A Study of Baricitinib (LY3009104) in Patients With Moderate to Severe Atopic Dermatitis: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for A Study of Baricitinib (LY3009104) in Patients With Moderate to Severe Atopic Dermatitis (NCT03334396), a randomized phase 3 study. The primary outcome is Percentage of Participants Achieving Investigator's Global Assessment (IGA) of 0 or 1 With a ≥ 2 Point Improvement (Placebo, 2 mg, or 4 mg Baricitinib).

2. Clinical Question

3. Trial Design

This was a randomized, double-blind, parallel-group, placebo-controlled Phase 3 study sponsored by Eli Lilly and Company. Its official title describes an evaluation of baricitinib’s efficacy and safety as monotherapy in adult patients with moderate to severe atopic dermatitis. The registry reports enrollment of 660 participants and a completed status.

The study began on 2017-11-23, and the primary completion date was 2018-12-06. The principal study groups received placebo or baricitinib at 1 mg, 2 mg, or 4 mg once daily. Matching placebo doses were used in the baricitinib groups. The record also lists Maximum Extended Enrollment cohorts for these groups. However, the primary outcome description says results are reported for the main global study groups and not for the Maximum Extended Enrollment groups.

The primary outcome was assessed at 16 Weeks. It evaluated the percentage reaching an IGA score of 0 or 1 with an improvement of at least 2 points. The primary analysis population is described as all participants randomized to placebo, 2 mg, or 4 mg study drug. The registry gives the arm denominators for this outcome as 249 for placebo, 123 for 2 mg baricitinib, and 125 for 4 mg baricitinib.

The registry record does not report the original sample-size justification, detailed randomization procedures, allocation concealment procedures, or the full protocol analysis plan. It also does not report interim-analysis procedures. Those details cannot be inferred from the randomized, double-blind, parallel-group designation alone.

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.

Only the first 6 of 8 groups are shown.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: All participants randomized to placebo, 2 mg, or 4 mg of study drug. As pre-specified in the analysis plan, outcome measures will not be reported for the Maximum Extended Enrollment (MEE) arms/groups but only for the main global study arms/groups.

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 randomized participants. Results for maximum extended enrollment (MEE) participants will be reported within 1-year after the study completion.

Only the first 6 of 9 groups are shown.

6. Primary Endpoint

The registry lists the primary outcome as Percentage of Participants Achieving Investigator's Global Assessment (IGA) of 0 or 1 With a ≥ 2 Point Improvement (Placebo, 2 mg, or 4 mg Baricitinib), measured over 16 Weeks.

The primary endpoint is the percentage of participants who, at 16 Weeks, had an Investigator’s Global Assessment (IGA) score of 0 or 1 and an improvement of at least 2 points. The IGA is a static, numeric 5-point scale ranging from 0 (clear skin) to 4 (severe disease). The registry describes the assessment as an investigator’s overall judgment of atopic dermatitis severity, based on erythema, papulation or induration, oozing or crusting, and lichenification.

Requiring both a low final score and an improvement of at least 2 points combines a status criterion with a change criterion. The status component identifies participants assessed as clear or almost clear at the endpoint; the change component requires improvement relative to baseline. This kind of responder endpoint can be easier to interpret clinically than an average score change because it expresses how many participants crossed a defined threshold. It also discards detail: participants just below the threshold and participants with little improvement are both counted as nonresponders, despite potentially different score histories.

The registry reports the endpoint as a percentage and identifies the primary outcome population as participants randomized to placebo, 2 mg, or 4 mg, excluding the Maximum Extended Enrollment groups from the reported efficacy outcomes. It does not report the detailed handling of missing endpoint observations in the record.

7. Results Posted in the Registry

• 2 mg Baricitinib: 11.4% (14/123) versus 4.8% (12/249) on control; risk difference 6.6 percentage points (95% CI 0.4 to 12.8), risk ratio 2.36 (1.13–4.95); unadjusted two-proportion z-test p = 0.020.
• 4 mg Baricitinib: 16.8% (21/125) versus 4.8% (12/249) on control; risk difference 12.0 percentage points (95% CI 4.9 to 19.1), risk ratio 3.49 (1.77–6.85); unadjusted two-proportion z-test p = <0.001.

