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
KNOCKOUT was a randomized, parallel, triple-masked phase 2 treatment trial in psoriasis with 20 participants allocated across two risankizumab dose groups. Its two primary endpoints examined CD8+ Trm1 and CD8+ Trm17 cells in lesional skin at baseline and Week 52 using longitudinal single-cell RNA-sequencing of full-thickness skin biopsy samples.
| Feature | KNOCKOUT |
|---|---|
| Trial acronym | KNOCKOUT |
| Phase | Phase 2 |
| Condition | Psoriasis |
| Intervention | Risankizumab |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Triple |
| Primary purpose | Treatment |
| Enrollment | 20 |
| Number of arms | 2 |
| Status | Completed |
| Start date | March 1, 2022 |
| Primary completion date | August 7, 2023 |
| Lead sponsor | Oregon Medical Research Center |
| ClinicalTrials.gov | NCT05283135 |
2. Clinical Question
The central question is whether different risankizumab dose groups are associated with changes in specific inflammatory T-cell populations in lesional skin over a 52-week period.
Population
Participants with psoriasis enrolled in a phase 2 treatment trial.
Intervention
Risankizumab, administered in the trial's 600 mg and 300 mg dose groups.
Comparator
The other randomized risankizumab dose group provides the parallel-group comparison.
Primary question
How do CD8+ Trm1 and CD8+ Trm17 cell populations in lesional skin change between baseline and Week 52?
3. Trial Design
600 mg Dose Group
- Risankizumab
- 10 participants at risk for the reported serious-adverse-event summary
- 0 participants with a serious adverse event in that summary
300 mg Dose Group
- Risankizumab
- 10 participants at risk for the reported serious-adverse-event summary
- 1 participant with a serious adverse event in that summary
4. Endpoints
The two registered primary endpoints are biological, longitudinal measures rather than conventional clinical time-to-event outcomes. Both are assessed at baseline and Week 52.
| Primary endpoint | Time frame | Registry definition |
|---|---|---|
| CD8+ Trm1 Cells in Lesional Skin at Baseline and Week 52 | 52 weeks | The number of lesional CD8+ Trm1 cells at baseline and Week 52 was calculated by longitudinal single-cell RNA-sequencing of full-thickness skin biopsy samples. Uniform Manifold Approximation Projection (UMAP) was used for T cell subtype visualization and CD8+ Trm1 cells were identified as IFNγ+/CD8+/CD69+ T cells. |
| CD8+ Trm17 Cells in Lesional Skin at Baseline and Week 52 | 52 weeks | The number of lesional CD8+ Trm17 cells at baseline and Week 52 was calculated by longitudinal single-cell RNA-sequencing of full-thickness skin biopsy samples. Uniform Manifold Approximation Projection (UMAP) was used for T cell subtype visualization and CD8+ Trm17 cells were identified as IFNγ+/CD8+/CD69+ T cells. |
What makes these endpoints unusual?
These endpoints are based on cellular measurements obtained from tissue rather than conventional clinical outcomes such as a symptom score or a binary response rate. Each participant can contribute observations at two time points, so the statistical structure is inherently longitudinal.
The use of single-cell RNA-sequencing also means that the endpoint is preceded by a biological classification step. UMAP provides a way to visualize cell populations in a lower-dimensional representation, while the registry definition identifies the target populations using the specified IFNγ, CD8, and CD69 characteristics.
5. Results
The ClinicalTrials.gov record indicates that results have been posted for both primary endpoints. However, no formal statistical analyses are posted in the record, and the available registry information does not provide numerical estimates, confidence intervals, or p-values for the two primary endpoints.
CD8+ Trm1 Cells in Lesional Skin
Registered primary endpoint
52-week longitudinal assessment of lesional CD8+ Trm1 cells using single-cell RNA-sequencing.
The registry defines CD8+ Trm1 cells as IFNγ+/CD8+/CD69+ T cells and specifies full-thickness skin biopsy samples for the longitudinal single-cell RNA-sequencing assessment. No numerical treatment-group estimate or inferential statistic is reported in the available record information.
The statistical question for this endpoint is fundamentally a change-over-time comparison. A useful analysis would distinguish the within-participant change from baseline to Week 52 from the difference in those changes between the 600 mg and 300 mg groups.
