number omitted from this endpoint analysis for having no post-baseline assessment or give a detailed account of deviations from randomized assignment. Further details on the statistical analysis plan, covariate selection for ANCOVA, treatment of protocol deviations, and handling of missing secondary outcomes are not reported in the registry record.

These omissions do not invalidate the posted findings, but they limit independent assessment of the robustness of the p-values and estimates. In particular, a reader cannot determine from the registry record whether multiplicity control was planned, whether secondary results were prespecified, or whether alternative missing-data analyses led to similar conclusions.

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 0% reporting threshold.

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: cellulitis (1.2%), pleural effusion (1.0%), gastric ulcer (1.0%), coronary artery disease (0.7%), atrial fibrillation (0.6%).

The registry reports serious adverse event counts for the initiation groups: 45 of 1046 participants at risk in the insulin glargine initiation group and 65 of 1045 at risk in the lispro LM initiation group. Other adverse event counts are 785 of 1024 at risk and 826 of 1026 at risk, respectively. These are counts alongside the stated at-risk denominators; they should be read as registry-reported group summaries, not as a complete description of each event’s severity, timing, duration, or relationship to treatment.

The record also lists serious adverse event terms, including cellulitis, pleural effusion, gastric ulcer, coronary artery disease, atrial fibrillation, and congestive cardiac failure, with percentages. These terms are not presented as a full account of all events or as evidence that an event was caused by a study treatment. Serious adverse events are distinct from the broader other-adverse-event counts.

Deaths are not reported in the registry record. The addendum and maintenance groups have separate risk sets and should not be combined with initiation groups as if they were parallel summaries of the same population and exposure period. The posted counts alone do not establish comparative safety or explain why participants experienced events.

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

The primary result describes a difference between group-average HbA1c values at the Week 24 initiation endpoint. The observed mean was 7.23 for lispro low mix and 7.33 for insulin glargine, and the recalculated unadjusted mean difference was -0.1. The registry’s own ANCOVA comparison reports p = 0.005. These statistics describe the analyzed groups under the stated endpoint and analysis definitions.

They do not show that every participant in one group had a better outcome than every participant in the other group. A group mean can conceal substantial variation among participants, as reflected by the reported standard deviations. Nor does a statistically significant comparison by itself show that the difference is clinically important; a threshold for meaningful change is not reported in the registry record.

The primary outcome is a Week 24 endpoint, not a direct measure of durability over the full follow-up period. Although the trial description includes maintenance and intensification pathways, this endpoint result does not establish how long an effect persisted, what occurred among participants entering later phases, or how all treatment pathways compared over time. It also does not describe outcomes beyond the measures posted in the registry.

The result does not identify the causal mechanism for any observed difference, and the p-value is not the probability that one regimen is superior. Interpretation is bounded by the open-label design, the stated LOCF approach, the analysis population, and missing information about the full statistical plan. The safety counts are a separate part of the evidence and should be considered alongside outcome data without treating event counts as proof of causation.

In short, the posted information supports a statistical description of the Week 24 HbA1c comparison and its uncertainty. It is not, on its own, a complete account of long-term outcomes, clinical importance, or the overall balance of benefits and harms.

18. Why This Trial Matters Statistically

This trial illustrates why statistical interpretation depends on keeping the outcome, analysis, and time point clear. The primary endpoint is mean HbA1c at the Week 24 initiation endpoint; the secondary change-from-baseline outcome is related but not identical. Distinguishing them prevents an endpoint comparison from being described as though it were a change score or a measure of long-term durability.

It also shows why an unadjusted reconstruction and a sponsor-reported adjusted analysis should be kept separate. The posted summaries allow a reader to understand the direction and uncertainty of a simple mean contrast, while the registry’s ANCOVA p-value represents the trial’s own reported analysis. Without the adjusted estimate and interval, the two analyses cannot be fully reconciled in terms of magnitude or precision. Naming the analysis behind each result is essential to transparent reporting.

The sample-size illustrations provide another statistical lesson: planning requirements depend on the effect a study is designed to detect. Smaller effects require more participants under otherwise unchanged assumptions. Such illustrations are useful for understanding design trade-offs, but cannot substitute for the trial’s original design rationale, which is not reported in the registry record.

Finally, the registry contains numerous secondary measures and analyses, as well as a LOCF endpoint approach. Multiplicity and missing-data assumptions can shape how strongly results should be interpreted. Recognizing those limits helps distinguish a reported statistical signal from a complete, definitive account of the treatment comparison.

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 registry record does not report the original sample-size justification or the assumptions used to design the trial. The planning illustrations based on the observed effect therefore cannot be used to infer the study’s original target or whether the primary endpoint was planned with a particular power.

The study is listed as unmasked. The registry record does not describe whether outcome assessment involved any protections against knowledge of treatment assignment. HbA1c is a laboratory measure, but other outcomes, including self-reported hypoglycemic episodes, may be more open to influence from awareness of assigned treatment. The record does not provide enough detail to assess that possibility.

The primary endpoint analysis uses LOCF and includes participants with at least one post-baseline assessment. The extent and pattern of missing endpoint information and results from alternative missing-data analyses are not reported in the registry record. The adjusted ANCOVA estimate, its confidence interval, and the covariates used are also not reported, limiting comparison with the unadjusted re-calculation.

The registry lists many outcomes and analyses but does not report a multiplicity strategy or clarify the prespecified status of each secondary analysis. The posted summaries also do not provide enough information to assess all protocol deviations, adherence, or the comparability of participants who entered later maintenance and addendum phases. Those later groups follow different pathways and do not extend the primary Week 24 comparison by themselves.

Finally, a mean endpoint difference does not establish clinical importance or individual-level response. The registry record does not report a threshold for a clinically meaningful HbA1c difference. These limitations constrain interpretation of the posted statistical evidence without changing what the endpoint summaries directly report.

20. Related Tutorials

21. Related Statistical Calculators

22. Sources

• ClinicalTrials.gov: The DURABLE Trial: Evaluating the Durability of Starter Insulin Regimens in Patients With Type 2 Diabetes (IOOV) — NCT00279201.
• Registry record last updated 2011-02-21; status last verified 2011-01.
• Results first posted 2011-02-21.
• 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

NCT00279201, the DURABLE trial, was a randomized, open-label, parallel-group Phase 4 study of insulin glargine versus lispro low mix, each added to existing oral therapy in participants with type 2 diabetes and inadequate glycemic control. Enrollment was 2091; the study is listed as completed. The primary initiation endpoint was mean HbA1c at Week 24, analyzed using LOCF in the intent-to-treat population with at least one post-baseline assessment.

Mean endpoint HbA1c was 7.33 with insulin glargine and 7.23 with lispro low mix. The recalculated unadjusted difference was -0.1, with a 95% confidence interval from -0.195 to -0.005 and p = 0.039. The registry-reported ANCOVA p-value was 0.005; its adjusted estimate and confidence interval are not reported in the registry record. The record also posts secondary outcomes, additional treatment phases, and adverse event counts. Multiplicity handling, detailed missing-data patterns, and the original sample-size rationale are not reported.