This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record.
1. Trial at a Glance
PRONTO-T1D was a randomized, double-blind, parallel phase 3 trial evaluating LY900014 in participants with type 1 diabetes. The registry reports 1392 enrolled participants, six arms, and primary efficacy analyses comparing LY900014 and LY900014 Postmeal with Insulin Lispro (Humalog).
| Feature | PRONTO-T1D |
|---|---|
| Trial name | PRONTO-T1D |
| Brief title | A Study of LY900014 in Participants With Type 1 Diabetes |
| Phase | Phase 3 |
| Status | Completed |
| Condition | Type 1 Diabetes Mellitus |
| Design | Randomized, double-blind, parallel |
| Primary purpose | Treatment |
| Enrollment | 1392 |
| Arms | 6 |
| Lead sponsor | Eli Lilly and Company |
| Sponsor type | Industry |
| ClinicalTrials.gov | NCT03214367 |
2. Clinical Question
The primary statistical question was whether LY900014 and LY900014 Postmeal could meet the prespecified non-inferiority criterion for change from baseline in HbA1c at Week 26 compared with Insulin Lispro (Humalog).
Population
Participants with type 1 diabetes mellitus enrolled in the phase 3 PRONTO-T1D trial.
Interventions
LY900014, LY900014 Postmeal, and other study groups involving Insulin Lispro, Insulin Glargine, and Insulin Degludec.
Comparator
Insulin Lispro (Humalog) served as the comparator in the posted primary analyses.
Primary question
Does LY900014 or LY900014 Postmeal provide a Week 26 HbA1c change that is no worse than the prespecified non-inferiority margin relative to Insulin Lispro?
3. Trial Design
The registry lists six study arms and four intervention products: LY900014, Insulin Lispro, Insulin Glargine, and Insulin Degludec. The posted primary efficacy analyses were restricted to the main global study arms rather than the Maximum Extended Enrollment (MEE) arms/groups.
Study timing
Study start
The registry lists July 17, 2017 as the study start date.
Primary completion
The registry lists September 6, 2018 as the primary completion date.
4. Analysis Populations and Estimands
The primary Week 26 efficacy analyses used all randomized participants with baseline and at least one post-baseline HbA1c measurement. The efficacy estimand for the primary endpoint included participant data when baseline and at least one post-baseline measurement were available prior to permanent discontinuation of study drug.
| Analysis feature | Registry description |
|---|---|
| Primary Week 26 population | All randomized participants with baseline and at least one post-baseline HbA1c data |
| Efficacy estimand | Participant data with baseline and at least one post-baseline measurement available prior to permanent discontinuation of study drug |
| MEE groups | Outcome measures were not reported for MEE arms/groups; analyses were reported for the main global study arms/groups |
| Week 52 HbA1c population | All randomized participants with baseline and at least one postbaseline HbA1c observation |
| Week 52 restriction | Only double-blind arms were analyzed because the open-label group ended at Week 26 |
This distinction is important statistically. Randomization defines the treatment comparison, but the longitudinal outcome analysis still depends on having baseline and post-baseline observations. The registry's stated efficacy estimand also specifies the treatment-discontinuation point relevant to the primary Week 26 analysis.
5. Primary Endpoint
| Endpoint | Definition | Time frame | Primary method |
|---|---|---|---|
| Change From Baseline in Hemoglobin A1c (HbA1c) Efficacy Estimand at Week 26 | HbA1c is the glycosylated fraction of hemoglobin A. HbA1c is measured to identify average plasma glucose concentration over prolonged periods of time. LS mean was determined by mixed model repeated measures (MMRM) with strata and treatment as fixed effects and baseline as a covariate. | Baseline, Week 26 | Mixed model repeated measures |
The prespecified strata in the MMRM were pooled country, type of basal insulin, prandial insulin dosing plan, and HbA1c stratum. Treatment was included as a fixed effect and baseline HbA1c as a covariate.
The registry specifies a 0.4% non-inferiority margin for HbA1c. The treatment comparison therefore concerns whether the observed difference remains within the prespecified boundary for acceptable loss of efficacy.