A difference between dose groups at Week 52 alone would not fully answer the longitudinal question because the groups could have differed at baseline. Conversely, a change within one group does not establish a between-group treatment effect.
With only 20 participants overall, estimates of change are expected to be imprecise. A confidence interval would therefore be particularly important for describing the uncertainty around the estimated difference in longitudinal change. A p-value, if reported, would address compatibility with a specified null hypothesis; it would not measure the magnitude or biological importance of the effect.
CD8+ Trm17 Cells in Lesional Skin
Registered primary endpoint
52-week longitudinal assessment of lesional CD8+ Trm17 cells using single-cell RNA-sequencing.
The registry defines CD8+ Trm17 cells as IFNγ+/CD8+/CD69+ T cells and specifies the same longitudinal single-cell RNA-sequencing approach using full-thickness skin biopsy samples. No numerical treatment-group estimate or inferential statistic is reported in the available record information.
The same longitudinal principle applies to Trm17 cells: the key comparison would ordinarily be the difference in baseline-to-Week-52 change between the two randomized dose groups.
Because cell-count measurements can be skewed and may have a substantial range across participants, the scale used for analysis matters. An arithmetic mean difference, median difference, ratio, or model-based contrast would answer somewhat different questions. The registry endpoint definition alone does not specify which effect measure or model was used.
Any inferential result should also be interpreted in light of the small sample size. Wide confidence intervals would indicate substantial uncertainty even if a point estimate appeared large, while a non-significant p-value would not by itself establish equivalence of the dose groups.
6. Planned Analysis
The registry endpoints call for longitudinal measurement of lesional CD8+ Trm1 and CD8+ Trm17 cells at baseline and Week 52. For an endpoint of this type, the statistical analysis would normally focus on the change within each participant and the contrast in change between the randomized dose groups.
| Statistical question | Typical approach | What it answers |
|---|---|---|
| Did cell counts change over time? | Paired analysis or a repeated-measures model | Whether the measured cellular population changed between baseline and Week 52. |
| Did the dose groups change differently? | Group-by-time comparison or analysis of change scores | Whether the longitudinal trajectory differed between the 600 mg and 300 mg groups. |
| How large was the between-group difference? | Estimated difference in change with a confidence interval | The magnitude and precision of the dose-group contrast. |
| How should non-normal cell counts be handled? | Transformation or an appropriate count/distributional model | A model that better reflects the scale and distribution of the cellular measurements. |
7. Statistical Methodology
Longitudinal change
Each primary endpoint contains two measurements per participant: a baseline measurement and a Week 52 measurement. This paired structure is statistically important because observations from the same participant are correlated.
For participant i, the change score compares the Week 52 cellular measurement with the corresponding baseline measurement. A between-group analysis then compares the distribution or model-based mean of these changes.
A simple change-score analysis is not the only possible approach. An analysis of covariance can model the Week 52 measurement while adjusting for baseline, and a repeated-measures model can explicitly represent the correlation between measurements. The appropriate choice depends on the measurement distribution, missing observations, and prespecified statistical analysis plan.
Between-group comparison
Randomization creates the framework for comparing the two dose groups. The principal treatment-related question is not simply whether either group changes over time, but whether the pattern of change differs between groups.
This contrast asks whether the average longitudinal change differs between the two randomized dose groups.
Repeated-measures modeling
A repeated-measures model can represent both the time effect and the treatment-group effect while accounting for the fact that baseline and Week 52 measurements from the same participant are not independent.
A general model might contain terms for dose group, time, and their interaction. The interaction is especially informative because it represents whether the change from baseline differs between the two groups.
Cell-count distributions
Cellular measurements may not satisfy the assumptions of a conventional normal-theory model. Depending on the actual distribution, an analyst might consider a transformation or a model designed for count data. The key principle is that the statistical scale should reflect how the endpoint behaves in the observed data.
Single-cell classification and UMAP
UMAP is primarily a visualization and dimensionality-reduction technique. It can help display relationships among individual cells and visualize candidate cellular subtypes. The registry separately defines the target CD8+ Trm1 and CD8+ Trm17 populations using specified cellular markers.
