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
JULIET was a phase 2, single-group study evaluating tisagenlecleucel in adults with diffuse large B-cell lymphoma. The registry identifies overall response rate in the Main cohort, as determined by an Independent Review Committee, as the primary endpoint.
| Feature | JULIET |
|---|---|
| Study acronym | JULIET |
| Phase | Phase 2 |
| Condition | Diffuse Large B-cell Lymphoma (DLBCL) |
| Design | Single-group |
| Allocation | NA |
| Masking | None |
| Primary purpose | Treatment |
| Enrollment | 115 |
| Number of arms | 1 |
| Intervention | Tisagenlecleucel; lymphodepleting chemotherapy |
| Status | Completed |
| Start date | 2015-07-29 |
| Primary completion date | 2022-12-22 |
| Lead sponsor | Novartis Pharmaceuticals |
| Sponsor type | Industry |
| ClinicalTrials.gov | NCT02445248 |
2. Clinical Question
The central clinical question was whether treatment with tisagenlecleucel, following lymphodepleting chemotherapy, produced a measurable overall response rate in adults with diffuse large B-cell lymphoma.
Population
Adults with diffuse large B-cell lymphoma (DLBCL) enrolled in the JULIET study.
Intervention
Tisagenlecleucel, with lymphodepleting chemotherapy identified as an associated drug intervention.
Comparator
There was no concurrent randomized comparator arm. The registry identifies a single-group study with one arm.
Primary question
What proportion of participants in the Main cohort achieved a best overall disease response of complete response or partial response according to Independent Review Committee assessment?
3. Trial Design
Tisagenlecleucel
- Tisagenlecleucel is the biological intervention identified in the registry.
- Lymphodepleting chemotherapy is identified as an associated drug intervention.
- The study contains one arm and does not include a randomized comparator.
Response assessment
- The primary endpoint is overall response rate in the Main cohort.
- Response is determined by an Independent Review Committee.
- The response framework includes complete response and partial response.
4. Trial Population and Analysis Structure
The registry reports 115 enrolled participants and a single treatment arm. Because there is no randomized comparator, the statistical structure differs fundamentally from a two-arm randomized trial. There is no between-arm treatment effect to estimate within JULIET.
| Feature | Registry description | Statistical implication |
|---|---|---|
| Enrollment | 115 | The enrolled cohort provides the population from which the observed response proportion and safety experience are summarized. |
| Number of arms | 1 | No randomized treatment-versus-control contrast is available. |
| Allocation | NA | Randomization-based balance and treatment-effect estimation do not apply. |
| Masking | None | The study is open-label rather than blinded. |
| Primary endpoint cohort | Main cohort | The primary endpoint is specifically defined for the Main cohort. |
| Response reviewer | Independent Review Committee | Independent review provides a standardized assessment framework for the response endpoint. |
The registry does not report baseline characteristics in the information summarized here. Accordingly, numerical comparisons of age, disease burden, performance status, or other baseline variables are not part of this analysis.
5. Primary Endpoint
| Endpoint | Registry definition | Time frame |
|---|---|---|
| Overall Response Rate (ORR) Per Independent Review Committee (IRC) in Main Cohort | ORR includes complete response (CR) and partial response (PR) in the Main cohort as determined by IRC assessment. ORR is the percentage of participants with a best overall disease response of CR or PR, where the best overall disease response is defined as the best disease response recorded from CTL019 infusion until progressive disease or start of new anticancer therapy, including ASCT, whichever comes first. Response was assessed according to Evaluation Criteria in diffuse large B cell lymphoma studies, based on Cheson Response criteria and the Lugano Classification (2014). | 60 months |
What exactly is being measured?
The endpoint is a proportion, not a time-to-event endpoint. A participant contributes to the numerator when the best overall disease response recorded during the specified assessment period is either complete response or partial response. The denominator is the relevant Main cohort used for the endpoint.