6. Primary Results: LY900014 vs Insulin Lispro
The first posted primary analysis compared LY900014 with Insulin Lispro (Humalog) for change from baseline in HbA1c at Week 26 using a mixed models analysis.
LS mean difference in HbA1c change
95% CI: -0.16 to 0.00 · P = 0.060
Non-inferiority margin: 0.4% for HbA1c
| Feature | LY900014 vs Insulin Lispro |
|---|---|
| Outcome | Change From Baseline in HbA1c Efficacy Estimand at Week 26 |
| Analysis population | All randomized participants with baseline and at least one post-baseline HbA1c data |
| Method | Mixed Models Analysis |
| Effect measure | LS Mean Difference |
| Estimate | -0.08 |
| 95% CI | -0.16 to 0.00 |
| P-value | 0.060 |
| Hypothesis | Non-inferiority |
| Non-inferiority margin | 0.4% |
The LS mean difference of -0.08 represents the model-adjusted difference in change from baseline in HbA1c between LY900014 and Insulin Lispro, with negative values favoring a greater reduction in HbA1c for LY900014 under this definition of the contrast.
The 95% confidence interval extends from -0.16 to 0.00. The interval describes statistical uncertainty around the estimated LS mean difference; it does not describe the range of individual patient responses.
The P-value of 0.060 is a measure of compatibility with the statistical null hypothesis under the specified analysis framework. It is not a measure of the magnitude or clinical importance of the treatment difference. In particular, the non-inferiority question is governed by the prespecified 0.4% margin rather than by whether a conventional superiority P-value is below 0.05.
The analysis is longitudinal because HbA1c was measured over time, and the MMRM incorporates repeated observations rather than treating a single post-baseline measurement as the entire longitudinal record.
7. Primary Results: LY900014 Postmeal vs Insulin Lispro
The second posted primary analysis compared LY900014 Postmeal with Insulin Lispro (Humalog) for the same Week 26 HbA1c endpoint and used the same mixed models approach.
LS mean difference in HbA1c change
95% CI: 0.04 to 0.22 · P = 0.003
Non-inferiority margin: 0.4% for HbA1c
| Feature | LY900014 Postmeal vs Insulin Lispro |
|---|---|
| Outcome | Change From Baseline in HbA1c Efficacy Estimand at Week 26 |
| Analysis population | All randomized participants with baseline and at least one post-baseline HbA1c data |
| Method | Mixed Models Analysis |
| Effect measure | LS Mean Difference |
| Estimate | 0.13 |
| 95% CI | 0.04 to 0.22 |
| P-value | 0.003 |
| Hypothesis | Non-inferiority |
| Non-inferiority margin | 0.4% |
The LS mean difference of 0.13 indicates that the model-adjusted Week 26 change in HbA1c was 0.13 percentage points higher for LY900014 Postmeal than for Insulin Lispro under the reported contrast.
The 95% confidence interval, 0.04 to 0.22, gives the uncertainty interval around that estimated difference. It does not imply that every participant experienced a difference within those limits.
The P-value of 0.003 indicates strong statistical evidence against a zero-difference null hypothesis in the posted analysis. It does not measure the size of the observed effect. For a non-inferiority analysis, the central question is instead whether the confidence interval remains within the prespecified 0.4% margin.
The positive direction of the estimate should also be kept separate from the non-inferiority conclusion. A non-inferiority design is not primarily testing whether the new treatment produces a statistically significant improvement over the comparator.
8. Reading the Two Primary Comparisons Together
| Comparison | LS Mean Difference | 95% CI | P-value | Hypothesis |
|---|---|---|---|---|
| LY900014 vs Insulin Lispro | -0.08 | -0.16 to 0.00 | 0.060 | Non-inferiority |
| LY900014 Postmeal vs Insulin Lispro | 0.13 | 0.04 to 0.22 | 0.003 | Non-inferiority |
Both confidence intervals are narrower than the 0.4% non-inferiority margin in the sense relevant to the reported comparison. The two estimates also illustrate why the P-value should not be used as a substitute for examining the confidence interval and the prespecified non-inferiority margin.
