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
TRANSCEND NHL 001 was a phase 1, non-randomized, parallel-design study evaluating two JCAR017 (lisocabtagene maraleucel) dosing schedules in patients with B-cell non-Hodgkin lymphoma and specified lymphoma subtypes. The registered primary endpoints address treatment-related adverse events, dose-limiting toxicities, and objective response rate.
| Feature | TRANSCEND NHL 001 |
|---|---|
| ClinicalTrials.gov identifier | NCT02631044 |
| Phase | Phase 1 |
| Status | Completed |
| Allocation | Non-randomized |
| Design model | Parallel |
| Masking | None |
| Primary purpose | Treatment |
| Enrollment | 387 |
| Number of arms | 2 |
| Lead sponsor | Juno Therapeutics, a Subsidiary of Celgene |
| Sponsor type | Industry |
2. Clinical Question
The central clinical-statistical question was how JCAR017 performs when administered according to either a single-dose schedule or a 2-dose schedule in patients with specified forms of B-cell non-Hodgkin lymphoma. Because allocation was non-randomized and there was no masking, the design is primarily suited to characterizing safety, dose-limiting toxicity, and response rather than establishing a randomized comparative treatment effect.
Population
Patients enrolled with non-Hodgkin lymphoma, including diffuse large B-cell lymphoma, follicular lymphoma, mantle-cell lymphoma, and primary mediastinal B-cell lymphoma.
Intervention
JCAR017 (lisocabtagene maraleucel), administered using either a single-dose schedule or a 2-dose schedule.
Comparator
There was no randomized control group. The two study arms represented different JCAR017 dosing schedules.
Primary question
What are the treatment-related adverse events and dose-limiting toxicities associated with JCAR017, and what objective response rate is observed over 24 months?
3. Trial Design
JCAR017 single-dose schedule
- JCAR017 (lisocabtagene maraleucel)
- Single-dose schedule
- Biological intervention
JCAR017 2-dose schedule
- JCAR017 (lisocabtagene maraleucel)
- 2-dose schedule
- Biological intervention
The non-randomized allocation is a central statistical feature of this study. Without random assignment, differences observed between the two dosing schedules can reflect differences in the populations receiving each schedule as well as differences associated with the schedule itself. Consequently, comparisons between arms should be interpreted descriptively unless a separate analysis explicitly addresses confounding.
Design characteristics and their statistical implications
| Design feature | Registry description | Statistical implication |
|---|---|---|
| Allocation | Non-randomized | Between-arm comparisons do not receive the protection against baseline confounding provided by randomization. |
| Design model | Parallel | Participants are evaluated within their assigned study schedule rather than through a crossover sequence. |
| Masking | None | Knowledge of treatment schedule can affect assessment and reporting, particularly for subjective outcomes. |
| Primary purpose | Treatment | The study evaluates a therapeutic intervention in the specified lymphoma population. |
4. Study Population and Conditions
The registry identifies the study conditions as non-Hodgkin lymphoma, diffuse large B-cell lymphoma, follicular lymphoma, mantle-cell lymphoma, and primary mediastinal B-cell lymphoma.
| Condition | Registry-listed condition |
|---|---|
| Non-Hodgkin lymphoma | Non-Hodgkin Lymphoma |
| Diffuse large B-cell lymphoma | Diffuse Large B Cell Lymphoma |
| Follicular lymphoma | Follicular Lymphoma |
| Mantle-cell lymphoma | Mantle-cell Lymphoma |
| Primary mediastinal B-cell lymphoma | Primary Mediastinal B-cell Lymphoma |
The registry enrollment is 387. No additional baseline demographic or disease-characteristic table is included here because those characteristics are not reported in the registry information summarized for this analysis.
5. Primary Endpoints
| Endpoint | Time frame | Registry description |
|---|---|---|
| Treatment-related adverse events (AEs) as assessed by CTCAE v4.03 | Up to 730 days after the final JCAR017 infusion | Physiological parameter |
| Dose-limiting toxicities of JCAR017 | 28 days after first (single-dose schedule) or second (2-dose schedule) JCAR017 infusion | Physiological parameter |
| Objective response rate (ORR) | 24 months | Lugano criteria |
Why the endpoint time frames matter
The three primary endpoints operate on different observation windows. Treatment-related adverse events are assessed through 730 days after the final infusion, whereas dose-limiting toxicity is focused on the first 28 days after the relevant infusion. ORR is evaluated over 24 months using Lugano criteria.
