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
SIGNATURE was a phase 2, single-group treatment study evaluating ceritinib in patients whose tumors had aberrations in ALK or ROS1. The registry reports an enrollment of 47 participants and two primary response endpoints assessed from baseline to approximately 16 weeks.
| Feature | SIGNATURE |
|---|---|
| Phase | Phase 2 |
| Population | Patients with tumors with aberrations in ALK or ROS1 |
| Design | Single-group, non-masked treatment study |
| Allocation | NA |
| Primary purpose | Treatment |
| Intervention | Ceritinib (drug) |
| Enrollment | 47 |
| Number of arms | 1 |
| Lead sponsor | Novartis Pharmaceuticals |
| Sponsor type | Industry |
| ClinicalTrials.gov | NCT02186821 |
| Status | Terminated |
2. Clinical Question
The central question was whether ceritinib produced measurable tumor responses and clinical benefit in participants whose tumors had aberrations in ALK or ROS1, using tumor-response criteria assessed at approximately 16 weeks.
Population
Patients with tumors having aberrations in ALK or ROS1.
Intervention
Ceritinib, administered as the study drug in the single treatment group.
Comparator
There was no comparator group. The study used a single-group design.
Primary question
What proportion of participants demonstrated overall response or clinical benefit, and what tumor-response categories were observed by approximately 16 weeks?
3. Trial Design
Ceritinib
- One treatment group
- Ceritinib (drug)
- No masking
- Primary purpose: treatment
- Tumor-response assessment through approximately 16 weeks
Single-group framework
- No randomized control group
- No between-arm treatment comparison
- Response endpoints describe outcomes within the treated group
- Interpretation therefore differs from a randomized comparative trial
The single-group structure is fundamental to the statistical interpretation. With one treatment group, the primary response measures describe the observed distribution of tumor responses among treated participants. They do not directly estimate a randomized treatment effect because there is no concurrent control group.
4. Study Timeline
Study start
The registry lists 2014-09-17 as the study start date.
Primary completion
The registry lists 2017-12-13 as the primary completion date.
Terminated
The ClinicalTrials.gov record lists the study status as terminated.
5. Primary Endpoints
| Endpoint | Time frame | Registry definition |
|---|---|---|
| Percentage of Participants Overall Response Rate (ORR) and Clinical Benefit Rate (CBR) Utilizing RECIST 1.1 at Approximately 16 Weeks | Baseline up to approximately 16 weeks | For solid tumors, responses included complete response (CR), partial response (PR), or stable disease (SD) using RECIST 1.1. ORR = CR + PR. Clinical benefit rate = CR + PR + SD. For hematologic tumors, other hematological response criteria applied. |
| Number of Participants With Tumor Responses Utilizing RECIST 1.1 at Approximately 16 Weeks | Baseline up to approximately 16 weeks | For solid tumors, participants were classified according to CR, PR, SD, or progressive disease (PD) using RECIST 1.1. For hematologic tumors, other hematological response criteria applied. |
RECIST 1.1 response definitions
The registry defines complete response as disappearance of all target lesions. Pathological lymph nodes, whether target or non-target, must have reduction in short axis to <10mm.
Partial response requires at least a 30% decrease in the sum of diameters of target lesions from the baseline sum diameters.
Stable disease means neither sufficient shrinkage to qualify for PR nor sufficient increase to qualify for progressive disease, taking as reference the smallest sum diameters while on study.
Progressive disease is defined as at least a 20% increase in the sum of diameter of measured target lesions from the smallest sum of diameter of all target lesions recorded at or after baseline, with an absolute increase of at least 5mm. One or more new lesions is also considered PD.
ORR therefore captures complete and partial responses, whereas CBR additionally counts stable disease according to the registry's definition.
6. Statistical Methodology
Descriptive analysis of tumor response
The primary endpoints are categorical response outcomes rather than time-to-event endpoints. A natural statistical summary is therefore the number and percentage of participants falling into each response category, followed by the derived ORR and CBR proportions.