Relative to the margin
The relevant benchmark for non-inferiority is the prespecified 0.4% HbA1c margin, not a generic threshold such as P < 0.05.
Relative to zero
The confidence intervals also show whether zero lies within the interval, which addresses a different question: whether the data are statistically compatible with no mean difference.
9. Secondary Endpoints
The registry reports five additional statistical analyses covering postprandial glucose excursions at Week 26 and HbA1c at Week 52. The Week 26 glucose endpoints were analyzed using ANCOVA, while Week 52 HbA1c used a mixed models analysis.
| Endpoint | Comparison | Method | Estimate | 95% CI | P-value | Hypothesis |
|---|---|---|---|---|---|---|
| Change From Baseline in 1-hour PPG Excursion During MMTT at Week 26 | LY900014 vs Insulin Lispro | ANCOVA | -27.9 | -35.3 to -20.6 | <0.001 | Superiority |
| Change From Baseline in 1-hour PPG Excursion During MMTT at Week 26 | LY900014 Postmeal vs Insulin Lispro | ANCOVA | 13.2 | 5.0 to 21.4 | 0.002 | Superiority |
| Change From Baseline in 2-hour PPG Excursion During MMTT at Week 26 | LY900014 vs Insulin Lispro | ANCOVA | -31.2 | -41.1 to -21.2 | <0.001 | Superiority |
| Change From Baseline in 2-hour PPG Excursion During MMTT at Week 26 | LY900014 Postmeal vs Insulin Lispro | ANCOVA | -6.7 | -17.6 to 4.3 | 0.235 | Superiority |
| Change From Baseline in HbA1c at Week 52 | LY900014 vs Insulin Lispro | Mixed Models Analysis | -0.06 | -0.16 to 0.03 | 0.184 | Superiority |
1-hour postprandial glucose excursion
LY900014 vs Insulin Lispro
95% CI: -35.3 to -20.6 · P < 0.001
Superiority analysis using ANCOVA
The LS mean difference of -27.9 mg/dL indicates a lower model-adjusted change in the 1-hour postprandial glucose excursion for LY900014 relative to Insulin Lispro under the reported contrast.
The 95% CI from -35.3 to -20.6 mg/dL indicates the statistical uncertainty around the estimated mean difference. The interval does not cross zero, and the reported P-value is <0.001.
This was a superiority analysis rather than the primary non-inferiority question. Consequently, the interpretation centers on evidence for a difference rather than whether the difference stays within a non-inferiority margin.
1-hour postprandial glucose excursion: LY900014 Postmeal
LY900014 Postmeal vs Insulin Lispro
95% CI: 5.0 to 21.4 · P = 0.002
Superiority analysis using ANCOVA
The estimate of 13.2 mg/dL is the reported LS mean difference for change in the 1-hour postprandial glucose excursion. Its positive direction contrasts with the negative estimate observed for LY900014 versus Insulin Lispro.
The 95% CI of 5.0 to 21.4 mg/dL does not include zero, and the reported P-value is 0.002. The statistical evidence therefore supports a difference under the specified superiority analysis.
The result should not be treated as interchangeable with the HbA1c endpoint. PPG excursion and HbA1c measure different aspects of glycemic control and operate on different time scales.
2-hour postprandial glucose excursion
LY900014 vs Insulin Lispro
95% CI: -41.1 to -21.2 · P < 0.001
Superiority analysis using ANCOVA
The reported LS mean difference of -31.2 mg/dL indicates a lower estimated change in 2-hour PPG excursion for LY900014 relative to Insulin Lispro. The 95% CI, -41.1 to -21.2 mg/dL, lies entirely below zero.
The P-value of <0.001 describes evidence against the zero-difference hypothesis under the posted analysis. It does not itself quantify the clinical magnitude of the effect; the estimate and confidence interval provide that information.
2-hour postprandial glucose excursion: LY900014 Postmeal
LY900014 Postmeal vs Insulin Lispro
95% CI: -17.6 to 4.3 · P = 0.235
Superiority analysis using ANCOVA
The estimated difference was -6.7 mg/dL. The 95% CI ranges from -17.6 to 4.3 mg/dL, crossing zero.