This distinction is statistically important because the denominators and opportunity for observation can differ by endpoint. A toxicity observed during a defined 28-day DLT window is not equivalent to an adverse event observed during a substantially longer follow-up period. Likewise, ORR over 24 months describes a response endpoint rather than a time-to-death endpoint.
6. Planned Analysis
The registry does not contain posted statistical analyses for the primary endpoints. The study therefore has to be understood from the endpoints it registered and the statistical structure normally appropriate for those endpoint types.
Treatment-related adverse events
AEs assessed using CTCAE v4.03 would ordinarily be summarized by treatment schedule, event type, severity or grade, and relevant observation period. Incidence proportions and event counts are natural descriptive summaries for a safety endpoint of this type.
Dose-limiting toxicities
DLT is a binary or categorical safety outcome within a defined 28-day assessment window. A typical analysis would summarize the number and proportion of participants experiencing a DLT in each schedule, with an appropriate confidence interval for the proportion.
Objective response rate
ORR is a proportion-based endpoint. Under Lugano criteria, a typical analysis would classify participants according to the prespecified response definitions and report the proportion achieving an objective response, together with a confidence interval.
Between-arm interpretation
Because the registry describes allocation as non-randomized, a simple difference between the two schedules would be descriptive rather than a randomized estimate of causal treatment effect.
7. Statistical Methodology
Safety-event analysis
Treatment-related adverse events are generally analyzed as categorical safety outcomes. For each treatment schedule, investigators can count the number of participants experiencing an event and calculate the corresponding proportion among the relevant safety population. CTCAE v4.03 provides a standardized framework for classifying adverse-event severity.
For a safety endpoint, the denominator is crucial. It should correspond to the population actually eligible for the relevant safety assessment rather than automatically assuming that the overall enrollment is the denominator for every analysis.
Dose-limiting toxicity
The DLT endpoint has a tightly defined assessment window: 28 days after the first JCAR017 infusion for the single-dose schedule or after the second JCAR017 infusion for the 2-dose schedule. This makes DLT a particularly focused early safety endpoint.
For a binary DLT endpoint, the primary descriptive quantity is a proportion. An exact binomial confidence interval or another prespecified binomial interval can quantify uncertainty around that proportion. In a phase 1 setting, the size of that interval can be as informative as the point estimate because small numbers of observed toxicities can produce substantial statistical uncertainty.
Objective response rate
ORR is defined in the registry using Lugano criteria and a 24-month time frame. The statistical unit is ordinarily the participant classified according to whether the prespecified response criteria are met.
The corresponding confidence interval quantifies uncertainty around the estimated response proportion. It does not describe the probability that a particular individual will respond.
Why survival analysis is not automatically implied by ORR
Although ORR has a 24-month time frame, the registry defines it as an objective response rate rather than a time-to-event endpoint. A time frame attached to a proportion does not by itself make the endpoint a survival endpoint. Kaplan-Meier methods would be appropriate for a separately defined time-to-response, duration-of-response, progression-free survival, or overall-survival endpoint, but those endpoints are not among the three registered primary endpoints listed here.
8. Non-Randomized Allocation and Comparative Interpretation
Randomization is one of the principal mechanisms used in clinical trials to balance both measured and unmeasured baseline characteristics in expectation. TRANSCEND NHL 001 is explicitly described as non-randomized, so that protection is absent.
The key statistical distinction
A difference in an outcome between the single-dose and 2-dose schedules can be observed and described, but the difference cannot automatically be attributed to the dosing schedule itself.
Suppose one schedule had a higher response rate than the other. A descriptive comparison would establish that the observed response proportions differed. It would not, by itself, establish that changing from one schedule to the other would cause the same difference in a comparable population.
The same issue applies to safety. If one arm has a higher observed adverse-event incidence, that difference can be clinically relevant, but the interpretation must account for the non-randomized allocation and for differences in who received each schedule.
9. Statistical Methods Explained
Why is this a phase 1 statistical problem?
Phase 1 studies commonly emphasize safety, tolerability, and identification of dose-related toxicities rather than definitive comparative efficacy. Here, the registered primary endpoints are treatment-related adverse events, dose-limiting toxicities, and ORR. That endpoint structure places safety characterization at the center of the statistical analysis.
Why is DLT assessed over 28 days?
The registry defines DLT using a fixed 28-day window after the relevant JCAR017 infusion. A fixed window creates a consistent period during which early dose-limiting toxicity can be assessed across participants. The statistical analysis should therefore respect the protocol-defined window rather than mixing early DLT with later adverse events.