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.

Percentage of Participants Achieving IGA of 0 or 1 With a ≥ 2 Point Improvement (Placebo, 1 mg Baricitinib)

• Endpoint type: Binary: the percentage of participants with the event (counts rebuilt from the posted percentages and group sizes).
• Test: Unadjusted two-proportion z-test, with a Wald 95% confidence interval for the risk difference.
• Result:
  • 1 mg Baricitinib: 11.8% (15/127) versus 4.8% (12/249) on control; risk difference 7.0 percentage points (95% CI 0.8 to 13.2), risk ratio 2.45 (1.18–5.08); unadjusted two-proportion z-test p = 0.013.
• Interpretation:
  • The 95% confidence interval excludes no difference, so the proportion with the event is statistically higher for 1 mg Baricitinib than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.

Percentage of Participants Achieving Eczema Area and Severity Index 75 (EASI75)

• Endpoint type: Binary: the percentage of participants with the event (counts rebuilt from the posted percentages and group sizes).
• Test: Unadjusted two-proportion z-test, with a Wald 95% confidence interval for the risk difference.
• Result:
  • 1 mg Baricitinib: 17.3% (22/127) versus 8.8% (22/249) on control; risk difference 8.5 percentage points (95% CI 1.0 to 16.0), risk ratio 1.96 (1.13–3.40); unadjusted two-proportion z-test p = 0.015.
  • 2 mg Baricitinib: 18.7% (23/123) versus 8.8% (22/249) on control; risk difference 9.9 percentage points (95% CI 2.1 to 17.6), risk ratio 2.12 (1.23–3.64); unadjusted two-proportion z-test p = 0.006.
• Interpretation:
  • The 95% confidence interval excludes no difference, so the proportion with the event is statistically higher for 1 mg Baricitinib than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.
  • The 95% confidence interval excludes no difference, so the proportion with the event is statistically higher for 2 mg Baricitinib than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.

Percentage of Participants Achieving EASI90

• Endpoint type: Binary: the percentage of participants with the event (counts rebuilt from the posted percentages and group sizes).
• Test: Unadjusted two-proportion z-test, with a Wald 95% confidence interval for the risk difference.
• Result:
  • 1 mg Baricitinib: 8.7% (11/127) versus 4.8% (12/249) on control; risk difference 3.8 percentage points (95% CI -1.7 to 9.4), risk ratio 1.80 (0.82–3.96); unadjusted two-proportion z-test p = 0.141.
  • 2 mg Baricitinib: 10.6% (13/123) versus 4.8% (12/249) on control; risk difference 5.7 percentage points (95% CI -0.3 to 11.8), risk ratio 2.19 (1.03–4.66); unadjusted two-proportion z-test p = 0.037.
• Interpretation:
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between 1 mg Baricitinib and the control. The interval gives the range of effects that remain compatible with the data.
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between 2 mg Baricitinib and the control. The interval gives the range of effects that remain compatible with the data.

The posted secondary outcomes cover additional clinical response definitions and symptoms, including IGA response at the 1 mg dose, EASI75, EASI90, SCORAD75, itch response, sleep, and skin pain. Across the posted binary response measures, the broad pattern is that the recorded response percentages generally rise across baricitinib dose groups compared with placebo. The secondary symptom outcomes address different aspects of disease experience and should not be assumed to follow the same pattern solely from the IGA result.

This pattern is broadly consistent with the primary endpoint’s higher recorded response percentages in the 2 mg and 4 mg groups than in placebo. Consistency across related endpoints can help describe whether the result is limited to one definition, but it does not make the outcomes interchangeable. EASI, SCORAD, itch, sleep, and pain measure different features or use different response thresholds. Their findings therefore provide context rather than a single combined confirmation of the primary result.

