This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record. Reported estimates are separated from statistical interpretation, which appears in clearly labelled boxes.
1. Trial at a Glance
ADvocate2 was a completed phase 3 randomized trial asking whether lebrikizumab, a biologic given as 250 mg every 2 weeks (Q2W), increased the proportion of patients with moderate to severe atopic dermatitis who reached clear or almost-clear skin and a 75% improvement in eczema severity after 16 weeks, compared with placebo.
| Feature | ADvocate2 |
|---|---|
| Phase | Phase 3 |
| Condition | Moderate to severe atopic dermatitis |
| Design | Randomized, parallel-group, quadruple-masked, placebo-controlled; treatment purpose |
| Enrollment | 445 participants |
| Interventions | Lebrikizumab (LY3650150; biological); placebo |
| Primary endpoints | IGA 0 or 1 with a ≥2-point reduction from baseline at Week 16; EASI-75 at Week 16 |
| Primary analysis | Cochran-Mantel-Haenszel risk difference, two-sided 95% CI, superiority hypothesis |
| Missing data | Markov Chain Monte Carlo multiple imputation (MCMC-MI) for the primary endpoints |
| Status | Completed; results posted |
| Dates | Start October 29, 2019; primary completion July 12, 2021 |
| Sponsor | Eli Lilly and Company (industry) |
| ClinicalTrials.gov | NCT04178967 |
2. Clinical Question
The central question was whether lebrikizumab 250 mg Q2W, compared with placebo, increases the proportion of participants who respond on two physician-rated measures of atopic dermatitis severity after a 16-week induction period. Because both primary endpoints are binary (responder versus non-responder), the natural effect measure is a difference in response proportions rather than a difference in means.
Population
Participants with moderate to severe atopic dermatitis, including adolescents (a PROMIS Anxiety analysis was restricted to adolescent participants).
Intervention
Lebrikizumab 250 mg every 2 weeks during the 16-week induction period, followed by blinded re-randomization of responders to Q2W, every 4 weeks (Q4W), or placebo.
Comparator
Placebo during induction, under quadruple masking.
Primary question
Does lebrikizumab Q2W increase the Week 16 proportion achieving IGA 0/1 with ≥2-point improvement, and the proportion achieving EASI-75, relative to placebo?
3. Trial Design
ADvocate2 combined a placebo-controlled induction comparison with a randomized-withdrawal style maintenance period. The two parts answer different questions and use different analysis populations.
Lebrikizumab 250 mg Q2W
- Lebrikizumab 250 mg every 2 weeks to Week 16
- Week 16 responders re-randomized (blinded) to lebrikizumab Q2W, lebrikizumab Q4W, or placebo
- Participants not continuing in blinded maintenance could enter an open-label escape arm
Placebo
- Matching placebo to Week 16
- Placebo responders continued in a blinded maintenance group
- Escape to open-label lebrikizumab was available at Week 16 and from Weeks 24 to 48
4. Analysis Populations and Handling of Missing Data
The registry describes the primary efficacy population as all randomized participants, analyzed regardless of whether they took the assigned treatment, received the correct treatment, or otherwise followed the protocol. This is an intention-to-treat style definition, with one reported exception: one investigational site with eighteen participants was excluded from analysis due to Good Clinical Practice (GCP) issues.
| Population | Definition in the registry | Used for |
|---|---|---|
| All randomized (site excluded) | All randomized participants, excluding the site with 18 participants | Primary endpoints; Week 16 IGA/EASI responder endpoints; EASI and SCORAD percent change |
| Symptom-restricted subsets | All randomized with baseline Pruritus NRS >0, ≥4 or ≥5; sleep-loss score >0 or ≥2; DLQI ≥4 or non-missing | Itch, sleep and quality-of-life responder and change endpoints |
| Observed-data populations | All randomized with observed BSA or POEM data | Mixed-model (MMRM) analyses |
| Adolescent subset | All randomized adolescent participants | PROMIS Anxiety |
| Maintenance population | Randomized to lebrikizumab Q2W at baseline, re-randomized at Week 16 to Q2W, Q4W or placebo, dosed at least once in maintenance, with evaluable Week 52 data | Week 52 maintenance-of-response endpoints |
Three missing-data strategies appear across the posted analyses: MCMC multiple imputation for the primary endpoints and most Week 16 responder and change endpoints; last observation carried forward (LOCF) for SCORAD, EQ-5D-5L and PROMIS Anxiety; and mixed models for repeated measures (MMRM), which use all observed visits without explicit imputation, for BSA and POEM. Each strategy makes a different assumption about why data are missing, and results should be read with that assumption in mind.