The P-value of 0.235 does not provide strong statistical evidence against a zero-difference hypothesis under this superiority analysis. Importantly, this is not evidence that the two treatments are exactly equivalent; failure to demonstrate superiority and demonstration of equivalence are different statistical claims.
Week 52 HbA1c
LY900014 vs Insulin Lispro
95% CI: -0.16 to 0.03 · P = 0.184
Superiority analysis using mixed models
The Week 52 LS mean difference was -0.06, with a 95% CI of -0.16 to 0.03. The confidence interval crosses zero, and the P-value was 0.184.
The registry specifies that only double-blind arms were analyzed for this outcome because the open-label group ended at Week 26. The result therefore belongs to a defined longitudinal analysis population rather than simply extending the Week 26 comparison without qualification.
10. Statistical Methods Explained
Why was a mixed-effects model used for HbA1c?
HbA1c is measured repeatedly over the study period, so the analysis has a longitudinal structure. The registry reports a mixed model repeated measures approach for the primary Week 26 endpoint, with strata and treatment as fixed effects and baseline as a covariate. This allows the analysis to use the repeated-measure framework rather than reducing the study to a single unadjusted comparison.
Why was ANCOVA used for postprandial glucose excursions?
The Week 26 PPG excursion endpoints were analyzed with ANCOVA. ANCOVA compares groups while adjusting for a baseline covariate, producing an adjusted treatment comparison. In this trial, the reported effect measure was an LS mean difference.
What does an LS mean difference of -27.9 mg/dL mean?
An LS mean difference of -27.9 mg/dL means that the model-adjusted mean change in the LY900014 group was estimated to be 27.9 mg/dL lower than the corresponding mean change in the Insulin Lispro group for the specified 1-hour PPG endpoint. The word "adjusted" matters: this is not necessarily the same quantity as a simple arithmetic difference between two observed sample means.
Why is the non-inferiority margin important?
The registry specifies a 0.4% non-inferiority margin for HbA1c. A non-inferiority trial asks whether the new treatment can be shown not to be unacceptably worse than the comparator according to a prespecified clinical boundary. The margin is therefore part of the hypothesis itself, rather than merely a reference value added after the analysis.
Why is the P-value not the same thing as the treatment effect?
A P-value quantifies statistical evidence against a specified null hypothesis under the model and assumptions used. It does not tell the reader how large the treatment difference is. For that, the LS mean difference is the primary effect estimate, while the confidence interval communicates its statistical precision.
What does the confidence interval add?
A 95% confidence interval provides an interval estimate around the treatment effect under the specified statistical framework. For example, the LY900014 versus Insulin Lispro Week 26 HbA1c estimate of -0.08 has a 95% CI of -0.16 to 0.00. The interval is informative about precision and about compatibility with zero, but it does not describe individual patient-level response ranges.
Why does the analysis distinguish MEE arms?
The registry states that outcome measures were not reported for the Maximum Extended Enrollment arms/groups and were instead reported for the main global study arms/groups. This is important because combining populations that were intentionally excluded from a posted analysis could change the estimand and make the resulting comparison inconsistent with the registered analysis.
11. Statistical Methodology
Mixed model repeated measures
The primary HbA1c endpoint was analyzed using a mixed model repeated measures approach. The registry specifies strata consisting of pooled country, type of basal insulin, prandial insulin dosing plan, and HbA1c stratum. Treatment was included as a fixed effect and baseline was included as a covariate.
The exact covariance specification is not described in the registry information summarized here, so the method is best understood at the level of the reported mixed-effects repeated-measures framework rather than as a reconstructed covariance model.
ANCOVA
ANCOVA was used for the Week 26 1-hour and 2-hour PPG excursion endpoints. The method produces an LS mean difference after accounting for the covariate structure specified by the analysis.
The LS mean difference is the model-adjusted treatment contrast rather than necessarily the raw difference between observed arithmetic means.
Non-inferiority analysis
The primary HbA1c analyses were designated as non-inferiority analyses, with a prespecified non-inferiority margin of 0.4%. The statistical interpretation therefore depends on the location of the confidence interval relative to that margin and the direction defined for the treatment contrast.