For a single-group study with 47 enrolled participants, an observed response proportion can be expressed conceptually as:
where x is the number of participants meeting the response definition and n is the relevant analysis population. Confidence intervals for a proportion can then quantify uncertainty around the observed response rate.
Why a single-group analysis is different
In a randomized two-arm trial, a response rate can be compared between treatment groups. SIGNATURE has one treatment group, so the central descriptive quantity is the response proportion within that group. A comparison with an external historical benchmark would require a prespecified reference rate and assumptions about whether the historical population is sufficiently comparable.
RECIST classification
The statistical analysis begins with the clinical classification itself. Participants are first categorized according to the response criteria, after which the categories are aggregated into ORR and CBR. This makes the endpoint definition clinically interpretable while preserving a straightforward binary summary for the percentage endpoints.
Hematologic tumors
The registry specifies that other hematological response criteria applied to hematologic tumors. Consequently, the response framework is not identical across all tumor types: the applicable response criteria depend on whether the tumor is solid or hematologic.
7. Planned Analysis
The registry identifies two primary response endpoints measured from baseline up to approximately 16 weeks. For the percentage endpoint, the appropriate descriptive analysis would summarize the proportion of participants with an overall response and the proportion with clinical benefit, with ORR defined as CR + PR and CBR defined as CR + PR + SD.
The second endpoint would ordinarily be summarized by the number of participants in each tumor-response category. For solid tumors, those categories are CR, PR, SD, and PD under RECIST 1.1. For hematologic tumors, the applicable hematological response criteria would be used.
| Statistical question | Typical analysis for this endpoint |
|---|---|
| How many participants responded? | Count and percentage with CR or PR |
| What is the overall response rate? | ORR = (CR + PR) / relevant analysis population |
| How many participants achieved clinical benefit? | Count and percentage with CR, PR, or SD |
| What is the clinical benefit rate? | CBR = (CR + PR + SD) / relevant analysis population |
| How precise is a response-rate estimate? | A binomial confidence interval for the proportion |
| How are individual responses categorized? | RECIST 1.1 for solid tumors; other hematological response criteria for hematologic tumors |
Results have not been posted on ClinicalTrials.gov in the form of formal statistical analyses for these endpoints. The registry record does, however, identify the response endpoints, their time frame, and the definitions used to classify tumor response.
8. Statistical Methods Explained
Why does ORR equal CR plus PR?
ORR is designed to capture objective tumor shrinkage meeting the prespecified response thresholds. Under the registry definition, complete response and partial response are counted as responses, while stable disease is not included in ORR.
Why does CBR include stable disease?
CBR answers a broader disease-control question. By including SD with CR and PR, it counts participants whose disease neither shrank enough for PR nor increased enough for PD according to the applicable criteria.
Why is the approximately 16-week time frame important?
A response endpoint is inseparable from its assessment window. The registry specifies baseline up to approximately 16 weeks, so the resulting proportion describes tumor-response status during that defined period rather than an unrestricted lifetime probability of response.
Why is RECIST 1.1 used?
RECIST 1.1 converts changes in measurable tumor burden into prespecified categorical outcomes. The rules reduce subjective interpretation by specifying the magnitude of shrinkage required for PR and the magnitude of increase required for PD.
Why is a single-group response rate not the same as a treatment effect?
A response percentage tells us what happened among participants who received the study treatment. Without a concurrent randomized comparator, it does not by itself isolate how much of the observed response rate is attributable to treatment rather than other factors such as the underlying disease course or differences between this study population and another population.
Why would a confidence interval be useful for ORR?
A response rate calculated from a finite number of participants is an estimate rather than an exact population quantity. A confidence interval communicates the statistical uncertainty associated with that estimate. It does not describe the range of responses that an individual participant could experience.
Why must the analysis population be defined?
The denominator determines the meaning of a percentage. If the analysis population differs from the total enrollment, the resulting ORR or CBR can differ even when the number of responders is unchanged. A rigorous statistical report therefore states which participants are included in the denominator.