5. Endpoints
Primary endpoints
| Endpoint | Registry definition | Time frame |
|---|---|---|
| IGA 0 or 1 with ≥2-point reduction | The IGA measures the investigator's global assessment of overall atopic dermatitis severity on a static, numeric 5-point scale from 0 (clear skin) to 4 (severe disease), based on erythema, papulation/induration, oozing/crusting, and lichenification. Responders have a score of 0 or 1 and a reduction of ≥2 points from baseline. | Baseline to Week 16 |
| EASI-75 | The EASI combines disease extent (0 = 0% to 6 = 90–100%) and severity of erythema, edema/papulation, excoriation and lichenification (each 0–3, half scores allowed) at 4 body sites; the weighted total ranges from 0 to 72. An EASI-75 responder achieves a ≥75% reduction from baseline in EASI score. | Baseline to Week 16 |
Secondary endpoints with posted analyses
- IGA 0/1 with ≥2-point reduction at Weeks 2, 4 and 16; EASI-90 at Weeks 4 and 16; percentage change in EASI at Week 16
- Pruritus NRS: percentage change at Week 16; ≥4-point reduction at Weeks 1, 2, 4 and 16 among participants with baseline NRS ≥4 or ≥5
- Change in percent body surface area (BSA), SCORAD percentage change, DLQI change and ≥4-point DLQI improvement, sleep-loss score change and response, EQ-5D-5L index and VAS, POEM, and PROMIS Anxiety (adolescents) at Week 16
- Maintenance of EASI-75, IGA 0/1 and pruritus response, and SCORAD percentage change, at Week 52 among re-randomized Week 16 responders
The registry notes that five further secondary or other analyses are posted beyond those summarized on this page.
6. Primary Result: IGA 0 or 1 with ≥2-Point Improvement at Week 16
Risk difference, lebrikizumab Q2W vs placebo
95% CI: 14.2 to 29.6 · P = 0.000004 (two-sided)
Cochran-Mantel-Haenszel · superiority · MCMC multiple imputation
The registry reports the comparison of lebrikizumab Q2W with placebo as a risk difference in the percentage of IGA responders. The same estimate, interval and p-value are also listed under a secondary-endpoint entry with an identical definition, which reflects duplicate registration of the Week 16 IGA outcome rather than an independent replication.
What the estimate means. A risk difference of 21.9 means that, after adjustment through the Cochran-Mantel-Haenszel weighting, the proportion of participants reaching IGA 0/1 with at least a 2-point improvement was about 21.9 percentage points higher with lebrikizumab than with placebo. On an absolute scale, roughly 22 additional responders per 100 treated participants would be expected under the conditions of the trial.
What it does not mean. It is not a 21.9% relative increase, and it does not state the response rate in either arm. It also does not mean that 21.9% of participants responded because of the drug in any individual-level sense; it is a population-average contrast between randomized groups.
Precision. The 95% CI of 14.2 to 29.6 spans about 15 percentage points. The whole interval lies above zero, so the data are incompatible (at the 5% level) with no difference, but the plausible absolute benefit ranges from modest to substantial.
The p-value is not the effect size. P = 0.000004 says the observed difference would be very unusual if lebrikizumab and placebo had identical response probabilities. It says nothing about how large or clinically important the difference is; that information is in the estimate and its interval.
Cautions. The result depends on the MCMC imputation model for missing Week 16 assessments and on the exclusion of one site with 18 participants. With two primary endpoints, the overall type I error depends on how the two tests were combined, which the registry entries do not describe.
7. Primary Result: EASI-75 at Week 16
Risk difference, lebrikizumab Q2W vs placebo
95% CI: 24.4 to 42.2 · P < 0.000001 (two-sided)
Cochran-Mantel-Haenszel · superiority · MCMC multiple imputation
What the estimate means. The proportion of EASI-75 responders was an estimated 33.3 percentage points higher with lebrikizumab Q2W than with placebo at Week 16. EASI-75 is a threshold on a relative change in a continuous score, so it captures a clinically meaningful degree of improvement rather than any improvement at all.
What it does not mean. The difference in EASI-75 cannot be compared directly with the IGA difference as if one endpoint were "better". The two endpoints use different instruments and different response thresholds, and a larger risk difference for EASI-75 partly reflects that EASI-75 is an easier threshold to cross than IGA 0/1.