Why use proportions for ORR?
ORR describes the fraction of participants who meet the prespecified objective-response definition. A proportion directly answers that question. A confidence interval is then used to communicate uncertainty around the estimated proportion.
Does a 24-month ORR endpoint mean that every patient must be observed for 24 months?
Not necessarily. The phrase "24 months" defines the registered time frame for the endpoint, but the precise handling of participants with incomplete observation, early discontinuation, or other assessment issues would require the detailed statistical analysis plan. The registry information does not provide those rules.
Why is non-randomization important when comparing the two schedules?
With non-randomized allocation, participants receiving the two schedules can differ systematically in characteristics that also affect safety or response. Consequently, an unadjusted difference between schedules can combine the effect of treatment schedule with differences in the populations receiving those schedules.
What would a confidence interval add to an ORR estimate?
The point estimate describes the observed response proportion, while the confidence interval describes statistical uncertainty around that estimate under the chosen inferential framework. A narrow interval indicates greater precision than a wide interval, but neither interval width nor the point estimate establishes causality in a non-randomized comparison.
Would a p-value establish that one JCAR017 schedule is better?
No. Even if a formal comparison produced a small p-value, the interpretation would still depend on the non-randomized design, the analysis population, the estimand, and the assumptions of the statistical model. Statistical significance does not remove confounding created by non-randomized allocation.
10. Safety Analysis Framework
The first registered primary endpoint is treatment-related adverse events assessed using CTCAE v4.03 for up to 730 days after the final JCAR017 infusion. This is a longitudinal safety endpoint with a substantially longer observation period than the DLT endpoint.
| Safety feature | Registered specification | Statistical consideration |
|---|---|---|
| Endpoint | Treatment-related adverse events | Participant-level incidence and event-level summaries can be informative. |
| Classification system | CTCAE v4.03 | Provides standardized adverse-event assessment. |
| Observation period | Up to 730 days after the final JCAR017 infusion | Late events can be captured beyond the early DLT period. |
| Allocation | Non-randomized | Between-schedule safety differences are subject to confounding. |
Safety interpretation should distinguish at least three concepts: whether an event occurred, how severe it was, and when it occurred. A single cumulative percentage can conceal the timing and severity structure of adverse events. The registry endpoint specifically identifies CTCAE v4.03, so severity classification is an important part of the intended assessment.
Why the 730-day window changes the interpretation
The 730-day window extends considerably beyond the 28-day DLT period. An event occurring during the early DLT window and an event occurring much later can both contribute to the broader treatment-related AE endpoint, but they answer different clinical questions. Early DLT focuses on acute dose-limiting toxicity; longer follow-up captures a broader safety experience.
11. Dose-Limiting Toxicity Analysis
The second primary endpoint is dose-limiting toxicities of JCAR017. The registry specifies separate timing according to dosing schedule: 28 days after the first infusion for the single-dose schedule and 28 days after the second infusion for the 2-dose schedule.
The different reference points are important. A direct comparison of crude DLT percentages can be misleading if the two schedules have different opportunities for exposure or different clinical characteristics among participants entering each schedule.
For a phase 1 DLT endpoint, the statistical emphasis is usually on the observed toxicity frequency and its uncertainty rather than on a conventional superiority hypothesis test. A small number of DLT events can result in a wide confidence interval, so the observed proportion should not be interpreted without considering its denominator and precision.
12. Objective Response Rate
The third registered primary endpoint is objective response rate over 24 months using Lugano criteria. Unlike a time-to-event endpoint, ORR is fundamentally a response classification endpoint.
What ORR measures
ORR summarizes the proportion of participants meeting the registry-defined objective-response criteria during the specified 24-month time frame.
An ORR analysis normally requires a clearly defined response-evaluable population and rules for participants who do not have an adequate post-baseline assessment. Those details can materially affect the denominator. Because the registry information summarized here does not specify those analysis-population rules, they should not be inferred from the enrollment count of 387.
ORR is not the same as durability
A response rate answers whether a participant achieved an objective response. It does not by itself describe how long the response lasted. Duration of response would require a separately defined time-to-event endpoint and appropriate censoring rules. No duration-of-response endpoint is listed among the three registered primary endpoints here.
13. Statistical Interpretation of a Non-Randomized Two-Arm Study
The two JCAR017 schedules provide an opportunity to describe safety and response patterns across the study arms. Because the allocation is non-randomized, however, an observed numerical difference between schedules is not equivalent to a treatment effect estimated from a randomized controlled comparison.