9. Safety
The registry reports serious adverse events by arm as 16 of 47 participants for the ceritinib 750 mg group.
Serious adverse events
Ceritinib 750 mg: affected / at risk
| Safety measure | Ceritinib 750 mg |
|---|---|
| Participants with serious adverse events | 16/47 |
| Study arms | 1 |
The reported safety measure is an affected-versus-at-risk count. It should be distinguished from an incidence rate over person-time, a severity distribution, or a comparative risk measure. Because there is no comparator arm, the 16/47 figure is a within-study safety description rather than a randomized between-treatment comparison.
10. What a Response Rate Does — and Does Not — Mean
An ORR is a proportion of participants who met the registry's definition of CR or PR during the specified assessment period. It is a categorical response measure, not a hazard ratio or a time-to-event effect estimate.
It does not mean that every responding participant experienced the same magnitude of tumor shrinkage, that the response was permanent, or that the treatment would produce the same response proportion in another population.
RECIST 1.1 uses specific thresholds. PR requires at least a 30% decrease in the sum of diameters of target lesions from baseline, while PD requires at least a 20% increase from the smallest recorded sum after baseline together with an absolute increase of at least 5mm, or the appearance of one or more new lesions. These rules create the categorical endpoint that is subsequently summarized statistically.
CBR includes stable disease in addition to complete and partial responses. Thus, two studies could have the same ORR but different CBRs if the proportion of participants with stable disease differs.
11. Single-Group Design and Causal Interpretation
The absence of a comparator is one of the most important statistical characteristics of SIGNATURE. A single-group phase 2 design can provide useful evidence about whether a treatment produces a measurable response signal in a defined population, but it does not provide the same randomized comparison available from a controlled trial.
What the design estimates directly
The observed distribution of tumor responses among participants treated with ceritinib, including the registry-defined ORR and CBR measures.
What the design does not directly estimate
A randomized difference in response between ceritinib and another treatment because no concurrent comparator group is included.
Why historical comparisons are difficult
Differences in eligibility, disease characteristics, assessment practices, follow-up, and other factors can make an external benchmark an imperfect counterfactual.
Why endpoint standardization helps
RECIST 1.1 provides prespecified rules for classifying tumor changes, improving consistency in how response is defined before the percentages are calculated.
12. Endpoint Construction in Detail
| Category | Definition in the registry | Contribution to ORR | Contribution to CBR |
|---|---|---|---|
| Complete response (CR) | Disappearance of all target lesions; pathological lymph nodes must reduce in short axis to <10mm. | Included | Included |
| Partial response (PR) | At least a 30% decrease in the sum of diameters of target lesions from baseline. | Included | Included |
| Stable disease (SD) | Neither sufficient shrinkage for PR nor sufficient increase for PD, using the smallest sum diameters while on study as the reference. | Not included | Included |
| Progressive disease (PD) | At least a 20% increase in measured target-lesion diameter from the smallest post-baseline sum, with an absolute increase of at least 5mm, or one or more new lesions. | Not included | Not included |
This construction illustrates why the response endpoint is more than a simple measurement of tumor size. A participant's classification depends on the baseline measurement, the smallest sum observed while on study, subsequent changes, and the appearance of new lesions.
13. Statistical Interpretation of the 47-Participant Study
The enrollment of 47 participants provides the population size against which the primary response percentages are defined. In a single-group study, every participant can have a substantial effect on the estimated proportion, particularly when the total sample is relatively small.
Numerator
The numerator for ORR consists of participants classified as CR or PR. The numerator for CBR consists of participants classified as CR, PR, or SD.
Denominator
The statistical denominator should correspond to the defined analysis population. Enrollment alone does not establish that every enrolled participant necessarily belongs to every efficacy-analysis population.
Precision
With 47 enrolled participants, uncertainty around a proportion can be material. Confidence intervals are therefore important complements to point estimates.