The registry lists 24 outcome measures and 84 analyses, and reports 23 secondary outcomes. When many outcomes and comparisons are considered, some nominally notable results may arise by chance even if there is no underlying difference for a particular outcome. The registry record does not report a multiplicity adjustment strategy for the secondary analyses. Consequently, unadjusted secondary results carry less evidential weight than a prespecified primary comparison protected by a stated analysis plan. They are useful for describing patterns and generating questions, but should not be treated as independent definitive findings without accounting for the number of tests and the outcomes’ relationships.

The record does not give enough information to establish how all secondary outcomes were prioritized, whether they were prespecified in a particular hierarchy, or how missing observations were handled for each measure. A cautious reading considers the overall direction and clinical variety of the outcomes while avoiding claims that exceed the registry’s analysis details.

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

• Percentage of Participants Achieving IGA of 0 or 1 With a ≥ 2 Point Improvement (Placebo, 1 mg Baricitinib)
• Percentage of Participants Achieving Eczema Area and Severity Index 75 (EASI75)
• Percentage of Participants Achieving EASI90
• Percent Change From Baseline in EASI Score
• Percentage of Participants Achieving SCORing Atopic Dermatitis 75 (SCORAD75)
• Percentage of Participants Achieving a 4-Point Improvement in Itch Numeric Rating Scale (NRS)
• Change From Baseline in the Score of Item 2 of the Atopic Dermatitis Sleep Scale (ADSS)
• Change From Baseline in the Skin Pain Numeric Rating Scale (NRS)

9. Statistical Interpretation

10. Statistical Methods Explained

The primary outcome is binary at the participant level: a participant either meets the IGA response definition or does not. The registry presents arm results as percentages and identifies the outcome parameter as a number of participants. With a binary outcome, comparisons can be expressed in several ways, including a risk difference, a risk ratio, or an odds ratio. These effect measures answer related but distinct questions and should not be treated as interchangeable.

What does a risk difference describe?

A risk difference is the response proportion in one group minus the response proportion in the comparator group. It is an absolute measure on the probability scale. The recalculated estimates are 0.066 for 2 mg versus placebo and 0.120 for 4 mg versus placebo. The intervals around these estimates express the uncertainty in the estimated absolute differences under the recalculation method. Because the recalculations are simple and unadjusted, they should be distinguished from the sponsor’s reported regression analysis.

What does the logistic-regression odds ratio describe?

Logistic regression models the odds of a binary outcome. The sponsor reports odds ratios of 2.61 for 2 mg versus placebo and 4.10 for 4 mg versus placebo. An odds ratio above 1 indicates higher estimated odds in the baricitinib group than in placebo under that analysis. It does not mean that the response percentage is multiplied by the odds ratio. The record names logistic regression, but does not report the model’s covariates, detailed coding choices, or other model specifications. Those features should not be inferred.

Why do analysis methods sometimes give different estimates?

Different effect measures summarize the same data on different scales. In addition, a regression model may adjust for variables or use modeling choices that differ from an unadjusted comparison of posted counts. The registry record does not include enough detail to attribute the discrepancy to a particular modeling choice. The sponsor’s estimates and tests are the reported trial analysis; the recalculations are unadjusted checks based on summary statistics. Given the recorded discrepancy, the sponsor’s own analysis takes precedence.

How should p-values be read?

A p-value measures how compatible the observed data are with a specified null hypothesis under the assumptions of the analysis. It is not the probability that the null hypothesis is true, nor is it a measure of the size or clinical importance of a difference. The sponsor reports p = 0.020 for 2 mg and p < 0.001 for 4 mg in the primary comparison. These reported values belong to the sponsor’s logistic-regression analyses. The recalculated p-values are unadjusted and should not be substituted for the sponsor’s reported tests.