Margin
The registry specifies a 0.4% HbA1c non-inferiority margin.
Estimate
The LS mean difference quantifies the estimated difference between the randomized treatment groups after model adjustment.
Confidence interval
The 95% CI communicates uncertainty around the estimated difference and is central to non-inferiority interpretation.
P-value
The reported P-value addresses a statistical null hypothesis but does not replace the prespecified non-inferiority margin.
12. Safety Results
The registry reports serious adverse events by arm as affected participants over participants at risk. The reported safety groups include the main Insulin Lispro, LY900014, and LY900014 Postmeal groups as well as the MEE groups.
| Study group | Serious adverse events affected / at risk |
|---|---|
| Insulin Lispro (Humalog) Lead-in | 48/1316 |
| Insulin Lispro (Humalog) | 67/442 |
| LY900014 | 54/451 |
| LY900014 Postmeal | 30/329 |
| Insulin Lispro (Humalog) Lead-in-MEE | 0/76 |
| Insulin Lispro (Humalog)-MEE | 9/31 |
| LY900014-MEE | 3/21 |
| LY900014 Postmeal-MEE | 2/22 |
These are serious-adverse-event counts rather than measures of treatment efficacy. They also illustrate why safety denominators matter: the number affected cannot be interpreted without considering the corresponding number at risk.
13. Understanding the Non-Inferiority Framework
Non-inferiority trials require a different reading strategy from conventional superiority trials. The question is not simply whether the new treatment differs from the comparator. The prespecified question is whether any disadvantage remains within a clinically acceptable boundary.
The registry's primary HbA1c analyses use this margin as the benchmark for the non-inferiority hypothesis.
For LY900014 versus Insulin Lispro, the reported estimate was -0.08 with a 95% CI of -0.16 to 0.00. For LY900014 Postmeal versus Insulin Lispro, the estimate was 0.13 with a 95% CI of 0.04 to 0.22. Both intervals are substantially narrower than the stated 0.4% margin.
The LY900014 primary comparison has a P-value of 0.060, while the LY900014 Postmeal comparison has a P-value of 0.003. Those values answer a question about evidence against a specified null hypothesis. They do not directly answer the non-inferiority question unless the null hypothesis and direction are explicitly defined around the non-inferiority margin.
The confidence interval and the prespecified margin therefore provide the key statistical framework for interpreting the primary objective.
14. Week 26 vs Week 52: Why the Time Frame Matters
| Time frame | Endpoint | Comparison | Method | Estimate | P-value |
|---|---|---|---|---|---|
| Baseline, Week 26 | HbA1c efficacy estimand | LY900014 vs Insulin Lispro | Mixed-effects model | -0.08 | 0.060 |
| Baseline, Week 26 | HbA1c efficacy estimand | LY900014 Postmeal vs Insulin Lispro | Mixed-effects model | 0.13 | 0.003 |
| Baseline, Week 52 | HbA1c | LY900014 vs Insulin Lispro | Mixed-effects model | -0.06 | 0.184 |
The Week 52 endpoint should not be treated as simply a delayed version of the Week 26 primary endpoint. The registry states that only double-blind arms were analyzed at Week 52 because the open-label group ended at Week 26. The analysis population and treatment exposure therefore differ in an important design-specific way.
15. Interpreting Effect Estimates and Confidence Intervals
An LS mean difference of -0.08 for HbA1c is an absolute difference in the HbA1c outcome scale. It should not be translated into a percentage reduction in risk or a relative treatment effect.
The 95% CI of -0.16 to 0.00 for LY900014 versus Insulin Lispro describes uncertainty around the model-based estimate. It is not a prediction interval for future participants and does not mean that individual participant outcomes must lie inside the interval.
The P-value of 0.060 does not mean the treatment effect has a 6.0% probability of being due to chance. It is a probability calculation under the specified null hypothesis and statistical model. The magnitude of the observed effect is described by the LS mean difference, while the confidence interval describes its uncertainty.
