A Follow-On Study of Donanemab (LY3002813) With Video Assessments in Participants With Alzheimer's Disease (TRAILBLAZER-EXT): Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for A Follow-On Study of Donanemab (LY3002813) With Video Assessments in Participants With Alzheimer's Disease (TRAILBLAZER-EXT) (NCT04640077), a non-randomized phase 2 study. The primary outcome is Part A: Intraclass Correlation Between On-Site and Video Teleconference (VTC) Assessments.

2. Clinical Question

The study addressed two related questions in symptomatic Alzheimer’s disease. In Part A, it asked whether selected clinical assessments administered by video teleconference (VTC) agreed with assessments conducted on site. In Part B, it followed participants who had received placebo in originating trials as they received donanemab. The registry’s primary endpoint is from Part A, not a comparison of donanemab with a concurrent control.

The record describes a two-group assessment sequence in Part A: one group completed on-site assessments before VTC assessments, while the other completed VTC assessments before on-site assessments. This ordering makes the central issue agreement between modes, rather than whether one mode produces a larger average score. The registry record does not report numerical ICC estimates or their confidence intervals, so the degree of agreement cannot be evaluated.

Part B had a different purpose and structure: it documented safety, tolerability, and efficacy outcomes during donanemab exposure. Because the listed allocation is non-randomized and the record identifies no concurrent like-for-like control, Part B observations should not be interpreted as a randomized treatment comparison.

3. Trial Design

TRAILBLAZER-EXT is listed as a Phase 2 study and is marked completed. The record gives an enrollment of 95, a start date of 2020-11-23, and a primary completion date of 2024-02-27. It describes a sequential study with two parts, rather than one single comparison serving all objectives.

Part A validated remote clinical scale assessments. Participants completed assessments both on site and at home by VTC, with the sequence reversed between the two groups. The primary outcome description says that Group 1 completed on-site assessments first and VTC assessments afterward; Group 2 completed VTC assessments first and on-site assessments afterward. Participants in this part did not receive study drug. Although the overall allocation field is non-randomized, the Part A description says participants were randomized 1:1 to the assessment-sequence groups. The registry record therefore contains different descriptions of allocation at the overall-study and Part A levels; it does not explain this distinction further.

Part B enrolled participants who had received placebo in originating trials. They received donanemab intravenously, beginning with 700 milligrams every 4 weeks for 3 doses and then 1400 milligrams every 4 weeks for up to 48 weeks. This sequence is not described as a randomized comparison against a concurrent control group. The record’s own caution is that the arms are not a like-for-like control comparison.

The primary endpoint belongs to Part A and concerns agreement across four scales. Part B lists several secondary outcomes, including cognitive and functional scales, imaging measures, and pharmacokinetics. The registry record does not report a sample-size justification, detailed statistical analysis plan, or rationale for the sequence and duration of Part B.

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: Participants in Part A were randomized 1:1 into two groups (Group 1 and Group 2) to have their cognitive and functional scales assessed.

Period 1

Reasons for not completing this period:

Period 2

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: All participants in Part A who had evaluable data for this outcome.

5. Baseline Characteristics

The table reproduces the baseline characteristics posted in the registry, which describe the participants analysed at baseline and help show whether the randomized groups were comparable.

Population: All enrolled participants in Part A.

6. Primary Endpoint

The registry lists the primary outcome as Part A: Intraclass Correlation Between On-Site and Video Teleconference (VTC) Assessments, measured over Baseline to 4 Weeks.

The primary endpoint is the intraclass correlation coefficient (ICC) between on-site and video teleconference assessments, with a 95% confidence interval. It is specified over baseline to 4 weeks and applies to participants in Part A with evaluable data for the outcome. The scales named are the ADAS-Cog13, ADCS-ADL, MMSE, and CDR-SB.

In general, an ICC is used to summarize how strongly repeated or paired measurements resemble one another. Here, the paired measurements are scores from different assessment modes. An ICC can help describe whether participants who score relatively high or low in one mode tend to have correspondingly high or low scores in the other, while also reflecting the variation among participants. It is therefore a natural type of measure for studying agreement or reliability between measurement approaches.