Importantly, visualization and inference are different stages of analysis. A visible separation of cell clusters on a UMAP plot is not, by itself, a treatment-effect test. Formal inference requires an appropriate unit of analysis and a prespecified statistical comparison.
8. Statistical Methods Explained
Why is a longitudinal analysis appropriate?
Because each primary endpoint is measured at baseline and Week 52, the two observations from an individual participant are related. Treating them as independent measurements would ignore useful within-participant information and could produce incorrect uncertainty estimates.
Why compare changes rather than only Week 52 values?
The endpoints are explicitly defined at both baseline and Week 52. Comparing changes accounts for where participants started. A Week 52-only comparison can be informative, particularly with baseline adjustment, but it does not directly describe the individual longitudinal change.
What does a group-by-time interaction mean?
The interaction asks whether the amount of change over time differs between the 600 mg and 300 mg groups. This is closely aligned with the scientific question when the endpoint is framed as baseline versus Week 52.
Why might cell counts require transformation?
Biological count measurements can be asymmetric, contain many small values, or vary substantially between participants. If the distribution is strongly skewed, a transformation or count-oriented model can provide a more appropriate representation of the data than an unmodified normal model.
What does UMAP contribute to the analysis?
UMAP reduces high-dimensional single-cell information into a lower-dimensional representation that can be visualized. It helps identify and display cellular subtypes, but the visual location of a cluster is not itself a p-value or treatment-effect estimate.
Why is the small sample size important?
With 20 participants overall, there is limited information for estimating between-group differences. Small studies can produce unstable point estimates and wide confidence intervals. They also have limited ability to distinguish modest biological differences from sampling variation.
9. Primary Endpoint Interpretation
Trm1 endpoint
The endpoint measures the number of lesional CD8+ Trm1 cells at baseline and Week 52 using longitudinal single-cell RNA-sequencing.
Trm17 endpoint
The endpoint measures the number of lesional CD8+ Trm17 cells at baseline and Week 52 using the same longitudinal single-cell approach.
Biological interpretation
A change in cellular abundance describes a biological measurement. It is not automatically equivalent to a clinical response unless a validated relationship between the cellular endpoint and clinical outcome has been established.
Statistical interpretation
A between-dose difference requires an estimate of the contrast, an appropriate measure of uncertainty, and a clearly defined analysis population and model.
10. Safety Results
The registry reports serious adverse events by dose group. The 600 mg Dose Group had 0 affected participants among 10 at risk, while the 300 mg Dose Group had 1 affected participant among 10 at risk.
| Safety measure | 600 mg Dose Group | 300 mg Dose Group |
|---|---|---|
| Serious adverse events | 0/10 | 1/10 |
These counts are descriptive rather than a basis for a reliable comparative safety inference. With 10 participants per dose group, a difference of one affected participant can materially change the observed proportion. Confidence intervals for adverse-event rates would be wide, and formal comparisons would have very limited precision.
11. How to Interpret the Safety Counts
The registry reports 0/10 serious-adverse-event-affected participants in the 600 mg Dose Group and 1/10 in the 300 mg Dose Group.
Those observations describe what occurred among the participants represented in the registry safety summary. They do not establish that one dose is safer than the other because the study is small and the event counts are sparse.
When each group contains 10 participants, one additional event changes the observed proportion by a substantial amount. Statistical uncertainty around a proportion can therefore be much larger than the point estimate alone suggests.
The cellular primary endpoints and serious-adverse-event counts answer different questions. A biological change does not establish clinical benefit, and a small difference in serious adverse events does not establish a comparative safety profile.
12. Missing Data and Longitudinal Measurements
Longitudinal tissue-based endpoints create an important analytical issue: a participant may have a baseline measurement but not a Week 52 measurement. The available registry information does not specify a missing-data or imputation method.
For a two-time-point endpoint, the treatment of missing Week 52 observations can materially affect the estimated change. Common approaches include analyses based on observed paired measurements, model-based repeated-measures methods, or prespecified imputation strategies. These approaches make different assumptions and should not be treated as interchangeable.