The primary HbA1c comparisons were non-inferiority analyses, whereas the posted PPG and Week 52 HbA1c analyses were labeled superiority analyses. The same numerical P-value threshold should not be used as a substitute for understanding the hypothesis being tested.
16. Multiplicity and Multiple Posted Analyses
The registry contains two posted primary analyses for the single registered primary endpoint and five additional posted secondary analyses. The primary endpoint therefore appears in two treatment comparisons, while the secondary results span several related glucose outcomes and a later HbA1c assessment.
| Analysis group | Number of posted analyses | Statistical role |
|---|---|---|
| Primary endpoint | 2 | Non-inferiority comparisons of HbA1c at Week 26 |
| Secondary endpoints | 5 | Superiority analyses of PPG excursions and Week 52 HbA1c |
| Total statistical analyses posted | 7 | Registry-posted statistical analyses |
When multiple hypotheses are examined, the nominal P-values should be interpreted in the context of the prespecified testing strategy. The registry information summarized here identifies the hypothesis type for each posted analysis but does not provide a complete multiplicity hierarchy for all seven analyses. Consequently, each result should be read according to its registered role rather than treating every P-value as an independent confirmatory test.
17. Crossover, Factorial Design, and Interim Analysis
The registry describes a parallel randomized design. The information available for this record does not identify a crossover design or factorial design, and no interim-analysis methodology is reported in the statistical analysis information summarized here.
Parallel design
Participants are compared according to their randomized parallel treatment groups rather than being used as their own crossover controls.
Non-inferiority
The primary HbA1c question uses a prespecified 0.4% margin rather than a superiority-only framework.
Mixed-effects model
Repeated HbA1c measurements are handled using the reported mixed model repeated measures approach.
ANCOVA
Week 26 PPG excursion outcomes were analyzed using ANCOVA with LS mean differences reported.
18. Missing Data and Treatment Discontinuation
The primary efficacy estimand is explicitly defined using participant data when baseline and at least one post-baseline measurement were available prior to permanent discontinuation of study drug. This means the registry's estimand definition incorporates treatment discontinuation into the specification of which observations contribute to the primary efficacy analysis.
The posted information does not identify a separate imputation method such as multiple imputation, last observation carried forward, or a specific pattern-mixture model. The appropriate interpretation is therefore the reported MMRM-based analysis and its stated efficacy-estimand population, rather than an assumed imputation procedure.
19. Stratification and Covariate Adjustment
The primary MMRM included four stratification factors and baseline HbA1c as a covariate:
| Model component | Role in primary MMRM |
|---|---|
| Pooled country | Stratification factor |
| Type of basal insulin | Stratification factor |
| Prandial insulin dosing plan | Stratification factor |
| HbA1c stratum | Stratification factor |
| Treatment | Fixed effect |
| Baseline | Covariate |
These adjustments serve a different purpose from randomization itself. Randomization establishes the treatment comparison, while covariate and stratification adjustment can improve precision and account for prespecified design factors in the statistical model.
20. Safety and Efficacy Use Different Statistical Populations
The primary efficacy analyses were defined around randomized participants with baseline and post-baseline efficacy observations. The registry separately reports serious adverse events for several study groups, including MEE groups.
This distinction is statistically meaningful because the estimand for an efficacy endpoint and the risk set for a safety endpoint need not be identical. Efficacy asks about the treatment comparison under the prespecified endpoint definition, whereas safety summaries are closely tied to exposure and the population at risk for adverse events.
21. Important Limitations and Interpretation Issues
- Non-inferiority interpretation: the primary HbA1c analyses must be interpreted against the prespecified 0.4% non-inferiority margin rather than by a generic superiority P-value rule.
- Analysis population: the primary efficacy population requires baseline and at least one post-baseline HbA1c observation, and the registry specifies a particular efficacy estimand tied to permanent study-drug discontinuation.
- MEE groups: the posted primary efficacy analyses exclude MEE arms/groups, while the safety information includes MEE groups. These populations should not be treated as interchangeable.
- Multiple analyses: the registry contains seven posted statistical analyses covering primary and secondary endpoints. Their hypothesis types and roles differ.