Precision. The 95% CI of 24.4 to 42.2 is about 18 percentage points wide and excludes zero by a wide margin. Even the lower limit represents roughly 24 additional responders per 100 participants.
The p-value is not the effect size. Reporting P < 0.000001 conveys only that the result falls far beyond what chance alone would readily produce under the null hypothesis. The clinical magnitude is described by 33.3 and its interval.
Cautions. Responder analyses dichotomize a continuous score, which discards information about participants just above or below the 75% threshold. The companion continuous analysis (percentage change in EASI, LS mean difference −33.57) is therefore a useful complement.
8. Secondary Endpoints at Week 16
All Week 16 secondary comparisons below are lebrikizumab Q2W versus placebo, reported with two-sided 95% confidence intervals under a superiority hypothesis. Responder endpoints are risk differences in percentage points (CMH); continuous endpoints are least-squares (LS) mean differences from ANCOVA or MMRM. Negative LS mean differences on symptom scales indicate a larger reduction with lebrikizumab.
| Endpoint (Week 16) | Method | Estimate | 95% CI | P-value |
|---|---|---|---|---|
| EASI-90 | CMH, RD | 20.7 | 13.3 to 28.1 | 0.000008 |
| Percentage change in EASI | ANCOVA, LS mean diff. | −33.57 | −41.2 to −26.0 | <0.000001 |
| Percentage change in Pruritus NRS (baseline >0) | ANCOVA, LS mean diff. | −27.53 | −34.9 to −20.2 | <0.000001 |
| Pruritus ≥4-point reduction (baseline NRS ≥4) | CMH, RD | 28.3 | 20.0 to 36.5 | <0.000001 |
| Pruritus ≥4-point reduction (baseline NRS ≥5) | CMH, RD | 29.7 | 21.0 to 38.4 | <0.000001 |
| Change in percent BSA | MMRM, LS mean diff. | −16.2 | −20.3 to −12.0 | <0.000001 |
| Percentage change in SCORAD (LOCF) | ANCOVA, LS mean diff. | −30.76 | −36.93 to −24.59 | <0.000001 |
| Change in DLQI | ANCOVA, LS mean diff. | −4.9 | −6.3 to −3.5 | <0.000001 |
| DLQI ≥4-point improvement (non-missing baseline) | CMH, RD | 32.9 | 22.2 to 43.6 | 0.000001 |
| DLQI ≥4-point improvement (baseline DLQI ≥4) | CMH, RD | 33 | 22.2 to 43.8 | 0.000001 |
| Percentage change in sleep-loss score | ANCOVA, LS mean diff. | −37.09 | −47.4 to −26.8 | <0.000001 |
| Change in sleep-loss score | ANCOVA, LS mean diff. | −0.70 | −0.9 to −0.5 | <0.000001 |
| Sleep-loss ≥2-point reduction (baseline ≥2) | CMH, RD | 18.9 | 9.6 to 28.1 | 0.000571 |
| EQ-5D-5L index, UK (LOCF) | ANCOVA, LS mean diff. | 0.1 | 0.1 to 0.1 | 0.000001 |
| EQ-5D-5L index, US (LOCF) | ANCOVA, LS mean diff. | 0.1 | 0.0 to 0.1 | 0.000001 |
| EQ-5D-5L VAS, mm (LOCF) | ANCOVA, LS mean diff. | 4.5 | 1.3 to 7.6 | 0.005304 |
| Change in POEM | MMRM, LS mean diff. | −6.0 | −7.7 to −4.3 | <0.000001 |
| PROMIS Anxiety T-score, adolescents (LOCF) | ANCOVA, LS mean diff. | −2.91 | −7.85 to 2.03 | 0.241275 |
The Week 16 secondary results are directionally consistent with the primary endpoints across physician-rated severity, itch, sleep and quality-of-life measures. Two entries deserve comment. The EQ-5D-5L index intervals (0.1 to 0.1 and 0.0 to 0.1) are displayed at one decimal place, so their apparent zero width is a rounding artefact rather than perfect precision. The PROMIS Anxiety analysis in adolescents has an interval spanning zero (−7.85 to 2.03), consistent with a small subgroup and an imprecise estimate; it neither demonstrates nor excludes an effect on anxiety.