Interpretation also depends on whether the analysis accounts for multiple comparisons. The registry reports numerous outcomes and analyses, but the multiplicity procedure for the primary and secondary comparisons is not reported in the registry record. Without that information, the reported p-values should be read as analyses on their stated terms, not as evidence that all comparisons were protected against the chance of false positive findings arising from many tests.

What do the analysis population and denominators tell us?

The primary outcome description states that the population includes all participants randomized to placebo, 2 mg, or 4 mg study drug. The arm denominators recorded for the endpoint are 249, 123, and 125, respectively. This description identifies the intended population, but does not establish how participants with missing endpoint data were treated in the analysis. The missing-data method, imputation approach, and any sensitivity analyses are not reported in the registry record. The record also states that outcome measures are not reported for the Maximum Extended Enrollment groups.

These distinctions matter because the denominator and the handling of missing outcomes can affect a binary response percentage. The registry’s description should be reported as written, without assuming a particular intention-to-treat implementation or missing-data rule. More detailed protocol or statistical-analysis information is not reported in the registry record.

What can the posted summaries support?

The arm percentages describe the recorded response levels, while confidence intervals communicate uncertainty around comparisons. The sponsor’s odds ratios summarize the reported regression analysis. The unadjusted recalculations provide a separate, transparent contrast based on the posted summary statistics. None of these summaries alone establishes the adequacy of the original design, the full set of prespecified analyses, or the effect of unreported analysis decisions. Those aspects require information that is not reported in the registry record.

11. Confidence Intervals and Precision

The sponsor’s primary comparisons include 95% confidence intervals for the odds ratios. For 2 mg baricitinib versus placebo, the reported odds ratio is 2.61, with a 95% confidence interval from 1.17 to 5.84. For 4 mg versus placebo, it is 4.10, with a 95% confidence interval from 1.93 to 8.70. These intervals indicate the range of effect estimates reasonably compatible with the data and the analysis assumptions, using the confidence-interval procedure. A confidence interval is not a guarantee that the true effect lies within its endpoints.

The intervals are fairly broad, especially for the 2 mg comparison, so the estimates are not highly precise. Their lower and upper limits also show that the data are compatible with a range of odds-ratio magnitudes. The interval for each sponsor comparison excludes 1, consistent with the direction of the reported superiority analysis, but interval width remains important: statistical evidence against a null value does not imply a narrowly estimated effect.

The unadjusted recalculated risk-difference intervals use a different scale. For 2 mg versus placebo, the estimate is 0.066, with a 95% confidence interval from 0.004 to 0.128. For 4 mg versus placebo, it is 0.120, with a 95% confidence interval from 0.049 to 0.191. These intervals communicate uncertainty in absolute response differences under the simple recalculation, not in the sponsor’s regression estimates. The registry’s discrepancy means the sponsor’s own analysis takes precedence.

The arm-level response percentages do not have confidence intervals attached in the posted summaries. Precision should therefore be judged primarily from the comparative intervals rather than from the percentages alone. The registry record does not report the full analysis details needed to assess all sources of uncertainty, including missing-data handling and multiplicity procedures.

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 sizes as participants per arm. They are not total enrollment targets and do not represent observed trial results. For the primary comparisons, the illustrations use the observed effect as a reference and show planning for 100%, 75%, and 50% of that observed effect. For example, the 2 mg comparison lists 271 participants per arm for 80% power and 362 participants per arm for 90% power when planning for 100% of the observed effect. The 4 mg comparison lists 105 participants per arm for 80% power and 140 participants per arm for 90% power under the same effect assumption.

The stated assumptions are a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation. The effect to detect is the stated fraction of the observed effect. The 80% power column gives the estimated participants per arm for a design with an 80% chance of detecting the specified effect under those assumptions, if that effect is the true one. The 90% power column gives the corresponding planning figure for a 90% chance. These are design calculations, not descriptions of what happened in the completed trial.