- Week 52 interpretation: the Week 52 HbA1c analysis includes only double-blind arms because the open-label group ended at Week 26.
- Confidence intervals: confidence intervals describe statistical uncertainty around the model-based estimates; they do not describe individual patient responses.
- P-values: P-values describe evidence under specified null hypotheses and do not quantify treatment-effect magnitude.
- Model assumptions: mixed-effects and ANCOVA analyses depend on their statistical modeling assumptions. The registry summary does not provide the complete covariance or residual specification needed to independently assess every assumption.
22. Why This Trial Matters Statistically
PRONTO-T1D is a useful teaching example because its posted analyses illustrate several important distinctions in clinical-trial statistics within a single randomized study: non-inferiority versus superiority, repeated-measures modeling versus ANCOVA, model-adjusted LS means versus raw means, and efficacy populations versus safety populations.
| Concept | How it appears in PRONTO-T1D |
|---|---|
| Randomization | Randomized parallel phase 3 design |
| Blinding | Double-blind study design |
| Non-inferiority | Primary Week 26 HbA1c comparisons with a 0.4% margin |
| Mixed-effects model | Primary Week 26 HbA1c and Week 52 HbA1c analyses |
| ANCOVA | Week 26 1-hour and 2-hour PPG excursion analyses |
| LS mean difference | Reported effect measure for the posted analyses |
| Confidence interval | 95% intervals accompany the posted estimates |
| P-value | Reported for each posted statistical analysis |
| Stratification | Pooled country, basal insulin, prandial insulin dosing plan, and HbA1c stratum in the primary MMRM |
| Analysis population | Prespecified randomized populations with required baseline and post-baseline observations |
23. Related Statistical Methods
Learn more about the methods used in this trial:
24. Related Statistical Calculators
Use these statistical tools to explore the core concepts represented in the trial:
25. Sources
- ClinicalTrials.gov: NCT03214367 — PRONTO-T1D.
- PubMed: PMID 38035641.
- PubMed: PMID 34347267.
- PubMed: PMID 33458803.
- PubMed: PMID 32728832.
Continue through the Clinical Biostats statistical library
Explore the methods represented in randomized clinical-trial analyses, from ANCOVA and mixed-effects models to confidence intervals and non-inferiority design.
26. Limitations
The registry provides a substantial description of the primary and secondary statistical analyses, but not every modeling detail needed to independently reproduce the full analysis dataset. In particular, the posted information identifies the mixed-effects and ANCOVA methods, covariates and strata for the primary MMRM, analysis populations, effect measures, confidence intervals, and P-values, but does not provide a complete statistical-programming specification.
The serious-adverse-event information also spans several groups, including MEE groups, while the primary efficacy analyses explicitly restrict reporting to the main global study arms/groups. Safety and efficacy results therefore require separate attention to their denominators and analysis populations.
Finally, the posted statistical analyses include both non-inferiority and superiority hypotheses. Their estimates should be interpreted according to the hypothesis attached to each endpoint rather than using a single generic significance rule across the entire trial.
27. Record Summary
PRONTO-T1D provides a compact example of several core principles in modern clinical-trial biostatistics. The trial used randomized, double-blind, parallel design with 1392 enrolled participants and six arms. Its primary Week 26 endpoint was change from baseline in HbA1c, analyzed using a mixed model repeated measures framework with prespecified stratification factors, treatment as a fixed effect, and baseline as a covariate.
The two posted primary comparisons were LY900014 versus Insulin Lispro, with an LS mean difference of -0.08 and 95% CI -0.16 to 0.00, and LY900014 Postmeal versus Insulin Lispro, with an LS mean difference of 0.13 and 95% CI 0.04 to 0.22. Both were evaluated under a prespecified HbA1c non-inferiority margin of 0.4%.
The secondary analyses demonstrate a different statistical framework: ANCOVA was used for 1-hour and 2-hour PPG excursions at Week 26, while mixed models were used for HbA1c at Week 52. The posted estimates range from statistically strong evidence of differences in some PPG comparisons to a confidence interval crossing zero for the LY900014 Postmeal 2-hour PPG comparison and the Week 52 HbA1c comparison.