The registry describes ORR as the percentage of participants whose best overall disease response is complete response or partial response. The precise denominator used for a posted analysis would need to follow the registry's prespecified analysis population.
The response window is important. This is not simply the percentage of participants who were responding at one arbitrary visit. The registry defines the best overall response from CTL019 infusion until progressive disease or initiation of new anticancer therapy, including ASCT, whichever occurs first.
6. Planned Analysis
For a single-group response endpoint such as this, the primary descriptive analysis would normally report the observed proportion of responders together with an appropriate confidence interval for a binomial proportion. The confidence interval quantifies statistical uncertainty around the observed response rate; it does not quantify the variability of individual patients' responses.
A typical confirmatory single-arm framework may specify a null response rate against which the observed ORR is tested. Such a hypothesis would take the general form:
Here p represents the underlying response probability and p0 represents a prespecified reference response rate. A valid formal test requires that reference value and the associated analysis specifications to be explicitly defined.
Because JULIET has no randomized comparator, the interpretation of ORR is fundamentally different from a randomized treatment comparison. An observed response proportion describes activity within the treated study population; it does not by itself estimate how much better tisagenlecleucel performed than an untreated or alternative-treatment population.
7. Statistical Methodology
Binomial response analysis
ORR is a binary participant-level endpoint: each participant either meets the response definition or does not. The resulting number of responders can therefore be modeled as a binomial outcome, with the response probability represented by p.
where X is the number of participants meeting the ORR definition, n is the relevant analysis population, and p is the underlying probability of response.
For an educational analysis of a single-group ORR, an exact binomial confidence interval is often preferable when sample sizes or response counts make normal approximations unreliable. The Wilson interval is another commonly used option. The appropriate method should follow the prespecified statistical analysis plan when one is available.
Independent Review Committee assessment
The registry assigns determination of the primary ORR endpoint to an Independent Review Committee. Independent review is important because tumor response involves clinical assessment and radiologic criteria that can otherwise be affected by knowledge of treatment assignment or other sources of assessment variability.
Best overall response
The endpoint uses the participant's best overall disease response over a defined period rather than a single response assessment. This distinction matters statistically: repeated assessments generate multiple opportunities for a participant to meet a response category, while the endpoint ultimately reduces those observations to the participant's best qualifying response under the prespecified rules.
Response criteria
The registry states that response was assessed according to evaluation criteria in diffuse large B-cell lymphoma studies, based on Cheson Response criteria and the Lugano Classification (2014). These criteria establish the clinical rules used to classify response rather than constituting a statistical test themselves.
Analysis cutoff created by subsequent therapy
The response definition stops the relevant assessment window at progressive disease or the start of new anticancer therapy, including ASCT, whichever comes first. This creates an explicit endpoint rule for situations in which treatment after the study intervention could otherwise complicate interpretation of subsequent disease assessments.
8. Statistical Methods Explained
Why is ORR analyzed as a proportion?
ORR classifies each participant according to whether the participant achieved a best overall response of CR or PR. Once those participant-level classifications are established, the primary summary is naturally expressed as a percentage of responders.
Why does the single-group design matter?
In a randomized trial, the difference between treatment groups can be attributed to the randomized treatment assignment, subject to the usual assumptions. JULIET has one arm, so there is no internal randomized comparison. Its ORR therefore describes response observed after treatment rather than a randomized treatment effect.
Why use an Independent Review Committee?
Independent review separates response determination from the treating investigator's assessment. This can provide a more standardized endpoint assessment and reduces the extent to which knowledge of treatment could influence classification of disease response.
What does a confidence interval for ORR mean?
A confidence interval around a response proportion describes statistical uncertainty about the underlying response probability under the chosen binomial framework. A narrower interval generally indicates greater statistical precision, while a wider interval indicates greater uncertainty. It does not mean that individual patients' response probabilities lie inside the interval.
Why does the response window matter?