Smaller effects are harder to detect, so the illustrations require more participants per arm when planning for a smaller fraction of the observed effect. For the 2 mg comparison, the listed figures rise from 271 participants per arm for 80% power at 100% of the observed effect to 880 when planning for 50% of it. For 4 mg, the corresponding figures are 105 and 314 participants per arm. The registry record does not report whether these assumptions match the trial’s original design rationale. These planning illustrations are not observed power.

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, p₀ and p₁ are the control and treatment proportions, and p̄ is their average.

Worked calculation for this trial

1. Control proportion p₀ = 0.048; treatment proportion p₁ = 0.114; average p̄ = 0.081.
2. First term: z_1−α/2 × √[2p̄(1−p̄)] = 1.96 × 0.3859 = 0.7563.
3. Second term: z_1−β × √[p₀(1−p₀) + p₁(1−p₁)] = 0.842 × 0.3831 = 0.3224.
4. n = (0.7563 + 0.3224)² / (0.0656)² = 270.1, rounded up to 271 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
p0 <- 0.0481927710843374                          # control proportion
p1 <- 0.113821138211382                          # treatment proportion
n_per_arm <- function(power, frac = 1) {
  q1 <- p0 + (p1 - p0) * frac               # smaller effect = fraction of the observed difference
  pbar <- (p0 + q1) / 2
  z_b <- qnorm(power)
  ceiling(round((z_a * sqrt(2 * pbar * (1 - pbar)) + z_b * sqrt(p0 * (1 - p0) + q1 * (1 - q1)))^2 / (q1 - p0)^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 (unrounded n per arm)
cat("power.prop.test:", round(power.prop.test(p1 = p0, p2 = p1, power = 0.80)$n, 1), "\n")

# Simulation: how often does the z-test reject at the 80%-power sample size?
set.seed(3334396)
n    <- n_per_arm(0.80)
reps <- 20000
x0 <- rbinom(reps, n, p0); x1 <- rbinom(reps, n, p1)
ph <- (x0 + x1) / (2 * n)
z  <- ((x1 - x0) / n) / sqrt(ph * (1 - ph) * 2 / n)
cat(sprintf("Simulated power at n = %s per arm: %.3f (%d replicates)\n", format(n, big.mark = ","), mean(abs(z) > z_a, na.rm = TRUE), reps))

Output from running the code

100% of observed effect: n per arm = 271 (80% power), 362 (90% power)
 75% of observed effect: n per arm = 437 (80% power), 584 (90% power)
 50% of observed effect: n per arm = 880 (80% power), 1,178 (90% power)
power.prop.test: 270.1 
Simulated power at n = 271 per arm: 0.817 (20000 replicates)

The simulation draws many pairs of event counts at the planned sample size and counts how often the two-proportion z-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 primary comparisons from logistic regression as odds ratios with 95% confidence intervals and p-values. The 2 mg comparison is reported as an odds ratio of 2.61, with a 95% confidence interval from 1.17 to 5.84 and p = 0.020. The 4 mg comparison is reported as an odds ratio of 4.10, with a 95% confidence interval from 1.93 to 8.70 and p < 0.001.

The simple recalculations use the posted arm-level summary counts and estimate unadjusted risk differences. They yield 0.066 for 2 mg versus placebo and 0.120 for 4 mg versus placebo. The point estimates, confidence intervals, and p-values do not exactly match the sponsor’s reported analysis because they are not the same analysis or effect scale. The registry record flags a discrepancy, but does not report its cause or provide enough model detail to resolve it.

For interpretation, the sponsor’s own analysis takes precedence. The recalculations are best treated as a transparent, unadjusted cross-check, not as a correction to the sponsor’s regression results. Their value is in making a simple comparison from the posted summaries explicit while keeping distinct the estimand, method, and assumptions behind each result.

15. Multiplicity, Missing Data and Analysis Populations

Multiplicity arises when a study evaluates several outcomes, treatment comparisons, or analysis choices. As the number of statistical tests grows, the chance of at least one apparently notable result can increase even if the tested effects are absent. This is especially relevant to a record listing 24 outcome measures and 84 analyses, including multiple baricitinib doses and multiple response definitions.