Confounding occurs when characteristics associated with the choice or receipt of a schedule are also associated with the outcome. Without randomization, observed schedule differences can therefore reflect both the schedule and differences in participant characteristics.
A confidence interval can describe statistical precision around an estimate, but a very precise estimate can still be biased if the study design does not support a causal comparison. Precision and validity are separate statistical concepts.
A p-value, when calculated, would address evidence against a specified null hypothesis. It would not measure the magnitude or clinical importance of a response-rate or safety difference, and it would not by itself resolve confounding from non-randomized allocation.
14. Missing Data and Analysis Populations
Endpoint-specific missing-data rules are important in this study because the primary endpoints have different observation periods and different outcome structures.
| Endpoint | Potential analysis issue | Why it matters |
|---|---|---|
| Treatment-related AEs | Incomplete safety follow-up | Shorter observation can reduce the opportunity to observe later events within the 730-day period. |
| DLT | Incomplete 28-day assessment | The DLT denominator should reflect the prespecified assessment population and rules for incomplete observation. |
| ORR | Missing or inadequate response assessments | The response-evaluable denominator and handling of missing assessments can materially affect the estimated response rate. |
The registry information does not specify a missing-data or imputation method for these endpoints. In particular, an ORR analysis should not automatically assume that missing response assessments are equivalent to response, non-response, or censoring without a prespecified analysis rule.
15. Multiplicity and Hypothesis Testing
The registry identifies three primary endpoints: treatment-related adverse events, dose-limiting toxicities, and ORR. These endpoints address different aspects of the study rather than representing repeated measurements of one identical clinical outcome.
| Endpoint | Primary role | Statistical character |
|---|---|---|
| Treatment-related adverse events | Primary safety endpoint | Safety-event incidence over up to 730 days |
| Dose-limiting toxicities | Primary early safety endpoint | Binary or categorical toxicity outcome in a 28-day window |
| Objective response rate | Primary efficacy endpoint | Proportion meeting Lugano response criteria over 24 months |
Because the endpoints address different questions, their statistical interpretation should remain separate. A favorable safety finding does not establish efficacy, and a favorable ORR does not establish a favorable long-term safety profile.
The registry information does not specify an alpha-allocation strategy, hierarchical testing sequence, multiplicity adjustment, or interim-analysis procedure for these endpoints. Those features should therefore not be assumed.
16. Interim Analysis and Bayesian Methods
The registry information summarized here does not report an interim-analysis procedure, alpha-spending strategy, Bayesian model, or Bayesian decision rule.
This distinction matters because phase 1 studies can use a variety of adaptive or model-based approaches in other settings, including Bayesian dose-escalation frameworks. The absence of such a method in the registered information means that a Bayesian model or adaptive stopping rule should not be attributed to TRANSCEND NHL 001 without explicit documentation.
17. Crossover, Factorial Design, and Treatment Switching
The study is registered as a parallel design with two JCAR017 schedules. The registry information does not describe a crossover design or factorial design.
Parallel design
The study's design model is parallel. The two arms represent separate JCAR017 dosing schedules.
No crossover analysis
The registry information does not identify crossover as a design feature, so crossover-specific statistical methods are not part of the analysis presented here.
No factorial structure
The registry describes two JCAR017 schedules rather than a factorial combination of independent treatment factors.
No treatment-effect assumption
Because allocation is non-randomized, schedule comparisons require more caution than a randomized two-arm comparison.
18. Study Timeline
Trial begins
The registry lists January 6, 2016 as the study start date.
Two JCAR017 schedules
The completed study includes two arms evaluating JCAR017 as a single-dose schedule or a 2-dose schedule.
Safety and response assessment
The primary endpoints cover treatment-related adverse events, dose-limiting toxicities, and objective response rate.
Primary completion
The registry lists May 16, 2024 as the primary completion date.
19. What a Proper Analysis Would Report
A statistically complete presentation of the primary endpoints would separate the analysis into safety and efficacy components rather than reducing the study to a single summary number.
| Endpoint | Core estimate | Precision / context |
|---|---|---|
| Treatment-related adverse events | Incidence by event and relevant CTCAE v4.03 classification | Observation through 730 days after the final infusion |
| Dose-limiting toxicities | DLT proportion | 28-day schedule-specific assessment window |
| Objective response rate | ORR proportion | Lugano criteria over 24 months |
For the safety endpoints, event counts and proportions are central. For ORR, the response proportion and its confidence interval are central. If investigators make formal between-schedule comparisons, the analysis should explicitly address the non-randomized allocation rather than treating the two arms as if they arose from randomized treatment assignment.