The registry does not define ORR simply as response at a single fixed visit. Instead, the best response is recorded from CTL019 infusion until progressive disease or initiation of new anticancer therapy, including ASCT, whichever occurs first. The timing rules therefore directly affect which observations can contribute to the endpoint.
What would a p-value mean in a single-arm ORR analysis?
A p-value would answer a specific hypothesis-testing question about the underlying response probability relative to a prespecified null value. It would not measure the magnitude of the response rate and would not establish how tisagenlecleucel compares with another treatment unless a valid comparative design and analysis were available.
9. Interpreting a Single-Arm Response Endpoint
An observed ORR summarizes the proportion of participants whose best overall response met the registry definition of complete response or partial response. It is therefore a direct measure of observed antitumor activity within the study population.
ORR alone does not establish a randomized causal treatment effect. Without a concurrent randomized comparator, a response proportion cannot be interpreted as the percentage of patients who responded because of treatment relative to what would have happened under another treatment.
Even with 115 enrolled participants, the observed proportion is an estimate of an underlying response probability rather than the probability itself. The confidence interval is therefore an important companion to the point estimate whenever formal response results are available.
A participant can satisfy an ORR definition without establishing how long the response lasts. ORR answers a response question; duration of response and time-to-event endpoints answer different questions. The registry information summarized here does not provide numerical duration-of-response results.
10. Safety Results
The registry reports serious adverse events for the single tisagenlecleucel arm as 84/115 affected participants out of 115 at risk.
| Safety measure | Tisagenlecleucel — All Patients |
|---|---|
| Serious adverse events | 84/115 |
Serious adverse events
Tisagenlecleucel — All Patients
The registry identifies 84 affected participants among 115 at risk.
Because JULIET has a single treatment arm, the serious-adverse-event figure is a description of the observed safety experience. It is not a relative risk or risk ratio comparing tisagenlecleucel with another treatment.
11. ClinicalTrials.gov Record and Study Timeline
Study start
The JULIET study began on 2015-07-29.
Single-group treatment study
The registry identifies JULIET as a phase 2, single-group study with one arm evaluating tisagenlecleucel in adults with DLBCL.
Primary response assessment framework
The primary endpoint is overall response rate per Independent Review Committee in the Main cohort, with a registry-defined time frame of 60 months.
Primary completion
The registry lists 2022-12-22 as the primary completion date.
12. Limitations and Interpretation Issues
- No randomized comparator: the single-group design prevents a within-trial randomized comparison with another treatment.
- Observed response versus causal effect: an ORR describes response among treated participants but does not by itself isolate the treatment effect from the natural history of disease or other factors.
- Endpoint definition: the response classification depends on the prespecified rules for best overall response, including the stopping point created by progressive disease or initiation of new anticancer therapy.
- Independent assessment: IRC assessment strengthens standardization of the response endpoint, but it does not convert a single-arm study into a randomized comparison.
- Precision: response proportions are estimates and should be interpreted with their confidence intervals when those intervals are reported.
- Safety interpretation: the reported serious-adverse-event count describes the single study arm and cannot establish comparative safety against another therapy.
- Missing numerical efficacy analysis: the registry information summarized here does not provide a posted statistical-analysis result for ORR, including a numerical estimate, confidence interval, or p-value.
- Population characterization: the registry information summarized here does not provide a numerical baseline-characteristics table, so detailed assessment of baseline representativeness is not possible from this record alone.