9. Early Time Points: Onset of Response
Several secondary endpoints examined the same response definitions at earlier visits. These analyses help describe the timing of separation between groups.
| Endpoint | Week | RD (pct. points) | 95% CI | P-value |
|---|---|---|---|---|
| IGA 0/1 with ≥2-point reduction | 2 | 0.7 | −0.3 to 1.7 | 0.308463 |
| IGA 0/1 with ≥2-point reduction | 4 | 8.1 | 4.1 to 12.0 | 0.001607 |
| EASI-90 | 4 | 4.9 | 1.4 to 8.4 | 0.022921 |
| Pruritus ≥4-point reduction (NRS ≥4) | 1 | 0.4 | −0.4 to 1.2 | 0.469221 |
| Pruritus ≥4-point reduction (NRS ≥4) | 2 | 2.7 | −0.1 to 5.4 | 0.113194 |
| Pruritus ≥4-point reduction (NRS ≥4) | 4 | 13.2 | 7.7 to 18.7 | 0.000230 |
| Pruritus ≥4-point reduction (NRS ≥5) | 1 | 0.4 | −0.4 to 1.3 | 0.468160 |
| Pruritus ≥4-point reduction (NRS ≥5) | 2 | 2.9 | −0.1 to 5.8 | 0.115770 |
| Pruritus ≥4-point reduction (NRS ≥5) | 4 | 14.2 | 8.2 to 20.1 | 0.000252 |
At Weeks 1 and 2 the intervals include zero, and by Week 4 the intervals for IGA, EASI-90 and itch response lie above zero. A non-significant early result is not evidence that lebrikizumab has no early effect; with very few responders in either group at Week 1 or 2, a risk difference is necessarily small and the test has little power. The pattern is best read as a trajectory rather than as a sequence of independent yes/no findings.
10. Maintenance Results at Week 52
The maintenance analyses compare re-randomized Week 16 lebrikizumab responders who were withdrawn to placebo with those who continued lebrikizumab at Q4W or Q2W. The registry lists each comparison as "placebo vs lebrikizumab" in the maintenance responder population.
| Week 52 endpoint (Week 16 responders) | Comparison | Estimate | 95% CI | P-value |
|---|---|---|---|---|
| Maintained EASI-75 | Placebo vs Q4W | RD 12.8 | −9.5 to 35.1 | 0.238245 |
| Maintained EASI-75 | Placebo vs Q2W | RD 4.8 | −17.8 to 27.3 | 0.612427 |
| Maintained IGA 0/1 with ≥2-point improvement | Placebo vs Q4W | RD 32.6 | 2.6 to 62.5 | 0.033876 |
| Maintained IGA 0/1 with ≥2-point improvement | Placebo vs Q2W | RD 13.4 | −17.5 to 44.3 | 0.407417 |
| Maintained pruritus response (NRS ≥4) | Placebo vs Q4W | RD 20.3 | −11.4 to 52.0 | 0.159281 |
| Maintained pruritus response (NRS ≥4) | Placebo vs Q2W | RD 20.4 | −10.6 to 51.4 | 0.164950 |
| Maintained pruritus response (NRS ≥5) | Placebo vs Q4W | RD 19.7 | −12.2 to 51.2 | 0.179704 |
| Maintained pruritus response (NRS ≥5) | Placebo vs Q2W | RD 24.3 | −6.5 to 55.1 | 0.083080 |
| Percentage change in SCORAD (LOCF) | Placebo vs Q4W | LS mean diff. −6.29 | −15.46 to 2.87 | 0.176748 |
| Percentage change in SCORAD (LOCF) | Placebo vs Q2W | LS mean diff. −6.25 | −15.48 to 2.99 | 0.183151 |
11. Safety: Serious Adverse Events
| Period / group | Participants with SAE / at risk |
|---|---|
| Induction — placebo | 4 / 149 |
| Induction — lebrikizumab 250 mg Q2W | 2 / 295 |
| Maintenance blinded — lebrikizumab responder groups | 1 / 30; 0 / 59; 2 / 59 |
| Maintenance blinded — placebo responder groups | 0 / 5; 2 / 9; 0 / 11 |
| Escape arm, Week 16 — maintenance open-label (first listed) | 1 / 108 |
| Escape arm, Week 16 — maintenance open-label lebrikizumab (second listed) | 5 / 125 |
| Escape arm, Weeks 24 to 48 | 0 / 11 |
SAEs were infrequent in every group. During induction, 4 of 149 placebo participants and 2 of 295 lebrikizumab participants had an SAE. With so few events, the trial is not designed to detect differences in SAE rates, and no formal SAE comparison is posted. The maintenance and escape groups were formed after Week 16 on the basis of response, so their SAE counts are not randomized comparisons with the induction arms and differ in exposure time.