The confidence interval communicates uncertainty around an ICC estimate. A narrow interval indicates greater precision than a wide interval, although interpretation also depends on the ICC estimate and the context in which the scale is used. The registry record does not report the ICC values, confidence intervals, or a threshold for what level of agreement would be considered acceptable. It also does not identify the ICC formulation or calculation details. The endpoint definition therefore clarifies what was intended to be measured, but does not by itself establish how close the assessment modes were in the observed data.

Agreement is not the same question as whether the average scores from the two modes are identical, nor does an ICC alone establish that a remote assessment can replace an on-site assessment for every use. In general, interpretation may also depend on systematic differences between modes and on the range of scores in the study population.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The primary analysis uses an intraclass correlation coefficient to summarize agreement between two ways of administering the same clinical scales. In this study, those modes were assessment on site and assessment at home by VTC. The endpoint description names four outcomes: ADAS-Cog13, ADCS-ADL, MMSE, and CDR-SB. In general, an ICC is useful when measurements are paired or repeated and the goal is to describe how closely they correspond, rather than merely whether two variables move together.

What an ICC summarizes

In general, an ICC relates variation between participants to variation in their repeated or paired measurements. Its interpretation depends on the exact ICC formulation and design. Some formulations emphasize consistency in participant ranking, while others are intended to reflect absolute agreement and can be affected by systematic differences between measurement occasions or methods. The registry record says the ICC was used to determine agreement, but it does not report the specific formulation, model, or calculation method. That limits how precisely an external reader can reproduce or interpret the analysis.

An ICC is also influenced by the spread of scores in the sample. In general, a sample with wide variation between participants can yield a different ICC from a sample with a restricted score range, even when the absolute discrepancies between paired measurements are similar. For that reason, the coefficient should be considered alongside the participant population, the scale, and any direct description of differences between modes. Such additional details are not reported in the registry record.

Why the confidence interval matters

The endpoint specifies a 95% confidence interval. In general, a confidence interval gives a range of values compatible with the data under the analysis assumptions and conveys the precision of an estimate. It is not a guarantee that the true value lies within the particular interval. The width of the interval depends on information in the data and the statistical method. Since neither the ICC estimates nor their intervals are reported, their precision cannot be assessed.

The record also does not state an acceptance threshold for agreement. A confidence interval alone would not establish whether a result met a practical standard unless that standard were defined and interpreted alongside the estimate. No equivalence or non-inferiority margin is reported. The stated endpoint should therefore be understood as a measure of agreement, not as a documented test against a specified pass criterion.

Assessment sequence and pairing

Part A used two assessment sequences: on-site followed by VTC in Group 1, and VTC followed by on-site in Group 2. In general, using more than one sequence can help distinguish mode differences from effects associated with completing assessments earlier or later. It does not automatically remove such effects; the analysis must account for the design, and the registry record does not describe how sequence was included in the analysis. The outcome population is described as all Part A participants with evaluable data for the relevant outcome, but no further definition of evaluability is given.

The registry record describes the Part A groups as randomized 1:1, while the overall allocation field labels the study non-randomized. These descriptions may refer to different levels of the study, but the record does not explain the relationship. The study’s overall sequential structure and the non-randomized Part B treatment period should not be conflated with the Part A assessment-sequence assignment.

Over-dispersion and count outcomes

In general, over-dispersion occurs when count data vary more than a statistical model expects. If a count model assumes less variability than is present, its uncertainty estimates can be too small and statistical evidence can appear more precise than warranted. This is a general consideration for count outcomes; the primary endpoint is an ICC, not a count. The registry record does not report a count-model method or an over-dispersion adjustment for the primary analysis.

Part B lists outcomes that include scale changes, imaging changes, and trough concentrations. The registry record does not specify the statistical models used for these secondary outcomes, how repeated measurements were handled, or how multiplicity was addressed. The endpoint list alone is not enough to infer those methods.

Planning illustrations are not trial results

The planning illustrations are based on an observed effect and stated assumptions, including a two-sided significance level of 0.05, equal allocation, and a normal approximation. The illustration considers detecting a stated fraction of the observed effect. These calculations are planning illustrations, not observed power and not the study’s original sample-size justification. For time-to-event illustrations, the total number of events is the planning quantity; other illustrations express participants per arm. The registry record does not report the trial’s own sample-size justification, and the primary numerical effect needed to assess those illustrations is not reported.