13. Why This Trial Matters Statistically
JULIET is a useful teaching example because it illustrates a statistical problem that differs sharply from the randomized survival analyses commonly used in oncology trials. The central endpoint is a single-arm response proportion, assessed through an independent review process and defined using disease-specific response criteria.
| Concept | How it appears in JULIET |
|---|---|
| Single-arm design | One treatment arm with 115 enrolled participants. |
| Binary efficacy endpoint | Participants are classified according to whether their best overall response is CR or PR. |
| Overall response rate | The primary endpoint is the percentage of Main-cohort participants meeting the CR or PR definition. |
| Independent review | The primary response endpoint is determined by an Independent Review Committee. |
| Best overall response | The endpoint uses the best disease response recorded during the specified assessment window. |
| Response criteria | Assessment is based on Cheson Response criteria and the Lugano Classification (2014). |
| Binomial inference | A response proportion can be analyzed using binomial methods and an appropriate confidence interval. |
| Reference-rate testing | A single-arm study can use a prespecified null response rate for formal hypothesis testing when such a value is specified by the protocol. |
| Safety description | Serious adverse events are reported as affected participants among those at risk in the single arm. |
| Interpretation of causality | Without randomization, an observed response rate does not by itself establish a comparative treatment effect. |
14. Statistical Interpretation vs Clinical Interpretation
Statistical interpretation
The primary endpoint is a binomial response proportion in a single treatment group. The principal inferential questions concern the observed response probability, its statistical precision, and, if prespecified, whether it exceeds a reference response rate.
Clinical interpretation
The clinical meaning of the endpoint depends on the magnitude and durability of tumor response, the population treated, the response-assessment criteria, and the safety experience. ORR alone does not summarize every dimension of treatment benefit.
15. What a Formal ORR Analysis Would Report
For a completed single-arm study with a response endpoint, a statistically complete presentation would normally put several pieces of information together rather than presenting a percentage in isolation.
Point estimate
The observed percentage of participants whose best overall response was CR or PR.
Confidence interval
An interval around the observed response proportion quantifying sampling uncertainty under the selected binomial method.
Analysis population
The exact Main-cohort denominator used to calculate ORR, including any prespecified rules governing evaluability.
Hypothesis test
If specified, a test comparing the response probability with a prespecified null or reference response rate.
The distinction between these components is important. A point estimate communicates the observed response level, a confidence interval communicates precision, and a p-value—if used—addresses a specific null hypothesis. None of these quantities alone describes individual patient outcomes.
16. Primary Endpoint Interpretation
The endpoint ultimately assigns each participant a response classification based on the best overall disease response during the specified window. This makes ORR straightforward to summarize but dependent on the rules used to determine response and the timing of assessments.
Independent Review Committee assessment can improve consistency and reduce assessment bias, but it does not provide the counterfactual information that randomization supplies. The study remains a single-group design.
If a protocol specifies a historical or reference response rate for hypothesis testing, that value is not equivalent to observing a concurrent control group under the same study conditions. Differences in patient selection, assessment, supportive care, and calendar time can affect historical comparisons.
The serious-adverse-event result of 84/115 describes observed events among JULIET participants. It should be interpreted independently from ORR rather than combined into a single numerical measure of overall treatment value.
17. Sources
- ClinicalTrials.gov: NCT02445248 — JULIET.
- PubMed: PubMed record 41252666.
- PubMed: PubMed record 35337365.
- PubMed: PubMed record 35167655.
- PubMed: PubMed record 34516954.
- PubMed: PubMed record 34515338.
Continue through the Clinical Biostats knowledge graph
Explore statistical methods, clinical-trial designs, and quantitative analysis tools across the Clinical Biostats library.
18. Record Summary
JULIET is a phase 2, single-group study of tisagenlecleucel in adults with diffuse large B-cell lymphoma, with 115 participants enrolled and one treatment arm. Its primary endpoint is overall response rate in the Main cohort, determined by an Independent Review Committee over a 60-month time frame. The endpoint defines response using complete response or partial response as the best overall disease response from CTL019 infusion until progressive disease or initiation of new anticancer therapy, including ASCT, whichever comes first.
Statistically, the most important feature of JULIET is the distinction between estimating response within a treated cohort and estimating a randomized treatment effect. The former is the natural framework for a single-arm ORR endpoint; the latter requires a valid comparator. The registry reports serious adverse events in 84 of 115 patients in the tisagenlecleucel arm, providing a separate description of the observed safety experience.