The registry record does not report a multiplicity adjustment procedure, a testing hierarchy, or whether secondary comparisons were considered confirmatory or exploratory. It also does not report whether the primary comparisons were adjusted for the number of dose comparisons. Therefore, the reported p-values should be interpreted in the context of the stated individual analyses, not as if control of the overall false-positive rate were established. The distinction is important: a result may be statistically notable under an individual test while the broader set of tests remains subject to multiplicity concerns.

The primary outcome description identifies participants randomized to placebo, 2 mg, or 4 mg study drug as the analysis population. However, the registry record does not report the detailed missing-data rules, imputation methods, or sensitivity analyses used for that outcome. It does not report whether all randomized participants contributed an observed endpoint or how any missing endpoint values were classified. The statement describing the population should not be expanded into an assumed missing-data method.

The registry record also does not report the original sample-size justification, interim-analysis plan, or full statistical analysis plan. Those omissions limit assessment of whether the final analyses followed a prespecified strategy and how the investigators addressed all planned comparisons. The record identifies logistic regression for the sponsor’s primary comparisons, but does not give its covariates or full model specification. These are limitations in the information available for statistical interpretation, rather than evidence that a particular method was or was not used.

16. 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: dermatitis atopic (0.8%), alcohol poisoning (0.8%), clavicle fracture (0.4%), rib fracture (0.4%), breast cancer (0.4%).

The registry reports serious adverse events and other adverse events by group, with the number at risk and the number affected. In the main placebo group, 6 of 249 participants had a serious adverse event; the corresponding figures were 1 of 127 in the 1 mg group, 0 of 123 in the 2 mg group, and 2 of 125 in the 4 mg group. No deaths are recorded in these groups. The Maximum Extended Enrollment groups are listed separately, and no deaths are recorded there either.

These are counts from the registry record, not a complete characterization of event timing, duration, severity, or causal attribution. The record lists serious-event terms including atopic dermatitis and alcohol poisoning at 0.8%, and clavicle fracture, rib fracture, breast cancer, and papillary thyroid cancer at 0.4% each. The term-level percentages do not establish that an event was caused by study treatment.

Adverse-event counts should be read descriptively. The registry record does not provide enough information to assess comparative risk with adjusted estimates, exposure-adjusted rates, or detailed event narratives. Small counts also leave limited information about uncommon events.

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

The primary registry results show higher recorded IGA response percentages at 16 Weeks in the 2 mg and 4 mg baricitinib groups than in placebo. The sponsor’s logistic-regression comparisons report odds ratios above 1, with confidence intervals that exclude 1 for both primary comparisons. The 4 mg group has the highest recorded response percentage and the larger sponsor-reported odds ratio. These statements describe the posted results and their reported statistical comparisons; they do not provide a complete account of each participant’s outcome.

The result concerns a specific responder definition: an IGA score of 0 or 1 together with an improvement of at least 2 points. It does not mean every participant in a baricitinib group met that definition, and the group percentages do not predict an individual participant’s outcome. The endpoint also does not capture every dimension of atopic dermatitis or establish effects on outcomes not reported at the stated assessment time.

Statistical evidence and clinical importance are distinct. A p-value describes compatibility with a null model under analysis assumptions; it does not measure the size of an effect or determine whether the effect matters to patients. The odds ratios and risk differences express comparisons on different scales. The confidence intervals indicate uncertainty, while the discrepancy between the recalculation and sponsor analysis calls for relying on the sponsor’s reported analysis as the primary inferential result.

The record does not report all design and analysis details needed to assess multiplicity, missing-data handling, or the original sample-size rationale. Secondary findings can add context but should be interpreted cautiously in view of the number of outcomes and analyses. The registry results alone do not establish longer-term outcomes, nor do they provide a basis for individual medical decisions.