10. Multiplicity, Missing Data and Analysis Populations

The registry record lists a primary Part A agreement endpoint and several Part B secondary outcomes. The secondary outcomes include changes in MMSE, ADAS-Cog13, iADRS, ADCS-iADL, and CDR-SB, as well as amyloid plaque deposition, whole brain volume, and donanemab trough concentrations. Multiple outcomes can create a multiplicity issue: when many hypotheses are examined, the chance of at least one apparently notable result can increase. In general, a prespecified testing hierarchy or adjustment can address this, but the registry record does not describe one.

The record does not report whether the secondary outcomes were adjusted for multiplicity, whether analyses were exploratory, or whether any hierarchical testing procedure was used. It also does not provide a statistical analysis plan, interim-analysis details, or p-values and confidence intervals for these outcomes. These items are not reported in the registry record, so they should not be inferred from the outcome list or the study’s completed status.

Missing-data handling is also not reported in the registry record. The primary population is described as Part A participants with evaluable data for the outcome, but the record does not define evaluability, give the number with available data for each measure, describe reasons for missing assessments, or state whether imputation or other methods were used. Without those details, it is not possible to assess how missing observations may have affected the ICC estimates.

Analysis populations beyond the stated evaluable-data population are not described. The record does not say how protocol deviations, incomplete assessment pairs, or participants who did not complete follow-up were handled in the primary analysis. These omissions matter because agreement estimates require paired information and may be affected by which participants contribute data.

11. Safety Results

The registry reports serious adverse events as affected participants divided by participants at risk.

Deaths

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: amyloid related imaging abnormality-oedema/effusion (5.5%), amyloid related imaging abnormality-microhaemorrhages and haemosiderin deposits (3.6%), syncope (2.2%), acute myocardial infarction (1.8%), atrial fibrillation (1.8%).

The registry reports serious adverse events separately for Part A Group 1, Part A Group 2, and Part B. It lists 0 of 50 participants affected in Part A Group 1, 1 of 45 in Part A Group 2, and 16 of 55 in Part B. The record lists 2 deaths for Part B and none for either Part A group. These are descriptive counts for the listed study periods; they do not, by themselves, establish that an event was caused by the study drug.

For other adverse events, the reported counts are 0 of 50 in Part A Group 1, 0 of 45 in Part A Group 2, and 40 of 55 in Part B. Part A involved no study drug, while Part B involved donanemab. The groups also differ in study period and treatment exposure, so their counts should not be treated as a randomized comparison of safety.

The listed serious-event terms include amyloid related imaging abnormality-oedema/effusion at 5.5%, amyloid related imaging abnormality-microhaemorrhages and haemosiderin deposits at 3.6%, syncope at 2.2%, acute myocardial infarction at 1.8%, atrial fibrillation at 1.8%, and cardiac failure congestive at 1.8%. These percentages describe the reported terms; they are not estimates of comparative risk. The registry record does not establish from these figures alone whether events were treatment-related or how the terms were adjudicated.

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

The record establishes the structure and intended questions of TRAILBLAZER-EXT: a Part A assessment-mode validation and a sequential Part B donanemab period. It specifies an ICC with a 95% confidence interval as the primary endpoint and names the four assessment scales. However, numerical ICC results and intervals are not reported in the registry record. The main question about the observed agreement between VTC and on-site assessment therefore cannot be answered from the information.

The study’s status is completed, but completion does not itself indicate a positive, negative, or precise result. The record’s stated study goals include safety and efficacy in Part B, yet the provided outcome information does not include numerical efficacy results. A list of planned or named outcomes is not a report of what those outcomes showed.

Part B should not be interpreted as a controlled estimate of treatment effect. Participants who had received placebo in originating trials received donanemab under the described dosing sequence, and the record identifies the overall allocation as non-randomized. It explicitly cautions that the arms are not a like-for-like control comparison. Observations during this period can describe what was recorded in the treated group, but they cannot by themselves separate treatment effects from other influences in a randomized comparison.

The serious adverse event and other adverse event counts are useful descriptions of the recorded safety experience in the listed groups. They do not establish causality, and the group differences should not be read as comparative safety estimates because the study periods and treatment exposure differ. The registry record does not report enough detail to determine attribution or provide a controlled comparison.