13. Multiple Primary Endpoints
KNOCKOUT has two registered primary endpoints: CD8+ Trm1 cells and CD8+ Trm17 cells, both evaluated over 52 weeks. When a study has more than one primary endpoint, the statistical interpretation depends on how success across those endpoints is defined and whether multiplicity is controlled.
| Endpoint | Role | Time frame | Statistical issue |
|---|---|---|---|
| CD8+ Trm1 cells | Primary | 52 weeks | Longitudinal cellular comparison |
| CD8+ Trm17 cells | Primary | 52 weeks | Longitudinal cellular comparison |
The available registry information does not specify an endpoint hierarchy, multiplicity adjustment, or alpha allocation for the two primary endpoints. Consequently, a p-value for either endpoint would need to be interpreted in the context of the prespecified multiplicity strategy rather than automatically treated as an isolated confirmatory test.
14. What a Confidence Interval Would Add
For these endpoints, a confidence interval around the estimated difference between dose groups would provide information that a p-value alone cannot provide.
The point estimate describes the observed or model-estimated difference, while the confidence interval describes the statistical uncertainty around that estimate under the specified analysis framework.
A narrow interval would indicate greater precision; a wide interval would indicate less precision. In a study with 20 participants, uncertainty is an especially important part of interpretation.
A confidence interval also does not describe the range of cellular changes experienced by individual participants. It describes uncertainty around the estimated population-level contrast under the statistical model used.
15. Statistical Unit of Analysis
Single-cell sequencing creates many individual cell-level observations, but the randomized treatment comparison is fundamentally a participant-level question. This distinction is critical.
Cells are not participants
Thousands of measured cells from one participant do not create thousands of independent randomized treatment assignments.
Randomization occurs at the participant level
The two dose groups are randomized groups of participants. Treatment-effect inference should respect that experimental unit.
Cell-level structure can still matter
Single-cell measurements contain rich within-sample information, but dependence among cells from the same participant must be recognized.
Pseudoreplication risk
Treating every cell as an independent participant can substantially understate uncertainty and produce misleadingly precise treatment comparisons.
16. Longitudinal Trial History
Trial start
The KNOCKOUT study began enrollment as a phase 2 randomized, parallel, triple-masked treatment trial in psoriasis.
Primary endpoint assessment
The registered primary endpoints evaluate lesional CD8+ Trm1 and CD8+ Trm17 cells at baseline and Week 52 using longitudinal single-cell RNA-sequencing of full-thickness skin biopsy samples.
Primary completion
The registry lists August 7, 2023 as the primary completion date.
Registry status
The ClinicalTrials.gov record lists the study as completed and indicates that results have been posted.
17. Limitations and Interpretation Issues
- Small sample size: the registry itself identifies small sample size as a limitation. With 20 participants overall, estimates of treatment differences are inherently imprecise.
- Sparse safety events: the serious-adverse-event summary contains 0 affected participants in one group and 1 in the other, limiting comparative safety inference.
- Two primary endpoints: the study has both Trm1 and Trm17 endpoints. The available registry information does not specify a multiplicity strategy for their joint interpretation.
- Longitudinal dependence: baseline and Week 52 measurements from the same participant are correlated and should not be treated as independent observations.
- Cell-level dependence: multiple cells from the same participant are not equivalent to independent randomized participants. Ignoring this hierarchy can understate uncertainty.
- Missing-data considerations: longitudinal tissue measurements can be incomplete, and the available registry information does not specify an imputation strategy.
- Biological endpoint interpretation: changes in CD8+ Trm1 or Trm17 cell populations are biological measurements and should not automatically be interpreted as clinical outcomes.
- Limited formal statistical detail: the registry endpoint descriptions identify the measurement and classification approach but do not provide a formal inferential model or numerical effect estimates in the available record information.
18. Why This Trial Matters Statistically
KNOCKOUT is a useful teaching case because it combines randomized treatment allocation with a highly specialized longitudinal biological endpoint. The statistical challenge is not simply to compare two numbers. It is to connect participant-level randomization, repeated measurements, cellular classification, and uncertainty into a coherent analysis.
| Concept | How it appears in KNOCKOUT |
|---|---|
| Randomization | Participants were randomized to 2 parallel dose groups. |
| Small-sample inference | 20 participants were enrolled overall. |
| Longitudinal analysis | Primary endpoints compare measurements at baseline and Week 52. |
| Repeated measurements | Each participant can contribute measurements at two time points. |
| Cellular phenotyping | CD8+ Trm1 and CD8+ Trm17 cells are identified using specified markers. |
| Dimensionality reduction | UMAP is used for T-cell subtype visualization. |
| Hierarchical data | Multiple single cells can be measured within an individual participant. |
| Between-group contrast | The central statistical comparison is the difference in longitudinal change between dose groups. |
| Multiplicity | There are 2 registered primary endpoints. |
| Safety description | Serious adverse events are reported by randomized dose group. |
19. A Statistical Framework for Reading the Primary Endpoints
For each participant, establish the baseline and Week 52 cellular measurement for the endpoint being studied. The analysis should preserve the participant as the fundamental experimental unit.