12. Statistical Methodology
Cochran-Mantel-Haenszel risk difference
All binary endpoints were analyzed with the Cochran-Mantel-Haenszel (CMH) method, which estimates a common risk difference across strata by combining stratum-specific differences with weights. Stratification protects against confounding by stratum and can improve precision when outcome risk differs across strata. The posted entries for the CMH analyses do not name the strata.
where p̂T,k and p̂C,k are response proportions in the treatment and control groups within stratum k. Larger, more balanced strata receive more weight.
Multiple imputation by MCMC
For the primary endpoints and most Week 16 endpoints, missing assessments were imputed with Markov Chain Monte Carlo multiple imputation. Several completed datasets are generated from a model for the missing values, the CMH or ANCOVA analysis is run in each, and the results are combined with Rubin's rules so that the final confidence interval reflects both sampling variability and uncertainty about the missing values.
Q̄ is the pooled estimate over m imputations, Ū the average within-imputation variance, and B the between-imputation variance.
ANCOVA for continuous Week 16 endpoints
Percentage and absolute changes in EASI, pruritus NRS, SCORAD, DLQI, sleep-loss, EQ-5D-5L and PROMIS Anxiety were analyzed with analysis of covariance, reported as LS mean differences. ANCOVA compares groups at a single time point while adjusting for covariates, typically including the baseline value, which reduces residual variance and narrows the confidence interval relative to an unadjusted comparison.
Mixed models for repeated measures
BSA and POEM were analyzed with MMRM. For BSA, the registry lists the model terms: treatment, baseline value, visit, baseline-by-visit interaction, treatment-by-visit interaction, geographic region, age group, and baseline IGA score. MMRM uses all observed post-baseline visits and is valid under a missing-at-random assumption without imputing individual values.
Last observation carried forward
SCORAD, EQ-5D-5L and PROMIS Anxiety analyses used LOCF, which replaces a missing value with the participant's last observed value. LOCF is simple, but it assumes no change after dropout and treats the imputed value as if it were observed, which can understate uncertainty.
Hypothesis framework
Every posted comparison was a superiority test with a two-sided 95% confidence interval. The primary null hypothesis for each endpoint was a risk difference of zero between lebrikizumab Q2W and placebo.
13. Multiplicity
ADvocate2 has two primary endpoints and more than thirty posted secondary comparisons across time points, subsets and maintenance doses. The registry entries do not describe a multiplicity-control strategy (for example, a fixed testing sequence or alpha split), so the p-values shown are nominal.
| Analysis family | Role | Interpretation |
|---|---|---|
| IGA 0/1 and EASI-75 at Week 16 | Primary | Confirmatory comparisons; both intervals exclude zero by a wide margin |
| Week 16 severity, itch, sleep and quality-of-life endpoints | Secondary | Supportive; consistency across endpoints is more informative than any single nominal p-value |
| Weeks 1, 2 and 4 endpoints | Secondary | Describe time course; early null results reflect low event counts |
| Week 52 maintenance endpoints | Secondary | Small re-randomized population; mostly imprecise; one nominally significant result among ten |
With many tests, some nominal p-values below 0.05 would be expected even without real effects. This matters most for isolated borderline results, such as the Week 52 IGA comparison with Q4W (P = 0.033876), and least for the Week 16 results, whose p-values are several orders of magnitude below conventional thresholds.
14. Statistical Methods Explained
A risk difference is expressed on the same percentage-point scale as the response rates themselves, so it translates directly into "additional responders per 100 treated". Odds ratios are harder to interpret when response rates are moderately high, as they can be for EASI-75, because odds ratios then diverge substantially from risk ratios.
CMH combines within-stratum comparisons into one weighted summary. When randomization is stratified, analyzing within those strata respects the design and prevents imbalances between strata from distorting the treatment comparison, while producing a single interpretable estimate.
A responder analysis requires a Week 16 assessment for everyone. Simply dropping participants with missing data changes the population away from "all randomized", and counting all of them as non-responders makes a strong assumption. Multiple imputation fills missing values from a model and carries the extra uncertainty into the confidence interval.