In general, agreement statistics need to be interpreted alongside their estimates, confidence intervals, scale properties, and a meaningful criterion for acceptable agreement. None of those numerical primary results or an acceptance threshold appears in the registry record. The careful conclusion is therefore limited: the study was designed to assess agreement across assessment modes and to observe outcomes during a follow-on donanemab period, but the posted record as does not establish the magnitude or precision of the primary result.

13. Why This Trial Matters Statistically

This study is statistically notable because its primary question concerns measurement rather than a conventional treatment-versus-control effect. In clinical research, remote assessment can change how outcomes are collected. Before results from remotely administered scales can be interpreted alongside on-site measurements, researchers need evidence about how closely the two modes correspond. The ICC is a familiar way to summarize that correspondence for continuous or scale-based outcomes.

The assessment sequence is also relevant. Part A used both on-site and VTC assessment, with the order reversed between groups. This paired approach focuses on within-participant agreement, while the two sequences recognize that order may matter. The record identifies the sequence but does not report the analysis details or results, illustrating why design descriptions and numerical estimates are both needed to evaluate validation work.

The study also makes clear why a follow-on treatment period must be distinguished from the primary measurement question. Part B involved donanemab exposure among participants who had received placebo in originating trials, but it is not described as a like-for-like randomized control comparison. Safety and outcome observations in that setting answer a different descriptive question from a controlled estimate of treatment effect.

Finally, this record illustrates the limits of registry summaries. It identifies the primary endpoint, outcome families, study periods, and some adverse-event counts, but does not report the numerical primary endpoint results or the statistical details needed to fully assess precision, missingness, and multiplicity. A complete interpretation depends not only on what a trial set out to measure, but also on the estimates and methods reported for that measurement.

14. 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 most important limitation for interpreting the primary question is that the registry record does not report the ICC estimates or their 95% confidence intervals. Without these values, the observed agreement and its precision cannot be assessed. The record also does not state an acceptable-agreement threshold, so even a numerical ICC would need a defined interpretive context.

The statistical details for the ICC are not described. The registry record does not identify the ICC formulation, model, or handling of assessment sequence. It also does not report results separately by assessment sequence or indicate whether possible period or order effects were considered. The overall allocation field says non-randomized, while the Part A narrative says participants were randomized 1:1 to groups; the relationship between these descriptions is not explained.

The primary population is limited to Part A participants with evaluable data, but the record does not define evaluability or report how many participants contributed paired data to each scale. Missing-data methods, reasons for missing assessments, and sensitivity analyses are not reported in the registry record. This prevents assessment of whether the available pairs represent all participants who entered Part A.

Part B has additional interpretive constraints. It is a sequential donanemab period among participants who received placebo in originating trials, and the record warns that the arms are not a like-for-like control comparison. The numerical secondary outcome results and detailed analysis methods are not reported. The record also does not report a sample-size justification, multiplicity strategy, or interim-analysis details. These limitations mean the listed objectives and outcomes should not be mistaken for evidence about the magnitude or certainty of effects.

15. Related Tutorials

16. Related Statistical Calculators

17. Sources

• ClinicalTrials.gov: A Follow-On Study of Donanemab (LY3002813) With Video Assessments in Participants With Alzheimer's Disease (TRAILBLAZER-EXT) — NCT04640077.
• Registry record last updated 2025-06-13; status last verified 2025-05.
• Results first posted 2025-06-13.
• Data retrieved from ClinicalTrials.gov on 2026-11-09.
• Group sizes, means, event counts and sponsor-reported p-values are reproduced as posted. Effect estimates, confidence intervals and p-values not attributed to the sponsor are independently calculated from the posted summary statistics.

18. Record Summary

TRAILBLAZER-EXT (NCT04640077) is a completed Phase 2 follow-on study sponsored by Eli Lilly and Company, with enrollment of 95. It had two parts: Part A assessed agreement between on-site and VTC clinical scales without study drug, and Part B involved donanemab in participants who had received placebo in originating trials.

The primary endpoint was the ICC between assessment modes for ADAS-Cog13, ADCS-ADL, MMSE, and CDR-SB, with a 95% confidence interval over baseline to 4 weeks. Numerical primary results are not reported in the registry record. Part B safety figures are reported by study group, but the non-randomized, sequential design does not provide a like-for-like control comparison. Key details including the primary estimates, missing-data methods, multiplicity approach, and sample-size justification are not reported in the registry record.