Calculate or model the change from baseline to Week 52. The choice between raw differences, transformed measurements, ratios, or a repeated-measures model depends on the underlying distribution and prespecified analysis.
Estimate how the longitudinal change differs between the 600 mg and 300 mg dose groups. This is the treatment comparison that most directly reflects the randomized design.
Report a confidence interval around the between-group effect estimate. With a small study, the width of this interval is essential to understanding the strength of the evidence.
Even a statistically clear difference in a cellular population would describe a biological effect. Its clinical significance requires a separate assessment of how that cellular measurement relates to psoriasis outcomes.
20. Statistical Interpretation vs Biological Interpretation
Statistical interpretation
The randomized design supports a between-dose comparison, while the baseline and Week 52 structure calls for a longitudinal analysis that accounts for within-participant dependence.
Biological interpretation
The primary endpoints characterize specific CD8+ T-cell populations in lesional skin using single-cell sequencing and marker-based identification.
Clinical interpretation
A change in a cellular population should not automatically be equated with symptomatic improvement, disease control, or another clinical outcome.
Precision
The small enrollment means that uncertainty around any between-group estimate is likely to be important when judging the strength of the evidence.
21. Registry-Based Summary of the Primary Evidence
| Domain | What the registry reports | Statistical implication |
|---|---|---|
| Population | Psoriasis | Defines the population to which the randomized comparison applies. |
| Design | Randomized, parallel, triple-masked, phase 2 | Provides a randomized framework for comparing the two dose groups. |
| Sample size | 20 | Small sample size limits precision and power for between-group inference. |
| Primary endpoint 1 | CD8+ Trm1 cells at baseline and Week 52 | Requires longitudinal analysis of participant-level cellular measurements. |
| Primary endpoint 2 | CD8+ Trm17 cells at baseline and Week 52 | Requires a second longitudinal analysis and consideration of multiple primary endpoints. |
| Measurement method | Longitudinal single-cell RNA-sequencing of full-thickness skin biopsy samples | Creates a hierarchical cellular dataset nested within participants. |
| Cell visualization | UMAP | Useful for visualizing T-cell subtype structure but not itself a treatment-effect test. |
| Cell identification | IFNγ+/CD8+/CD69+ T cells | Defines the cellular populations used for the registered endpoints. |
22. Sources
- ClinicalTrials.gov: NCT05283135 — KNOCKOUT: High Dose Risankizumab for Psoriasis.
Continue through the Clinical Biostats knowledge graph
Explore the broader statistical concepts behind randomized trials, longitudinal measurements, small-sample inference, and biological endpoint analysis.
23. Record Summary
KNOCKOUT is a small randomized phase 2 study in psoriasis evaluating risankizumab across 2 parallel dose groups. Its statistical interest comes from the combination of randomized allocation and highly specialized longitudinal biological endpoints: lesional CD8+ Trm1 and CD8+ Trm17 cells measured at baseline and Week 52 through single-cell RNA-sequencing of full-thickness skin biopsy samples.
The central statistical comparison is the difference in longitudinal change between dose groups. That analysis must respect the repeated-measures structure within participants and the hierarchical structure created by measuring many cells within individual participants. UMAP provides visualization of T-cell subtype structure, while the specified IFNγ+/CD8+/CD69+ markers define the registered cellular populations.
The registry reports that results have been posted, but no formal statistical analyses or numerical estimates for the two primary endpoints are contained in the available record information. The safety summary reports 0/10 serious-adverse-event-affected participants in the 600 mg Dose Group and 1/10 in the 300 mg Dose Group. Given the small sample size, uncertainty is central to interpreting both efficacy and safety observations.