20. Important Limitations and Interpretation Issues
- Non-randomized allocation: the study does not use randomization, so differences between the two dosing schedules can be confounded by differences in the participants receiving each schedule.
- No masking: the registry describes the study as unmasked. Knowledge of treatment schedule can affect reporting or assessment of some outcomes.
- Phase 1 design: the study's primary endpoint structure emphasizes safety, dose-limiting toxicity, and objective response rather than a randomized confirmatory comparison.
- Different endpoint windows: DLT is assessed within 28 days of the relevant infusion, treatment-related AEs are followed for up to 730 days after the final infusion, and ORR uses a 24-month time frame.
- Missing-data rules: the registry information does not specify the detailed rules for incomplete response assessments or incomplete endpoint follow-up.
- Analysis populations: the enrollment count of 387 should not automatically be treated as the denominator for every primary endpoint without the endpoint-specific analysis population.
- Unreported inferential framework: the registry information does not specify an alpha-spending procedure, multiplicity adjustment, Bayesian model, or interim-analysis method.
- No posted formal statistical analyses: ClinicalTrials.gov does not contain posted formal statistical analyses for the registered primary endpoints.
21. Why This Trial Matters Statistically
TRANSCEND NHL 001 is a useful teaching example because it illustrates a statistical situation that differs fundamentally from the familiar randomized phase 3 trial. The study has two treatment schedules, but the registry describes allocation as non-randomized. That single design feature changes how nearly every between-arm comparison should be interpreted.
| Concept | How it appears in TRANSCEND NHL 001 |
|---|---|
| Phase 1 design | Primary emphasis on safety, dose-limiting toxicity, and response. |
| Non-randomized allocation | Two JCAR017 schedules without randomized assignment. |
| Parallel design | Separate single-dose and 2-dose schedules. |
| Safety monitoring | Treatment-related AEs assessed using CTCAE v4.03. |
| Early toxicity endpoint | DLT assessed during a 28-day schedule-specific window. |
| Response endpoint | ORR assessed over 24 months using Lugano criteria. |
| Proportion estimation | Natural framework for DLT and ORR estimates. |
| Confidence intervals | Important for describing uncertainty around safety and response proportions. |
| Confounding | Central issue when interpreting differences between non-randomized schedules. |
| Endpoint-specific denominators | Important because safety, DLT, and ORR have different observation requirements. |
22. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The registered endpoints define three distinct statistical questions: the frequency of treatment-related adverse events, the frequency of dose-limiting toxicities within a defined early window, and the proportion of participants achieving objective response over 24 months.
Clinical interpretation
The study is structured to characterize the safety and activity of JCAR017 across two dosing schedules in specified B-cell non-Hodgkin lymphoma populations. The absence of randomized allocation limits direct causal comparison between schedules.
The distinction is important. A statistical analysis can precisely estimate an observed response proportion without establishing that the corresponding treatment schedule caused that response rate. Similarly, a well-characterized adverse-event rate describes the observed safety experience but does not by itself establish comparative causality.
23. A Practical Analysis Workflow
The most important analytical discipline is to preserve the endpoint definitions. Treatment-related AEs should be analyzed within their 730-day follow-up framework; DLT should use the relevant 28-day window; and ORR should retain the 24-month Lugano-based definition. Combining these endpoints into a single measure would obscure the different questions the study was designed to answer.
24. Sources
- ClinicalTrials.gov: NCT02631044 — TRANSCEND NHL 001.
Continue through the Clinical Biostats clinical trial library
Clinical trial analysis is most useful when design, endpoint definition, statistical method, and interpretation are considered together.
25. Record Summary
TRANSCEND NHL 001 is a phase 1, non-randomized, parallel study of JCAR017 (lisocabtagene maraleucel) using a single-dose schedule or a 2-dose schedule in patients with specified B-cell non-Hodgkin lymphoma conditions. Its three registered primary endpoints address treatment-related adverse events assessed by CTCAE v4.03 for up to 730 days after the final infusion, dose-limiting toxicities assessed during a 28-day schedule-specific window, and objective response rate over 24 months using Lugano criteria.
Statistically, the most important feature is the non-randomized allocation. The study can characterize the observed safety and response experience of the two JCAR017 schedules, but numerical differences between schedules should not automatically be interpreted as randomized treatment effects. Proper interpretation therefore requires attention to endpoint-specific denominators, observation windows, uncertainty around proportions, and the potential for confounding.