18. Why This Trial Matters Statistically

This trial illustrates why statistical interpretation requires more than comparing arm percentages. The primary endpoint is a binary response, and its observed percentages give an accessible description of the groups. Comparative analysis adds uncertainty estimates and formal tests, but the effect measure matters: the sponsor reports odds ratios, while the simple recalculation reports risk differences. Those quantities answer different questions and cannot be read as if they were the same scale.

The record also makes the distinction between a sponsor analysis and a summary-statistic recalculation concrete. The sponsor used logistic regression, while the recalculation is unadjusted. Because the registry record flags a discrepancy, the sponsor’s analysis takes precedence. Reporting both with their methods made explicit avoids presenting an unadjusted check as though it were the trial’s principal result.

Precision matters alongside statistical evidence. The sponsor’s confidence intervals are informative about the range of estimates compatible with the analyses, and their width cautions against treating a point estimate as exact. The sample-size illustrations further show, as a general planning principle, that smaller effects require more participants per arm to achieve a chosen power under stated assumptions. Those illustrations are not evidence about the trial’s original design rationale.

Finally, the record includes many outcome measures and analyses. Without a reported multiplicity strategy, secondary results should be interpreted as a broader pattern rather than as a collection of independent confirmatory claims. These are transferable lessons for reading registry results: identify the estimand, distinguish effect scales, examine uncertainty, and consider the analysis population and the number of comparisons before drawing conclusions.

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.
• Participant counts were reconstructed from posted percentages and arm sizes, so small rounding differences are possible.

The primary endpoint is reported at 16 Weeks, so the record does not establish how the response pattern may extend to other time points. The endpoint is a threshold-based IGA response, which simplifies interpretation but does not preserve all information about participants’ score changes or other aspects of disease severity. Arm-level percentages are posted without confidence intervals in the summaries, leaving the comparative estimates to carry most of the displayed precision information.

The registry record reports sponsor logistic-regression odds ratios and unadjusted recalculations from summary statistics, and it flags a discrepancy between them. It does not report enough detail to identify the source of that discrepancy, such as model covariates or other analysis specifications. The sponsor’s analysis takes precedence, but the missing detail limits independent assessment of the model.

The registry lists numerous outcomes and analyses, while a multiplicity procedure is not reported. The handling of missing data, sensitivity analyses, and detailed analysis populations beyond the primary population description are also not reported in the registry record. The original sample-size justification and interim-analysis plan are not reported either. These gaps limit evaluation of how the design and analysis plan addressed all planned questions.

The primary outcome description excludes the Maximum Extended Enrollment groups from reported efficacy outcomes, even though those cohorts appear in the safety record. The resulting efficacy interpretation therefore applies to the main global study groups described for the endpoint. The registry record also does not supply detailed adverse-event narratives or exposure-adjusted safety comparisons, so the safety counts should remain descriptive.

20. Related Tutorials

21. Related Statistical Calculators

22. Sources

• ClinicalTrials.gov: A Study of Baricitinib (LY3009104) in Patients With Moderate to Severe Atopic Dermatitis — NCT03334396.
• Registry record last updated 2020-08-18; status last verified 2020-01.
• Results first posted 2020-01-18.
• 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

NCT03334396 is a completed, randomized, double-blind, parallel-group, placebo-controlled Phase 3 study of baricitinib monotherapy in adults with moderate to severe atopic dermatitis. The registry reports 660 participants enrolled. The primary endpoint was the percentage achieving IGA 0 or 1 with at least a 2-point improvement at 16 Weeks, for placebo, 2 mg, and 4 mg groups.

Observed response percentages were 4.8% for placebo, 11.4% for 2 mg baricitinib, and 16.8% for 4 mg baricitinib. The sponsor reported logistic-regression odds ratios of 2.61 and 4.10 for the 2 mg and 4 mg comparisons, respectively. Because the registry record flags a discrepancy with simple unadjusted recalculations, the sponsor’s analysis takes precedence. The record lists multiple secondary outcomes and adverse-event counts, while several important analysis details, including multiplicity and missing-data methods, are not reported in the registry record.