External validity
A response proportion observed in this study population should not automatically be treated as the response probability for every patient with an ALK- or ROS1-aberrant tumor.
14. Limitations
- Single-group design: the absence of a concurrent comparator limits direct estimation of a treatment effect.
- Small enrollment: 47 participants means that individual observations can materially affect response proportions and their statistical precision.
- Response endpoint window: the primary assessment is baseline up to approximately 16 weeks, so it should not be interpreted as a lifetime response probability.
- Endpoint dependence on classification: ORR and CBR depend on the prespecified categorization of tumor response.
- Solid versus hematologic tumors: the registry specifies RECIST 1.1 for solid tumors but other hematological response criteria for hematologic tumors, so response assessment is not necessarily identical across tumor types.
- External comparators: comparisons with historical response rates can be affected by differences in patient populations, disease characteristics, treatment history, assessment methods, and other study features.
- Safety interpretation: the reported 16/47 serious-adverse-event count is descriptive and cannot establish comparative safety without a control group.
- Formal inferential results: the registry does not post formal statistical analyses with effect estimates, confidence intervals, or p-values for the primary endpoints.
- Generalizability: the applicability of the observed response distribution to other populations depends on how closely those populations resemble the participants enrolled in the study.
15. Why This Trial Matters Statistically
SIGNATURE is a useful teaching example because it illustrates how statistical interpretation changes when a clinical trial has a single treatment group rather than a randomized comparator. The primary endpoints are response-based, the assessment window is explicitly defined, and the registry provides detailed operational definitions for CR, PR, SD, and PD.
| Concept | How it appears in SIGNATURE |
|---|---|
| Single-group design | One treatment group with 47 enrolled participants. |
| Descriptive proportion | ORR and CBR are percentages based on categorical tumor-response outcomes. |
| RECIST 1.1 | Provides standardized definitions for CR, PR, SD, and PD in solid tumors. |
| Response endpoint | ORR is CR + PR; CBR is CR + PR + SD. |
| Assessment window | Baseline up to approximately 16 weeks. |
| Denominator | Response percentages require a clearly defined analysis population. |
| Confidence intervals | Can quantify uncertainty around a single-group response proportion. |
| Historical comparison | Requires caution because an external population is not equivalent to a concurrent randomized control group. |
| Safety description | Serious adverse events are reported as affected / at risk: 16/47 for ceritinib 750 mg. |
| Heterogeneous response criteria | RECIST 1.1 applies to solid tumors; other hematological criteria apply to hematologic tumors. |
16. Related Statistical Concepts
The statistical structure of SIGNATURE connects naturally to several core concepts in clinical-trial methodology, particularly categorical endpoints, binomial proportions, confidence intervals, and the interpretation of single-arm studies.
17. Sources
- ClinicalTrials.gov: NCT02186821 — Ceritinib (LDK378) for Patients Whose Tumors Have Aberrations in ALK or ROS1 (SIGNATURE).
Continue through the Clinical Biostats knowledge graph
Explore additional clinical-trial statistical methods, educational tutorials, and analysis tools.
18. Record Summary
SIGNATURE is a phase 2, single-group treatment study of ceritinib in patients with tumors having aberrations in ALK or ROS1. Its statistical structure is centered on tumor-response classification rather than a randomized comparison: participants were assessed from baseline up to approximately 16 weeks, with ORR defined as CR + PR and CBR defined as CR + PR + SD. RECIST 1.1 supplies the response framework for solid tumors, while the registry specifies other hematological response criteria for hematologic tumors.
The most important statistical distinction is that a single-group response proportion describes what was observed among treated participants but does not by itself quantify a randomized treatment effect. Interpretation therefore depends on the response definitions, assessment window, analysis denominator, statistical precision, and the limitations inherent in comparing a single treatment group with any external benchmark. The registry reports serious adverse events for the ceritinib 750 mg group as 16/47, providing a separate descriptive safety measure.