After covariate adjustment, the average percentage reduction in EASI from baseline to Week 16 was about 33.57 percentage points greater with lebrikizumab than with placebo. It is a difference in group means of a percentage change, not a statement that each participant improved by that amount.
The Week 16 comparison includes the whole randomized population; the Week 52 comparisons include only re-randomized lebrikizumab responders split into three groups. Smaller groups mean larger standard errors, so confidence intervals such as −17.8 to 27.3 are several times wider. Wide intervals that include zero indicate uncertainty, not absence of effect.
ANCOVA analyzes the Week 16 value alone (after imputation or LOCF), whereas MMRM models all visits jointly and handles dropouts under a missing-at-random assumption. Both estimate an adjusted treatment difference at Week 16, but they rely on different assumptions about the missing data, so their estimates are not interchangeable.
15. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
Both primary risk differences (21.9 and 33.3 percentage points) had 95% confidence intervals entirely above zero, and the Week 16 secondary endpoints were consistent in direction. Early-visit and Week 52 maintenance comparisons were less precise and mostly inconclusive.
Clinical interpretation
An absolute gain of roughly one additional IGA responder for every five participants treated, and one additional EASI-75 responder for every three, describes the size of the Week 16 effect. Durability after re-randomization is not established with precision by the posted Week 52 comparisons.
16. Important Limitations and Interpretation Issues
- Site exclusion: one investigational site with eighteen participants was excluded from analysis due to GCP issues, so the analyzed population is not strictly all randomized participants.
- Arm-level rates not shown: the posted analyses report risk differences without the corresponding arm-level percentages here, so relative measures and numbers needed to treat must be read from the differences alone.
- Missing-data assumptions: MCMC-MI, LOCF and MMRM make different assumptions; results for endpoints analyzed with LOCF may understate uncertainty.
- Multiplicity: the registry does not describe how type I error was controlled across two primary endpoints and many secondary analyses.
- Maintenance design: the Week 52 comparisons apply only to Week 16 lebrikizumab responders and have limited precision; they do not describe outcomes for all treated participants.
- Subset analyses: itch, sleep and DLQI responder endpoints are restricted to participants above baseline thresholds, and PROMIS Anxiety to adolescents; these populations differ from the primary analysis set.
- Safety precision: SAEs were rare, so the trial provides limited information about uncommon serious harms.
17. Why This Trial Matters Statistically
ADvocate2 is a useful teaching case for responder-based dermatology trials, where the main questions concern proportions, missing data and the transition from induction to maintenance.
| Concept | How it appears in ADvocate2 |
|---|---|
| Risk difference | Absolute difference in responder proportions for IGA, EASI, itch, sleep and DLQI endpoints |
| Cochran-Mantel-Haenszel test | Stratified analysis of all binary endpoints |
| Multiple imputation | MCMC-MI for missing Week 16 assessments |
| ANCOVA | Adjusted LS mean differences for continuous Week 16 outcomes |
| Mixed-effects model | MMRM for BSA and POEM with visit and interaction terms |
| Confidence intervals | Narrow at Week 16, wide in the small Week 52 maintenance groups |
| P-values | Extremely small primary p-values that convey evidence, not magnitude |
| Randomized withdrawal | Re-randomization of Week 16 responders to Q2W, Q4W or placebo |
| Multiplicity | Two primary endpoints plus many secondary comparisons |
18. Related Tutorials
Learn more about the methods used in this trial:
19. Related Calculators
20. Sources
- ClinicalTrials.gov: NCT04178967 — Evaluation of the Efficacy and Safety of Lebrikizumab (LY3650150) in Moderate to Severe Atopic Dermatitis.
- PubMed: PMID 41435984.
- PubMed: PMID 41206702.
- PubMed: PMID 40663228.
- PubMed: PMID 39249591.
- PubMed: PMID 39123054.
Continue exploring clinical biostatistics
Connect the endpoints and methods in this trial to step-by-step tutorials and calculators.
21. Record Summary
ADvocate2 illustrates the statistical anatomy of a modern responder-based phase 3 trial: two binary primary endpoints analyzed as stratified risk differences, multiple imputation for missing assessments, ANCOVA and MMRM for continuous outcomes, and a re-randomized maintenance phase with much smaller groups. The most informative reading combines the absolute risk differences at Week 16, their confidence intervals, the assumptions behind each missing-data method, and the reduced precision of the Week 52 maintenance comparisons.